遇见数据集

Data for paper "Characterizing AI Manipulation Risks in Brazilian YouTube Climate Discourse"

收藏
Zenodo2025-11-11 更新2026-05-26 收录
官方服务:

资源简介:

DATA DESCRIPTIONThis dataset supports the study of climate discourse in Portuguese language YouTube videos, focusing on Brazilian Portuguese content. The data is split into three files: youtube_video.parquet: containing video metadata, and channel, topic and persuasion annotations, for a total of 226,775 videos. youtube_comments.parquet: containing comment metadata, hashed user ids to preserve privacy, and theory of mind annotations for a total of 2,756,165 comments. aggregated_youtube.parquet: containing the aggregated data for each of the 226,775 videos, such as the percentage of comments presenting each of the theory of mind categories. The dataset can be rehydrated through the id field for the youtube_comments file and videoId field for youtube_video and aggregated_youtube files. We remove the textual content from all data to prevent leakage of personally identifiable information. DATA ACCESS INSTRUCTIONS Access to this data is restricted. To request access, follow the steps below: Login to zenodo or enter your details (email and name) in the Files section. In the message box, describe the intended usage of the data. Request access. Your request will be reviewed within a week. When access is granted, you will receive a notification via email. If you have any questions, contact us at: wenchao[dot]dong[at]mpi-sp[dot]org or marcelo[dot]sartori-locatelli[at]mpi-sp[dot]org ETHICAL USAGE COMMITMENT By requesting and accessing this dataset, you agree to comply with the following ethical usage requirements: The dataset may only be used for research and educational purposes, such as the study of online discourse and social media analysis. The data must not be used for political, or malicious purposes. The dataset may not be used for harmful applications, including, but not limited to: training or fine-tuning generative models to amplify, reproduce or legitimize harmful content (e.g. misinformation, denialist narratives, conspiracies). You will take appropriate measures to avoid unauthorized access, data leakage or other actions that may infringe on the confidentiality of the dataset. ADDITIONAL INFORMATION For the full version of the paper check the arxiv. DATASET STRUCTUREyoutube_video.parquet Field Description videoId A unique identifier for the video in the YouTube platform query Query used to retrieve the video using the YouTube api channelId A unique identifier for the creator of the video in the YouTube platform viewCount Number of views associated with the entry likeCount Number of likes associated with the entry commentCount Number of comments associated with the entry tags YouTube given tags associated with the entry duration Duration of the video in ISO8601 format video_duration_minutes Duration of the video in minutes channel_subscriber_count Number of subscribers for the creator of the video channel_subscriberCount_log Number of subscribers for the creator of the video (log scale) logical_appeal Indicates the presence of logical appeal in the vide's content emotional_appeal Indicates the presence of emotional appeal in the video's content statistical_evidence Indicates the presence of statistical evidence in the video's content social_norm Indicates the presence of social norm in the video's content authority Indicates the presence of authority in the video's content personal_stories Indicates the presence of personal stories in the video's content moral_appeal Indicates the presence of moral appeal in the video's content reciprocity Indicates the presence of reciprocity in the video's content scarcity Indicates the presence of scarcity in the video's content common_ground Indicates the presence of common ground in the video's content video_like_ratio likeCount divided by viewCount for the entry video_comment_ratio commentCount divided by viewCount for the entry channel_classification Category associated with the video's creator (e.g. Government, individual creator) channel_idx Index for the category associated with the video's creator topic Topic associated with the video's contents topic_label Index for the topic associated with the video's contents publishedAt Creation date of the entry in ISO8601 format publish_year Year of creation for the entry publish_month Month of creation for the entry publish_day Day of creation for the entry publish_time Time of creation for the entry (hh:mm:ss) publishedAt_datetime Creation date of the entry in UTC format youtube_comments.parquet: Field Description id A unique identifier for the comment in the YouTube platform hashed_userId A pseudonymized user id associated with the entry videoId A unique identifier for the video associated with the entry in the YouTube platform comment_like_count Number of likes associated with the entry comment_reply_count Number of replies associated with the entry beliefs_exists Indicates the presence of belief theory of mind label in the video's content intentions_exists Indicates the presence of intention theory of mind label in the video's content desires_exists Indicates the presence of desires theory of mind label in the video's content emotions_exists Indicates the presence of emotions theory of mind label in the video's content knowledge_exists Indicates the presence of knowledge theory of mind label in the video's content percepts_exists Indicates the presence of percepts theory of mind label in the video's content non_literal_exists Indicates the presence of non-literal communication theory of mind label in the video's content comment_length Length of the content associated with the entry in words comment_delay_days Number of days between the posting of this entry and the video associated with it comment_publish_year Year of creation for the entry comment_publish_month Month of creation for the entry comment_publish_day Day of creation for the entry channel_subscriberCount_log Number of subscribers for the creator of the video associated with the entry (log scale) logical_appeal, emotional_appeal, statistical_evidence, social_norm, authority, personal_stories, moral_appeal, reciprocity, scarcity, common_ground Persuasion strategies employed in the video associated with this entry, matching the ones in file youtube_video.parquet for the same videoId aggregated_youtube.parquet: Field Description videoId A unique identifier for the video in the YouTube platform comment_like_count Average number of likes of comments associated with this entry comment_reply_count Average number of replies of comments associated with this entry video_duration_minutes Duration of the video in minutes channel_subscriberCount_log Number of subscribers for the creator of the video (log scale) comment_length Average length (in words) of comments associated with this entry beliefs_exists Proportion of comments containing belief theory of mind label associated with this entry intentions_exists Proportion of comments containing intentions theory of mind label associated with this entry desires_exists Proportion of comments containing desires theory of mind label associated with this entry emotions_exists Proportion of comments containing emotions theory of mind label associated with this entry knowledge_exists Proportion of comments containing knowledge theory of mind label associated with this entry percepts_exists Proportion of comments containing percepts theory of mind label associated with this entry non_literal_exists Proportion of comments containing non-literal communication theory of mind label associated with this entry logical_appeal Indicates the presence of logical appeal in the vide's content emotional_appeal Indicates the presence of emotional appeal in the video's content statistical_evidence Indicates the presence of statistical evidence in the video's content social_norm Indicates the presence of social norm in the video's content authority Indicates the presence of authority in the video's content personal_stories Indicates the presence of personal stories in the video's content moral_appeal Indicates the presence of moral appeal in the video's content reciprocity Indicates the presence of reciprocity in the video's content scarcity Indicates the presence of scarcity in the video's content common_ground Indicates the presence of common ground in the video's content

提供机构:
Zenodo
创建时间:
2025-11-07
二维码
社区交流群
二维码
科研交流群
商业服务