Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual
Collection
Datasets used in the paper: Q2E: Query-to-Event Decomposition for Zero-Shot Multilingual Text-to-Video Retrieval • 37 items • Updated
query stringlengths 16 163 | prequel listlengths 0 15 | sequel listlengths 0 15 | during listlengths 0 15 | metadata dict | num_of_frames int64 32 32 | frame_captions listlengths 32 32 | video_id stringlengths 9 9 | frame2video_caption stringlengths 380 2.6k |
|---|---|---|---|---|---|---|---|---|
a person is connecting something to system | [
"A person holding a cable and approaching a computer to establish a connection to a system",
"Plugging a USB device into a port for connection to a system",
"Configuring network settings on a device for connection to a system",
"Connecting a power cord to a socket for connection to a system",
"A person typi... | [
"A person plugging a cable into a port for connection to a system",
"A person typing on a keyboard to configure the connection to a system",
"A device being attached to a network router during connection to a system",
"A person scanning a QR code to connect to a system",
"A person entering a password to aut... | [
"A person holding a cable for connection to a system",
"A person plugging a cable into a port for connection to a system",
"A person turning on a device to connect to a system",
"A person configuring settings on a screen during connection to a system",
"A person testing their connection to a system"
] | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a person is connecting something to system<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: To extract events from the query \"a person is connecting something to system\", it makes sense to consider the physical actions involved in connecting a device or object to a system, as these actions can be visualized in a video. The events should be specific and concrete, focusing on the physical interactions between the person and the system.\n\nEVENTS:\n1. A person holding a cable\n2. A person plugging a cable into a port\n3. A person turning on a device\n4. A person configuring settings on a screen\n5. A person testing the connection",
"en_article_excerpt": "a person is connecting something to system",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a person is connecting something to system<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: To extract events related to the query \"a person is connecting something to system\", it makes sense to consider common scenarios where someone is linking a device or component to a larger network or infrastructure, such as a computer system or a power grid. This could involve physical actions like plugging in cables or configuring settings, which can be visualized in a video.\n\nEVENTS:\n1. A person holding a cable and approaching a computer\n2. Plugging a USB device into a port\n3. Configuring network settings on a device\n4. Connecting a power cord to a socket\n5. A person typing IP address into a device to connect to a network",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a person is connecting something to system<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are directly related to the action of connecting something to a system, focusing on the physical or visual aspects of the connection process. This could include events that show the person interacting with the system or the object being connected, as these would be concrete and visualizable in a video.\n\nEVENTS:\n1. A person plugging a cable into a port\n2. A person typing on a keyboard to configure the connection\n3. A device being attached to a network router\n4. A person scanning a QR code to connect to a system\n5. A person entering a password to authenticate the connection",
"video_URL": "https://youtube.com/watch?v=video9770",
"video_description": null,
"video_id": "video9770"
} | 32 | [
"A person is plugging a cable into a desktop computer tower.",
"A person is plugging a cable into a desktop computer tower.",
"A person is connecting a component or cable to the side panel of the desktop computer tower.",
"The person is reaching towards the front of the desktop computer tower, possibly to adj... | video9770 | A person is assembling or reassembling a desktop computer tower, focusing on connecting cables, installing components, and securing the side panel to the chassis. The individual ensures all parts are properly attached and aligned, using tools such as a screwdriver to tighten screws and secure the panel in place. The process involves multiple steps of adjustment and tightening to guarantee a secure fit. Once the side panel is attached, the person checks the overall assembly, making any necessary adjustments before completing the task and finally adjusting the monitor to its proper position. |
a little girl does gymnastics | [
"A little girl runs towards a springboard during a gymnastics performance",
"little girl jumps over a hurdle during a gymnastics performance",
"little girl balances on a balance beam during a gymnastics performance",
"Little girl performs a cartwheel on a mat during a gymnastics performance",
"the little gi... | [
"Little girl performs a cartwheel during a gymnastics performance",
"She balances on one foot on a balance beam during a gymnastics performance",
"The girl executes a backflip on a spring floor during a gymnastics performance",
"She hangs from the uneven bars during a gymnastics performance",
"The little gi... | [
"little girl runs towards a springboard during a gymnastics performance",
"She jumps over a mini hurdle during a gymnastics performance",
"Girl performs a cartwheel on a mat during a gymnastics performance",
"She balances on a balance beam during a gymnastics performance",
