Who can verify this? Finding authorities for rumor verification in Twitter paper
We provide the following:
- Main folder: data
Note: Users metadata can be downloaded from here, and the users collection indexes can be downloaded from here.
We provide the rumors in a JSON format file. The file contains a list of JSON objects representing rumors. For each rumor, we provide the following entries:
{
rumor_id [unique ID for the rumor]
tweet_id [unique tweet ID as provided by Twitter]
tweet_text [tweet text as collected from Twitter]
category [category of the rumor which is either politics, sports, or health]
}
Examples:
{
"rumor_id": "AuFIN_024",
"tweet_id": 1355061889122889729,
"tweet_text": "خبر سار لجماهير الاهلي.. بي ان سبورت تختار احمد الطيب للتعليق علي مباراة الاهلي والدحيل ف افتتاح كاس العالم للاندية https://t.co/JPNHMcNyZM",
"category": "sports"
},
...,
{
"rumor_id": "AuFIN_034",
"tweet_id": 1329019940435849217,
"tweet_text": "بعد تعينه نائب رئيس الجمعية الدستورية القطرية الرئيس قيس سعيد يقسم اليمين امام صاحبة السمو الشيخة موزة المسند فى عاصمة #قطر الدوحة https://t.co/kCYFDQHYcS",
"category": "politics"
},
...
We provide the rumors 5 folds used in our experiments. Each fold file containing 30 rumors, i.e., 10 from each of the following categories: politics, sports, and health.
We provide qrels files in .txt format. It is a TAB-separated file in TREC format. Each rumor ID is associated with user IDs of the authorities. The file is in the following format
rumor_id 0 user_id relevance
where:
- rumor_id: Unique ID for the given rumor
- 0: Literally 0 (this column is needed to comply with the TREC format)
- user_id: Unique ID for the given authority
- relevance: 2 if the authority is highly relevant to the rumor (has higher priority to be contacted); 1 if she is relevant
Note: In the qrels file, only pairs with relevance = 1 or 2 are reported. Relevance = 0 is assumed for all pairs not appearing in the qrels file.
Example:
rumor_id | 0 | user_id | relevance |
---|---|---|---|
AuFIN_042 | 0 | 129518457 | 2 |
AuFIN_042 | 0 | 45522292 | 2 |
AuFIN_042 | 0 | 722486044059398144 | 1 |
AuFIN_042 | 0 | 745198763552145409 | 1 |
AuFIN_042 | 0 | 43580047 | 1 |
AuFIN_013 | 0 | 259372802 | 1 |
AuFIN_013 | 0 | 931668301 | 1 |
AuFIN_013 | 0 | 745581936421257216 | 2 |
... |
We provide a collection of 1000 JSON files with users information. Each file has a list of JSON objects representing users. For each user we provide the following entries:
{
user_id [the unique user ID as provided by Twitter]
name [the name of the user, as they’ve defined it on their profile]
description [the text of this user's profile description (also known as bio), if the user provided one]
translated_name [the name of the user translated by us into Arabic]
translated_desc [the user's profile description translated by us into Arabic]
following_count [the number of Twitter users this user is following]
followers_count [the number of Twitter users following this user]
verified [indicates if this user is a verified Twitter User (1 or 0)]
lists_count [the number of Twitter lists this user is member of]
lists_ids [list of unique IDs of the Twitter lists the user is member of if exists]
collected_Arabic_tweets_ids [unique IDs of Arabic tweets posted by this user (recent at the collection time)]
}
Example:
{
"user_id": 1303766076,
"name": "UNDPIraq",
"description": "Committed to supporting the people & Government of Iraq during the transition towards reconciliation, reform & stability.",
"translated_name": "برنامج الأمم المتحدة الإنمائي العراق",
"translated_desc": "ملتزمون بدعم حكومة الشعب العراقي خلال فترة الانتقال نحو المصالحة الإصلاحية والاستقرار",
"following_count": 295,
"followers_count": 21814,
"verified": 1,
"lists_count": 196,
"lists_ids": "1441652508531646465,1438861026968092672,1437411195066142729,........
"collected_Arabic_tweets_ids": "1425811783835557897,1401106018323599363,1313790186816446464,...
}
We provide a collection of 1000 JSON files with Twitter lists information. Each file has a list of JSON objects representing Twitter lists. For each Twitter list, we provide the following entries:
{
list_id [the unique list ID as provided by Twitter]
name [the name of the list, as defined when creating the list]
description [a brief description to let users know about the list, if the user provided one]
translated_name [the name of the list translated by us into Arabic]
translated_desc [the list's description translated by us into Arabic]
member_count [the number of users that are part of this list (added as members by the owner)]
follower_count [the number of users following this list]
created_at [the UTC datetime that the list was created on Twitter]
owner_id [unique ID of the user who owns (created) this list]
owner_name [the name of the user who owns (created) this list, as they’ve defined it on their profile]
}
Example:
{
"list_id": 61845497,
"name": "Tech News",
"description": "Talking about the latest technological inventions and events.",
"translated_name": "أخبار التكنولوجيا",
"translated_desc": "الحديث عن أحدث الاختراعات والأحداث التكنولوجية",
"member_count": 13,
"follower_count": 2,
"created_at": "Wed Dec 28 19:43:53 +0000 2011",
"owner_id": 180411223,
"owner_name": "𝙷𝚊𝚒𝚍𝚊𝚛 𝚉𝚎𝚒𝚗𝚎𝚍𝚍𝚒𝚗𝚎"
},
We release both English and Arabic keywords we used to collect our seed of users Twitter account by either streaming or searching. The English and Arabic keywords are in two separated .txt files.
We provide seven indexes for the users collection indexed using Pyserini. We give a description of each below:
- bio_index: each user is represented by his translated Twitter profile name and description.
- lists_index: each user is represented by concatenating his translated Twitter lists names and descriptions.
- timeline_index: each user is represented by concatenating the collected Arabic tweets from his timeline.
- bio_lists_index: each user is represented by concatenating his translated Twitter profile name and description and his translated Twitter lists.
- lists_timeline_index: each user is represented by concatenating his translated Twitter lists and the collected Arabic tweets from his timeline.
- bio_lists_timeline_index: each user is represented by concatenating his translated Twitter profile name and description, his translated Twitter lists names and descriptions, and the collected Arabic tweets from his timeline
- Arabic Wikipedia2vec: a pre-trained Wikipedia2vec model with Arabic Wikipedia 2022. Download from here
- Fatima Haouari (Qatar University)
- Tamer Elsayed (Qatar University)
- Watheq Mansour (Qatar University)
@article{AuFIN, title = {Who can verify this? Finding authorities for rumor verification in Twitter}, journal = {Information Processing & Management}, volume = {60}, number = {4}, pages = {103366}, year = {2023}, issn = {0306-4573}, author = {Fatima Haouari and Tamer Elsayed and Watheq Mansour}, }
This work was made possible by GSRA grant# GSRA6-1-0611-19074, and NPRP grant# NPRP 11S-1204-170060 from the Qatar National Research Fund. The statements made herein are solely the responsibility of the authors.