| _version_ | 1866902226176311296 |
|---|---|
| author | Mohammad I. Fahmi Nababan, Adli Abdillah |
| author_facet | Mohammad I. Fahmi Nababan, Adli Abdillah |
| contents | <p>This dataset contains 10,699 English-language tweets collected from Twitter (X) using keywords related to mental health: "depression", "anxiety", "stress", and "mental health". The data was gathered from January to May 2025 using the Twitter API. Each entry includes metadata such as tweet text, timestamps, engagement metrics (likes, replies, retweets), and media links (if available).</p> <p>The dataset is intended for use in Natural Language Processing research on sentiment classification, especially in the domain of mental health detection using transformer models like BERT and RoBERTa.</p> <p>Missing data is present in fields such as <code>location</code> and <code>username</code>, which have been anonymized or stripped for ethical and privacy compliance.</p> <h3><strong>Dataset Overview</strong></h3> <ul> <li> <p><strong>Total Rows (Entries)</strong>: 10,699</p> </li> <li> <p><strong>Total Columns</strong>: 15</p> </li> <li> <p><strong>Language Focus</strong>: Mostly English (<code>lang = en</code> in 10,687 out of 10,699 rows)</p> </li> <li> <p><strong>Content</strong>: The dataset contains tweets related to <strong>mental health topics</strong> such as anxiety, depression, and stress, collected from <strong>Twitter (X)</strong> between <strong>January and May 2025</strong>.</p> </li> </ul> <h3><strong>Main Columns & Descriptions</strong></h3> <div> <div> <table> <thead> <tr> <th>Column Name</th> <th>Description</th> </tr> </thead> <tbody> <tr> <td><code>conversation_id_str</code></td> <td>Unique conversation/thread identifier</td> </tr> <tr> <td><code>created_at</code></td> <td>Timestamp of the tweet</td> </tr> <tr> <td><code>favorite_count</code></td> <td>Number of likes/favorites</td> </tr> <tr> <td><code>full_text</code></td> <td>The actual tweet text content</td> </tr> <tr> <td><code>id_str</code></td> <td>Tweet ID</td> </tr> <tr> <td><code>image_url</code></td> <td>URL to any attached image (if any)</td> </tr> <tr> <td><code>in_reply_to_screen_name</code></td> <td>Username if the tweet is a reply</td> </tr> <tr> <td><code>lang</code></td> <td>Language of the tweet (mostly <code>en</code>)</td> </tr> <tr> <td><code>location</code></td> <td>Tweet author's location (all <code>null</code> in this dataset)</td> </tr> <tr> <td><code>quote_count</code></td> <td>Number of times the tweet was quoted</td> </tr> <tr> <td><code>reply_count</code></td> <td>Number of replies</td> </tr> <tr> <td><code>retweet_count</code></td> <td>Number of retweets</td> </tr> <tr> <td><code>tweet_url</code></td> <td>Direct URL to the tweet</td> </tr> <tr> <td><code>user_id_str</code></td> <td>ID of the user who posted the tweet</td> </tr> <tr> <td><code>username</code></td> <td>Username of the author (all <code>null</code> in this dataset)</td> </tr> </tbody> </table> </div> </div> <h3><strong>Data Highlights</strong></h3> <ul> <li> <p><strong>Tweets with images</strong>: 2,999 rows contain a non-null <code>image_url</code></p> </li> <li> <p><strong>Replies</strong>: About 4,725 tweets are replies (<code>in_reply_to_screen_name</code> not null)</p> </li> <li> <p><strong>Engagement</strong>:</p> <ul> <li> <p><code>favorite_count</code>: ranges from 0 to 34,364</p> </li> <li> <p><code>retweet_count</code>: ranges from 0 to 1,958</p> </li> <li> <p>Most tweets have low engagement (median values are 0)</p> </li> </ul> </li> <li> <p><strong>Top mentioned account</strong>: <code>elonmusk</code> (mentioned in replies 61 times)</p> </li> <li> <p><strong>Missing Data</strong>:</p> <ul> <li> <p><code>location</code> and <code>username</code>: completely null</p> </li> <li> <p><code>image_url</code>: ~72% of tweets don’t include images</p> </li> </ul> </li> </ul> <p> </p> <p> </p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17144715 |
| institution | Zenodo |