"The little girl lands a dismount ... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little girl does gymnastics<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: To extract events from the query \"a little girl does gymnastics\", it makes sense to consider common gymnastics movements and routines that a young girl might perform, as these would be visually representative in a video. The extracted events should be specific and concrete enough to be captioned in a video frame, showcasing the girl's actions and movements.\n\nEVENTS:\n1. The little girl runs towards a springboard\n2. She jumps over a mini hurdle\n3. The girl performs a cartwheel on a mat\n4. She balances on a balance beam\n5. The little girl lands a dismount from a mini trampoline",
"en_article_excerpt": "a little girl does gymnastics",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little girl does gymnastics<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: To extract events related to the query \"a little girl does gymnastics\", it makes sense to consider common actions and movements typically performed in gymnastics, as well as the setting in which these activities usually take place. This will help to identify specific, visual events that could be captured in a video of a little girl participating in gymnastics.\n\nEVENTS:\n1. A little girl runs towards a springboard\n2. The little girl jumps over a hurdle\n3. The little girl balances on a balance beam\n4. The little girl performs a cartwheel on a mat\n5. The little girl lands a dismount from the uneven bars",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little girl does gymnastics<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: Given the query \"a little girl does gymnastics\", it makes sense to extract events that are typical movements or poses seen in gymnastics, as they could be visualized in a video and are relevant to the activity. The extracted events should be specific to a young girl participating in gymnastics, taking into account common gymnastics moves and routines.\n\nEVENTS:\n1. The little girl performs a cartwheel\n2. She balances on one foot on a balance beam\n3. The girl executes a backflip on a spring floor\n4. She hangs from the uneven bars\n5. The little girl lands a dismount from the vault",
"video_URL": "https://youtube.com/watch?v=video9771",
"video_description": null,
"video_id": "video9771"
} | 32 | [
"A young child is standing on a gym floor, wearing a sleeveless athletic outfit. The child has their hair styled in two buns and appears to be smiling. The background includes a red and black stripe on the wall and a blue mat or equipment to the right.",
"The young child, still wearing the sleeveless athletic out... | video9771 | A young child, initially wearing a sleeveless athletic outfit and later a purple leotard, is participating in a gymnastics routine in a gymnasium. The child begins by standing on the gym floor, displaying focus and determination, before transitioning into various dynamic poses and movements. They perform a series of jumps, leaps, and landings on a blue mat, showcasing their gymnastics skills. The child also runs, walks, and stands on the mat, transitioning between different elements of their routine. Throughout the video, the child maintains a focused and determined expression, indicating their engagement and concentration on the gymnastics activities. The setting remains consistent, with a red and black striped wall and various gym equipment visible in the background, emphasizing the athletic and competitive nature of the event. |
a woman creating a fondant baby and flower | [
"Kneading the fondant to soften it when creating a fondant baby",
"Kneading the fondant to soften it for creating a flower",
"How to shape fondant into a baby figure when creating a fondant baby",
"Shaping the fondant into a baby figure while creating a flower decoration",
"Coloring the fondant to add detai... | [
"Kneading and coloring the fondant for creating a fondant baby",
"Kneading and coloring the fondant for creating a flower",
"Shaping the fondant into a baby figure while creating a fondant baby",
"Shaping the fondant into a baby figure for cake decorating",
"Creating delicate flower petals from fondant for ... | [
"Kneading the fondant to soften it when creating a fondant baby",
"Kneading the fondant to soften it for creating a flower",
"Shaping the fondant into a baby figure while creating a fondant baby",
"Shaping the fondant into a baby figure while creating a flower decoration",
"Adding details to the fondant bab... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a woman creating a fondant baby and flower<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: To decompose the event of a woman creating a fondant baby and flower, it makes sense to extract specific steps or actions involved in the process that could be visually represented in a video, considering the typical techniques and materials used in fondant modeling. This will allow for a clear and concrete visualization of the event in a video.\n\nEVENTS:\n1. Kneading the fondant to soften it\n2. Shaping the fondant into a baby figure\n3. Adding details to the fondant baby's face\n4. Creating flower petals from colored fondant\n5. Assembling the fondant flowers into a decorative arrangement",
"en_article_excerpt": "a woman creating a fondant baby and flower",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a woman creating a fondant baby and flower<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: To extract events related to the query \"a woman creating a fondant baby and flower\", it makes sense to focus on the specific steps involved in the creative process of working with fondant, as these steps can be visually represented in a video. The events should be concrete and detailed enough to be captioned in a frame of a video showcasing the woman's creative process.\n\nEVENTS:\n1. Kneading the fondant to soften it\n2. Shaping the fondant into a baby figure\n3. Coloring the fondant to add details to the baby's face and body\n4. Creating flower petals from fondant and assembling them into a flower\n5. Placing the fondant flower in the baby's hand or near the baby figure",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a woman creating a fondant baby and flower<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events related to the process of creating a fondant baby and flower, as these are specific and visual tasks that can be captured in a video. The events should focus on the steps involved in shaping and decorating the fondant to create the desired designs.\n\nEVENTS:\n1. Kneading and coloring the fondant\n2. Shaping the fondant into a baby figure\n3. Creating delicate flower petals from fondant\n4. Assembling the flower and adding details\n5. Placing the fondant baby in a scene with the finished flower",
"video_URL": "https://youtube.com/watch?v=video7020",
"video_description": null,
"video_id": "video7020"
} | 32 | [
"A pair of hands is shaping a piece of dough or clay, using a tool to refine its form on a work surface.",
"A pair of hands is continuing to shape a piece of dough or clay, pressing and smoothing it with their fingers on a work surface.",
"The hands are now carefully placing a green dough piece, possibly a leaf... | video7020 | A person is creating a decorative figure using dough, starting with shaping a small dough figure resembling a turtle. They then add a green dough leaf to the figure, carefully positioning and smoothing it out to ensure it is securely attached and visually appealing. After completing the decoration of the turtle figure, the person moves on to work with yellow dough, flattening it into a circle and smoothing out its surface using a rolling pin. They then use a flower-shaped cutter to cut out decorative shapes from the yellow dough and refine the edges using scissors, aiming to create a precise and symmetrical floral design. Throughout the process, the person carefully handles the dough, making adjustments and refinements to achieve a polished and finished look for their creations. |
a boy plays grand theft auto 5 | [
"A boy sitting in front of a TV or computer screen playing a video game",
"The boy holding a gaming controller while playing a video game",
"The Grand Theft Auto 5 game loading screen appearing on the screen while playing a video game",
"The boy's character in the game walking or driving around the virtual ci... | [
"The boy's character drives a car when playing a video game",
"The boy's character shoots at enemies while playing a video game",
"The boy's character completes a mission while playing a video game",
"The boy's character interacts with non-playable characters (NPCs) while playing a video game",
"The boy's c... | [
"A boy sitting in front of a TV or computer screen playing a video game",
"The boy holding a gaming controller while playing a video game",
"What does the Grand Theft Auto 5 game interface look like when playing a video game",
"The boy's character driving a car while playing a video game",
"The boy's charac... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a boy plays grand theft auto 5<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: To extract events from the query \"a boy plays Grand Theft Auto 5\", it makes sense to consider the actions and scenes that are likely to occur when someone is playing this video game, taking into account the game's content and typical gameplay. The extracted events should be specific and visual, allowing them to be easily captioned in a video frame.\n\nEVENTS:\n1. A boy sitting in front of a TV or computer screen\n2. The boy holding a gaming controller\n3. The Grand Theft Auto 5 game interface displayed on the screen\n4. The boy's character in the game driving a car\n5. The boy's character in the game engaging in a shootout with non-player characters",
"en_article_excerpt": "a boy plays grand theft auto 5",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a boy plays grand theft auto 5<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: It makes sense to extract events that are directly related to the action of playing a video game, specifically Grand Theft Auto 5, as these events can be visually represented in a video. The extracted events should be concrete and specific to the game and the setting of a boy playing it.\n\nEVENTS:\n1. A boy sitting in front of a TV or computer screen\n2. The boy holding a gaming controller\n3. The Grand Theft Auto 5 game loading screen appearing on the screen\n4. The boy's character in the game walking or driving around the virtual city\n5. The boy reacting to gameplay, such as jumping back in surprise or smiling at a successful mission completion",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a boy plays grand theft auto 5<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are directly related to the gameplay of Grand Theft Auto 5, as the query specifies that the boy is playing the game. The extracted events should be concrete and visual, such as actions performed by the player character or interactions within the game environment.\n\nEVENTS:\n1. The boy's character drives a car in the game\n2. The boy's character shoots at enemies\n3. The boy's character completes a mission\n4. The boy's character interacts with non-playable characters (NPCs) in the game\n5. The boy's character crashes a car in the game",
"video_URL": "https://youtube.com/watch?v=video9773",
"video_description": null,
"video_id": "video9773"
} | 32 | [
"The image depicts a police motorcycle chase on a city street. Multiple police officers on motorcycles are pursuing another motorcycle. The scene takes place in an urban environment with tall buildings lining the street. There is a police car also visible in the chase. The road appears wet, suggesting recent rain."... | video9773 | The video depicts two distinct scenes. The first part shows a high-speed police motorcycle chase on a wet city street, involving multiple police officers on motorcycles and a police car pursuing another motorcycle. The scene is set in an urban environment with tall buildings.