| language | eng |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Mental Health Twitter Dataset for Sentiment Analysis (Jan–May 2025) Mohammad I. Fahmi Nababan, Adli Abdillah Mental Health NLP Sentiment Analysis BERT RoBERTa <p>This dataset contains 10,699 English-language tweets collected from Twitter (X) using keywords related to mental health: "depression", "anxiety", "stress", and "mental health". The data was gathered from January to May 2025 using the Twitter API. Each entry includes metadata such as tweet text, timestamps, engagement metrics (likes, replies, retweets), and media links (if available).</p> <p>The dataset is intended for use in Natural Language Processing research on sentiment classification, especially in the domain of mental health detection using transformer models like BERT and RoBERTa.</p> <p>Missing data is present in fields such as <code>location</code> and <code>username</code>, which have been anonymized or stripped for ethical and privacy compliance.</p> <h3><strong>Dataset Overview</strong></h3> <ul> <li> <p><strong>Total Rows (Entries)</strong>: 10,699</p> </li> <li> <p><strong>Total Columns</strong>: 15</p> </li> <li> <p><strong>Language Focus</strong>: Mostly English (<code>lang = en</code> in 10,687 out of 10,699 rows)</p> </li> <li> <p><strong>Content</strong>: The dataset contains tweets related to <strong>mental health topics</strong> such as anxiety, depression, and stress, collected from <strong>Twitter (X)</strong> between <strong>January and May 2025</strong>.</p> </li> </ul> <h3><strong>Main Columns & Descriptions</strong></h3> <div> <div> <table> <thead> <tr> <th>Column Name</th> <th>Description</th> </tr> </thead> <tbody> <tr> <td><code>conversation_id_str</code></td> <td>Unique conversation/thread identifier</td> </tr> <tr> <td><code>created_at</code></td> <td>Timestamp of the tweet</td> </tr> <tr> <td><code>favorite_count</code></td> <td>Number of likes/favorites</td> </tr> <tr> <td><code>full_text</code></td> <td>The actual tweet text content</td> </tr> <tr> <td><code>id_str</code></td> <td>Tweet ID</td> </tr> <tr> <td><code>image_url</code></td> <td>URL to any attached image (if any)</td> </tr> <tr> <td><code>in_reply_to_screen_name</code></td> <td>Username if the tweet is a reply</td> </tr> <tr> <td><code>lang</code></td> <td>Language of the tweet (mostly <code>en</code>)</td> </tr> <tr> <td><code>location</code></td> <td>Tweet author's location (all <code>null</code> in this dataset)</td> </tr> <tr> <td><code>quote_count</code></td> <td>Number of times the tweet was quoted</td> </tr> <tr> <td><code>reply_count</code></td> <td>Number of replies</td> </tr> <tr> <td><code>retweet_count</code></td> <td>Number of retweets</td> </tr> <tr> <td><code>tweet_url</code></td> <td>Direct URL to the tweet</td> </tr> <tr> <td><code>user_id_str</code></td> <td>ID of the user who posted the tweet</td> </tr> <tr> <td><code>username</code></td> <td>Username of the author (all <code>null</code> in this dataset)</td> </tr> </tbody> </table> </div> </div> <h3><strong>Data Highlights</strong></h3> <ul> <li> <p><strong>Tweets with images</strong>: 2,999 rows contain a non-null <code>image_url</code></p> </li> <li> <p><strong>Replies</strong>: About 4,725 tweets are replies (<code>in_reply_to_screen_name</code> not null)</p> </li> <li> <p><strong>Engagement</strong>:</p> <ul> <li> <p><code>favorite_count</code>: ranges from 0 to 34,364</p> </li> <li> <p><code>retweet_count</code>: ranges from 0 to 1,958</p> </li> <li> <p>Most tweets have low engagement (median values are 0)</p> </li> </ul> </li> <li> <p><strong>Top mentioned account</strong>: <code>elonmusk</code> (mentioned in replies 61 times)</p> </li> <li> <p><strong>Missing Data</strong>:</p> <ul> <li> <p><code>location</code> and <code>username</code>: completely null</p> </li> <li> <p><code>image_url</code>: ~72% of tweets don’t include images</p> </li> </ul> </li> </ul> <p> </p> <p> </p> |
| title | Mental Health Twitter Dataset for Sentiment Analysis (Jan–May 2025) |
| topic | Mental Health NLP Sentiment Analysis BERT RoBERTa |
| url | https://doi.org/10.5281/zenodo.17144715 |