The video then transitions to a video game scenario where a green sports car is driven on various roads, initially in a rural setting and later in a cityscape. The car performs a significant jump from a ramp and soars through the air, flying over a set of bleachers and the cityscape, before landing back on the road and continuing at high speed through an urban environment. The car's journey is dynamic, capturing a sense of motion and adventure as it navigates through different landscapes. |
a man is giving a review on a vehicle | [
"The man walking around the vehicle to inspect its exterior during a review of a vehicle",
"The man sitting in the driver's seat and starting the engine during a review of a vehicle",
"The man test driving the vehicle on a road or test track during a review of a vehicle",
"The man pointing out specific featur... | [
"The man walking around the vehicle to inspect it during a review of a vehicle",
"The man sitting in the driver's seat and starting the engine during a review of a vehicle",
"The man driving the vehicle to test its performance during a review of a vehicle",
"The man pointing out specific features of the vehic... | [
"A man standing next to a vehicle during a review of a vehicle",
"The man walking around the vehicle to inspect it during a review of a vehicle",
"The man sitting in the driver's seat of the vehicle during a review",
"The man gesturing to the vehicle's features while speaking during a review of a vehicle",
... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man is giving a review on a vehicle<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: It makes sense to extract events that are visually related to a person providing a review of a vehicle, such as interacting with the vehicle or expressing their opinions. These events should be concrete and easily visualizable in a video, focusing on the reviewer's actions and the vehicle itself.\n\nEVENTS:\n1. A man standing next to a vehicle\n2. The man walking around the vehicle to inspect it\n3. The man sitting in the driver's seat of the vehicle\n4. The man gesturing to the vehicle's features while speaking\n5. The man test-driving the vehicle",
"en_article_excerpt": "a man is giving a review on a vehicle",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man is giving a review on a vehicle<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: To extract events related to the query \"a man is giving a review on a vehicle\", it makes sense to consider actions that are typically involved in reviewing a vehicle, such as inspecting the vehicle's features, test driving, and sharing opinions. These events should be concrete and visual, allowing them to be reasonably captioned in a video frame associated with the query.\n\nEVENTS:\n1. The man walking around the vehicle to inspect its exterior\n2. The man sitting in the driver's seat and starting the engine\n3. The man test driving the vehicle on a road or test track\n4. The man pointing out specific features of the vehicle's interior\n5. The man speaking directly to the camera to share his review and opinions",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man is giving a review on a vehicle<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are directly related to the action of reviewing a vehicle, such as the man inspecting the vehicle, driving it, or pointing out its features. These events are likely to be visual and concrete, making them suitable for a video caption.\n\nEVENTS:\n1. The man walking around the vehicle to inspect it\n2. The man sitting in the driver's seat and starting the engine\n3. The man driving the vehicle to test its performance\n4. The man pointing out specific features of the vehicle, such as the infotainment system or safety features\n5. The man checking the vehicle's trunk space or cargo area",
"video_URL": "https://youtube.com/watch?v=video7026",
"video_description": null,
"video_id": "video7026"
} | 32 | [
"A person is off-roading or driving in a rugged, rocky terrain. The vehicle is on a rocky surface surrounded by large boulders and sparse vegetation, indicating a desert or rocky environment.",
"The vehicle continues its off-roading journey through the rugged, rocky terrain, navigating around large boulders and s... | video7026 | The video depicts a person driving a vehicle through a rugged, rocky terrain, likely in a desert or mountainous area, at a slow speed. The vehicle navigates around large boulders and sparse vegetation, with the driver occasionally checking the dashboard, which displays information such as speed and gear. The speed remains relatively slow, around 1-2 MPH, with occasional increases and decreases. The vehicle's descent control system is available, and the dashboard displays various messages, including "Descent available" and "Speed unavailable," indicating potential issues with the speedometer.
After driving through the rocky terrain, the scene shifts to the person inside the vehicle, parked or moving slowly through a forested or wooded area. The driver is engaged in a conversation, gesturing with their hands while seated in the driver's seat and occasionally holding the steering wheel. The vehicle's interior remains consistent throughout, with a black dashboard, center console, and seats. A black thermos cup is visible on the center console during the conversation. The person continues to talk and gesture, indicating an ongoing engagement in the conversation, as the vehicle remains stationary or moves slowly through the wooded area. |
a man speaks to children in a classroom | [
"A man writes on a blackboard while teaching",
"A man writes on a blackboard during a public speaking engagement",
"Children raising their hands to answer a question during teaching",
"Children raising their hands to answer a question during public speaking",
"The man hands out worksheets to the students du... | [
"The man writes on the blackboard while teaching",
"The man writes on the blackboard during a public speaking engagement",
"Children taking notes during teaching",
"Children taking notes during public speaking events",
"The man distributes handouts to the students during a teaching session",
"The man dist... | [
"A man stands at the front of the classroom teaching and writes on the blackboard",
"A man stands at the front of the classroom and writes on the blackboard while engaged in public speaking",
"Children look up at the man with interest and take notes during a teaching session",
"Children look up at the man wit... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man speaks to children in a classroom<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: It makes sense to extract events that are typical of a classroom setting and involve interaction between a teacher and students, as these are common and visual elements that could be captured in a video. The events should focus on the man's actions and the children's reactions or responses to create a clear and concrete visualization.\n\nEVENTS:\n1. A man stands at the front of the classroom and writes on the blackboard\n2. Children look up at the man with interest and take notes\n3. The man gestures as he explains a concept to the class\n4. Children raise their hands to ask questions\n5. The man walks around the classroom to assist individual students",
"en_article_excerpt": "a man speaks to children in a classroom",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man speaks to children in a classroom<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: It makes sense to extract events that are typical in a classroom setting and involve interaction between a teacher and students, as these events are likely to be visualized in a video associated with the query. The extracted events should be concrete and specific, such as actions the man or children might be doing, to serve as reasonable captions for a frame of a video.\n\nEVENTS:\n1. A man writes on a blackboard\n2. Children raise their hands to answer a question\n3. The man hands out worksheets to the students\n4. Children take notes while the man lectures\n5. The man helps a student with a problem on their desk",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a man speaks to children in a classroom<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are typical interactions or activities that occur in a classroom setting, considering the context of a man speaking to children. These events should be concrete and visual, allowing them to be easily depicted in a video.\n\nEVENTS:\n1. The man writes on the blackboard\n2. Children take notes\n3. The man distributes handouts to the students\n4. Students ask the man questions\n5. The man uses visual aids to explain a concept",
"video_URL": "https://youtube.com/watch?v=video9775",
"video_description": null,
"video_id": "video9775"
} | 32 | [
"In the image, a group of five children is sitting in a classroom setting. They appear to be engaged in an activity or listening to someone off-camera. The children are seated on shelves filled with books and storage bins. The classroom is decorated with educational posters and charts on the walls.",
"The childre... | video9775 | A group of children, approximately five in number, are engaged in a learning session in a classroom setting. The session involves active participation and discussion, led by a man who interacts with the children, explaining educational topics and facilitating their engagement. The children, including a girl in a pink top and a boy named Alton, show interest and enthusiasm, asking questions and sharing ideas. The classroom environment is filled with educational materials, posters, and a calendar, indicating a structured learning environment. Throughout the session, Alton is particularly active, standing up to speak and share his thoughts, demonstrating his engagement and participation in the discussion. The overall context suggests a supportive and interactive learning environment where children are encouraged to express themselves and learn from each other and the adult facilitator. |
one micky mouse is talking to other | [
"Mickey Mouse walking up to a group of friends in a park and starting a conversation with them",
"Mickey Mouse shaking hands with Goofy while talking to him and other characters",
"Mickey Mouse having a conversation with Minnie Mouse while talking to others",
"Mickey Mouse laughing with Pluto and Donald Duck ... | [
"Mickey Mouse shaking hands with another character while talking to others",
"Mickey Mouse laughing with his friends while talking to them",
"Mickey Mouse having a conversation with Minnie Mouse while talking to others",
"Mickey Mouse gesturing to a group of characters while talking to them",
"Mickey Mouse ... | [
"Mickey Mouse standing in front of a colorful background and speaking while talking to others",
"Other Disney characters gathering around Mickey Mouse as he talks to them",
"Mickey Mouse gesturing with his hands while talking to his friends",
"Other characters responding or reacting to what Mickey Mouse is sa... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: one micky mouse is talking to other<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: Given the query \"one Mickey Mouse is talking to other\", it makes sense to extract events that are typical of a scene where Mickey Mouse is interacting with others, likely in a cartoon or theme park setting. The events should be concrete and visual, such as Mickey Mouse engaging in conversation or interacting with his friends.\n\nEVENTS:\n1. Mickey Mouse standing in front of a colorful background and speaking\n2. Other Disney characters gathering around Mickey Mouse to listen\n3. Mickey Mouse gesturing with his hands while talking to his friends\n4. Other characters responding or reacting to what Mickey Mouse is saying\n5. Mickey Mouse laughing or smiling while interacting with the other characters",
"en_article_excerpt": "one micky mouse is talking to other",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: one micky mouse is talking to other<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: It makes sense to extract events that involve Mickey Mouse interacting with other characters, as this is a common scenario in cartoons and animated videos. Given the context of Mickey Mouse, it's likely that the events will be lighthearted and family-friendly, and may involve other well-known Disney characters.\n\nEVENTS:\n1. Mickey Mouse walking up to a group of friends in a park\n2. Mickey Mouse shaking hands with Goofy\n3. Mickey Mouse having a conversation with Minnie Mouse\n4. Mickey Mouse laughing with Pluto and Donald Duck\n5. Mickey Mouse waving at a crowd of Disney characters",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: one micky mouse is talking to other<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are typical interactions or scenarios involving Mickey Mouse, given his character and common portrayals in media. The events should be concrete and visual, considering what could be reasonably depicted in a video featuring Mickey Mouse conversing with others.\n\nEVENTS:\n1. Mickey Mouse shaking hands with another character\n2. Mickey Mouse laughing with his friends\n3. Mickey Mouse having a conversation with Minnie Mouse\n4. Mickey Mouse gesturing to a group of characters\n5. Mickey Mouse listening to Pluto bark in response",
"video_URL": "https://youtube.com/watch?v=video9776",
"video_description": null,
"video_id": "video9776"
} | 32 | [
"The image shows an animated character resembling a cow, wearing a yellow hat and an orange cape. The background is colorful, featuring a pink and purple pattern. There is no visible text or additional context in the image.",
"The animated character, still resembling a cow, is now seen smiling and winking, mainta... | video9776 | The video features an animated cow character wearing a yellow hat and an orange cape, set against a vibrant and colorful background with pink and purple patterns. The character's expressions and actions change throughout the video, ranging from smiling and winking to looking surprised, curious, and determined. At one point, the character appears to be engaged in a task or interaction, then transitions to a moment of pause or reflection. The setting later includes a windmill in the background, adding to the whimsical atmosphere. The character also holds a wand with a star on top at some points, further enhancing the playful and magical tone. Throughout the video, the cow character's expressions and actions convey a sense of joy, wonder, and engagement, creating a lively and cheerful atmosphere. |
a naked child runs through a field | [
"A child takes off their clothes while running through a field",
"A child plays outside without clothes on a warm day while running through a field",
"child runs away from lake or pool without clothes",
"a child being chased while running through a field",
"A child playing tag or catch while running through... | [
"A child running through a field of tall grass",
"A child laughing and spinning while running through a field",
"A parent or guardian chasing after a child running through a field",
"A child running through a field playing with flowers or wild plants",
"A child running through a field and eventually stoppin... | [
"A child runs through a field without clothes",
"The child's legs and arms move as they run through a field",
"A child's hair flowing in the wind as they run through a field",
"A child laughs or smiles while running through the field",
"A child running through a field with their bare feet touching the grass... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a naked child runs through a field<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: Given the query \"a naked child runs through a field\", it makes sense to extract events that are directly related to the action of a child running in a field without clothes, considering typical scenes that could be visualized in a video. The extracted events should focus on the child's movement and interaction with the field environment.\n\nEVENTS:\n1. A child runs through a field without clothes\n2. The child's legs and arms move as they run\n3. The child's hair flows in the wind as they run\n4. The child laughs or smiles while running through the field\n5. The child's bare feet touch the grass or dirt in the field",
"en_article_excerpt": "a naked child runs through a field",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a naked child runs through a field<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: Given the query \"a naked child runs through a field\", it makes sense to extract events that could lead to this scene, such as the child removing their clothes or being in a situation where they would not be wearing clothes, and then running through the field. These events should be concrete and visual, allowing them to be reasonably captioned in a video frame.\n\nEVENTS:\n1. A child takes off their clothes near a field\n2. A child plays outside without clothes on a warm day\n3. A child runs away from a body of water, such as a lake or pool, without clothes\n4. A child is chased by someone or something while running through the field\n5. A child is playing tag or a game of catch in the field without clothes",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a naked child runs through a field<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: Given the query \"a naked child runs through a field\", it makes sense to extract events that are visually plausible and related to the action of a child running in a field without clothes, considering typical scenarios and environments where such an event might occur. The extracted events should be concrete and capable of being depicted in a video, taking into account common aspects of a field setting and the implications of a child running naked.\n\nEVENTS:\n1. A child running through tall grass\n2. The child laughing and spinning in the field\n3. A parent or guardian chasing after the child\n4. The child playing with flowers or wild plants in the field\n5. The child eventually stopping to look back or play with an object found in the field",
"video_URL": "https://youtube.com/watch?v=video7025",
"video_description": null,
"video_id": "video7025"
} | 32 | [
"A young child is running across a grassy area near a body of water, with several people lounging and sitting in the background. The setting appears to be a beach or park area.",
"The young child continues to run across the grassy area near the body of water. The child's motion suggests they are engaged in play, ... | video7025 | The video depicts two distinct scenes. The first part of the video shows a young child playing and running in a grassy area near a body of water, likely a beach or park, on a sunny day. Several people are lounging and sitting in the background, enjoying the outdoors under colorful umbrellas. As the scene progresses, a woman is seen jogging and then standing, and the child continues to play. The atmosphere is leisurely and relaxed, with people enjoying the sunny day.
The scene then shifts to a beach setting where a larger group of people are gathered, sitting, lying, and standing along the shoreline. The atmosphere remains relaxed, with people enjoying the beach and the calm water.
The video then abruptly changes to a black and white portrait of a young man with a serious expression, looking directly at the camera. He is dressed in formal or uniform-like attire, suggesting a context of significance or ceremony. This portrait continues for an extended period, with no changes in the individual's expression or the camera's focus, maintaining a historical or archival feel. The contrast between the lively, outdoor beach scenes and the solemn, black and white portrait is striking, suggesting a narrative that may be exploring different themes or periods. |
a little boy singing in front of judges and crowd | [
"A little boy standing on a stage with a microphone during a singing performance",
"A little boy standing on a stage with a microphone at a talent show or competition",
"Judges seated at a table with scorecards and pens during a singing performance",
"Judges seated at a table with scorecards and pens at a tal... | [
"The little boy stepping up to the microphone for a singing performance",
"The little boy stepping up to the microphone at a talent show or competition",
"Judges taking notes and giving feedback during a singing performance",
"Judges taking notes and giving feedback at a talent show or competition",
"the cr... | [
"A little boy stepping up to a microphone for a singing performance",
"a little boy stepping up to a microphone at a talent show or competition",
"The boy beginning to sing in front of the crowd during a singing performance",
"the boy beginning to sing in front of the crowd at a talent show or competition",
... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little boy singing in front of judges and crowd<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: This query appears to be related to a talent show or competition setting, where a young boy is performing in front of an audience and a panel of judges, making it reasonable to extract events that are typical of such a scenario. The events extracted should be concrete and visual, allowing them to be easily captioned in a video frame.\n\nEVENTS:\n1. A little boy stepping up to a microphone\n2. The boy beginning to sing in front of the crowd\n3. Judges looking at the boy with expressions of interest or surprise\n4. The audience applauding or cheering the boy on\n5. The boy finishing his song and bowing to the audience",
"en_article_excerpt": "a little boy singing in front of judges and crowd",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little boy singing in front of judges and crowd<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: It makes sense to extract events related to a talent show or competition setting, as the presence of judges and a crowd suggests a performance-based scenario. The events should be concrete and visual, allowing for a clear mental image of what the scene might look like in a video.\n\nEVENTS:\n1. A little boy standing on a stage with a microphone\n2. Judges seated at a table with scorecards and pens\n3. A crowd of people cheering and applauding in the audience\n4. The little boy taking a deep breath before starting to sing\n5. Confetti and balloons falling from the ceiling after the performance",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: a little boy singing in front of judges and crowd<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: It makes sense to extract events that are typical of a talent show or competition setting, where a child is performing in front of an audience and judges, and the events should be concrete and visual enough to be captured in a video. The extracted events should reflect the emotions, actions, and reactions of the participants and the audience in this setting.\n\nEVENTS:\n1. The little boy stepping up to the microphone\n2. Judges taking notes and giving feedback\n3. The crowd cheering and applauding\n4. The little boy singing with passion and emotion\n5. Confetti falling or trophies being awarded after the performance",
"video_URL": "https://youtube.com/watch?v=video9778",
"video_description": null,
"video_id": "video9778"
} | 32 | [
"A woman with curly hair is sitting in a red chair, smiling and laughing, seemingly reacting to something happening off-camera. She is wearing a black outfit and is seated in what appears to be a television studio setting.",
"The woman continues to sit in the red chair, still smiling and laughing, maintaining her... | video9778 | A woman appears to be a guest or participant on a television show, likely "The X Factor," given the setting and the presence of a judge and audience. Initially, she is seated and then stands at a podium, engaging with the audience and possibly responding to questions or participating in a discussion. Her interaction with the audience and her body language suggest a dynamic and engaging presentation or conversation.
The scene then shifts to a young boy performing on stage, singing in front of an "X Factor" backdrop. He is dressed casually and maintains a confident demeanor throughout his performance. A judge, wearing a "GUILTY" T-shirt, is shown clapping and reacting positively to the boy's singing, indicating his approval and enjoyment of the performance.
The video depicts an interactive and engaging atmosphere, with participants and judges showing enthusiasm and support for the performances. The setting of a television studio with a live audience and the presence of judges suggest that the event is a competition or talent show, where contestants perform and receive feedback from the judges and audience. Overall, the video captures the energy and excitement of a live performance event, highlighting the interactions between participants, judges, and the audience. |
fireworks are being lit and exploding in a night sky | [
"Fireworks being set up on the ground or on a fireworks launch pad at night for a fireworks display",
"People gathering at night to watch the fireworks display",
"A fuse being lit, igniting a firework during a nighttime fireworks display",
"Fireworks exploding in the night sky with bursts of color and light d... | [
"Fireworks display at night with explosions into colorful patterns in the sky",
"Sparks and embers falling from the explosions during a nighttime fireworks display",
"People watching the nighttime fireworks display with reactions of amazement and delight",
"Fireworks rockets shooting up into the air before ex... | [
"Fireworks shooting up into the night sky during a fireworks display at night",
"Explosions of colorful lights and patterns in the sky during a nighttime fireworks display",
"Sparks and embers falling back down to the ground during a nighttime fireworks display",
"People watching and reacting to the nighttime... | {
"article_url": null,
"during_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to decompose the ongoing event in multiple simple events. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: fireworks are being lit and exploding in a night sky<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"during_raw_generation": "EXPLANATION: Given the query, it makes sense to extract events that describe the visual and dynamic aspects of fireworks exploding in the night sky, as these events can be easily visualized in a video. The extracted events should focus on the actions and movements of the fireworks, as well as the surrounding environment.\n\nEVENTS:\n1. Fireworks shooting up into the night sky\n2. Explosions of colorful lights and patterns in the sky\n3. Sparks and embers falling back down to the ground\n4. People watching and reacting to the fireworks display\n5. The night sky being illuminated by the colorful explosions",
"en_article_excerpt": "fireworks are being lit and exploding in a night sky",
"en_article_url": null,
"event_category": null,
"event_name": null,
"language": null,
"prequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what previous events could lead to this given event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: fireworks are being lit and exploding in a night sky<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"prequel_raw_generation": "EXPLANATION: It makes sense to extract events that are directly related to the preparation and display of fireworks, as these are the most likely to be visualized in a video of fireworks being lit and exploding in a night sky. Events that involve people setting up or reacting to the fireworks could also be relevant, as they provide context and add to the visual narrative of the video.\n\nEVENTS:\n1. Fireworks being set up on the ground or on a fireworks launch pad\n2. People gathering to watch the fireworks display\n3. A fuse being lit, igniting a firework\n4. Fireworks exploding in the night sky, with bursts of color and light\n5. Sparks and embers falling to the ground after a firework has exploded",
"sequel_prompt": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nCutting Knowledge Date: December 2023\nToday Date: 26 Jul 2024\n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nYou are given a video search query. Assume the video search query is an ongoing event. You have to find out what events can be the outcome of this ongoing event. The extracted events should be concrete enough to be visualized in a video.\n\nHere are some guidelines: \n- Extract events that could be visualized in a video.\n- Before extracting an event, consider whether the event text could serve as a reasonable caption for a frame of a video associated with the query.\n- Give specific events related to the query, use your previous WORLD KNOWLEDGE to extract events that are likely to be seen in a video. For example, if the query is \"2019 Christmas Eve Protests in Hong Kong\", you might return \"Protesters marching in the streets\", but not \"Protestors marching in snow\" because it is not likely to have snow in Hong Kong, though its pretty common in other places. \n- First, explain your reasoning in 1-2 sentences about which events it makes sense to extract. Then, provide a numbered list with the events (anywhere from one to five event; use your best judgement to determine how many to extract)\n- Structure your response into two sections, \"\"EXPLANATION:\"\" and then \"\"EVENTS:\"\". The events sections should contain nothing but the events. \n\nHere is the query I'd like you to extract events from: fireworks are being lit and exploding in a night sky<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
"sequel_raw_generation": "EXPLANATION: Given the query \"fireworks are being lit and exploding in a night sky\", it makes sense to extract events that are directly related to the visual and auditory experience of a fireworks display, as these events can be easily visualized in a video. The extracted events should focus on the fireworks themselves, the surrounding environment, and the potential reactions of people watching the display.\n\nEVENTS:\n1. Fireworks exploding into colorful patterns in the night sky\n2. Sparks and embers falling from the explosions\n3. People watching the fireworks display with reactions of amazement and delight\n4. Fireworks rockets shooting up into the air before exploding\n5. The night sky being illuminated by the bright colors of the fireworks",
"video_URL": "https://youtube.com/watch?v=video9779",
"video_description": null,
"video_id": "video9779"
} | 32 | [
"The image shows a night sky with fireworks exploding. There are three bright bursts of light and one trail of sparks moving upwards. A section of a boat with flags is visible in the lower right corner.",
"The image captures a night sky filled with fireworks, showing two bursts of light and one trail of sparks mo... | video9779 | A fireworks display takes place over a body of water, with a boat adorned with colorful flags visible in the foreground. The video depicts a continuous and vibrant display of fireworks, with bursts of light and trails of sparks illuminating the night sky. The maritime setting and the festive atmosphere created by the fireworks suggest a celebratory event, such as a national holiday, festival, or special occasion. The video showcases a dynamic and visually engaging display of fireworks, with a variety of colors and patterns, maintaining a consistent and captivating visual theme throughout. The presence of the boat with colorful flags reinforces the festive and celebratory mood, creating a sense of excitement and joy. Overall, the video captures a lively and entertaining fireworks display in a maritime setting, conveying a sense of celebration and festivity. |
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 16
Size of downloaded dataset files:
5.34 MB
Size of the auto-converted Parquet files:
5.34 MB
Number of rows:
1,000