[논문 리뷰] ArCOV-19: The First Arabic COVID-19 Twitter Dataset with Propagation Networks

Fatima Haouari, Maram Hasanain외 2명|arXiv (Cornell University)|2020. 04. 13.

Misinformation and Its Impacts인용 수 46

한 줄 요약

ArCOV-19는 가장 인기 있는 부분에 대해 원본 트윗과 전파 네트워크를 포함하는 최초의 공개된 Arabic COVID-19 Twitter 데이터셋(Jan 2020–Jan 2021)입니다.

ABSTRACT

In this paper, we present ArCOV-19, an Arabic COVID-19 Twitter dataset that spans one year, covering the period from 27th of January 2020 till 31st of January 2021. ArCOV-19 is the first publicly-available Arabic Twitter dataset covering COVID-19 pandemic that includes about 2.7M tweets alongside the propagation networks of the most-popular subset of them (i.e., most-retweeted and -liked). The propagation networks include both retweets and conversational threads (i.e., threads of replies). ArCOV-19 is designed to enable research under several domains including natural language processing, information retrieval, and social computing. Preliminary analysis shows that ArCOV-19 captures rising discussions associated with the first reported cases of the disease as they appeared in the Arab world. In addition to the source tweets and propagation networks, we also release the search queries and language-independent crawler used to collect the tweets to encourage the curation of similar datasets.

연구 동기 및 목표

• NLP, IR, 및 사회 컴퓨팅 연구를 가능하게 하기 위해 COVID-19에 초점을 맞춘 1년 간의 Arabic Twitter 데이터셋을 선별합니다.
• 가장 인기 있는 부분의 트윗에 대한 전파 네트워크(리트윗 및 대화 스레드)를 제공합니다.
• 다른 연구자들이 재현 및 확장을 용이하게 하도록 크롤링 방법론, 쿼리 및 도구를 공개합니다.

제안 방법

• 수동으로 작성된 Arabic 쿼리를 사용하여 언어별 Twitter 검색 크롤러로 매일 원본 트윗을 수집합니다.
• 리트윗과 중복을 제외하고; 검색 API 창의 제한에 도달했을 때 GetOldTweets3를 사용해 오래된 트윗을 복구합니다.
• 일별로 인기 점수(리트윗 + 즐겨찾기 합계)에 따라 상위 1,000개 트윗을 식별하고 이들의 리트윗과 답글을 수집하여 전파 네트워크를 형성합니다.
• 연구자들을 위해 소스 트윗 ID, 검색 쿼리, 상위 부분 ID 및 전파 네트워크(리트윗 및 대화 스레드)를 공개합니다.
• 아랍 세계의 초기 COVID-19 기간 커버리지를 검증하기 위해 데이터셋의 시간적, 지리적, 주제적 특성을 분석합니다.

실험 결과

연구 질문

• RQ1 What are the temporal dynamics of Arabic COVID-19 discourse on Twitter from Jan 2020 to Jan 2021?
• RQ2 What is the geographical distribution of Arabic COVID-19 tweets and how does it reflect outbreak timelines in Arab countries?
• RQ3 What topics, hashtags, and domains dominate Arabic COVID-19 discussions and how do they evolve over time?
• RQ4 How do propagation networks (retweets and replies) for the most popular tweets illuminate information diffusion and potential misinformation dynamics?

주요 결과

• The dataset comprises about 2.7 million source tweets from over 690k users; 18.66% are from verified users and 25.40% include URLs.
• Top subset includes 370,132 tweets (13.84% of source tweets); total retweets in propagation networks reach 7,925,821 and replies reach 1,476,950 (as of April 2020 for replies).
• Geolocated tweets (place attribute) number 60,873 (2.28%), posted by 24,072 users; geotagged tweets (coordinates) number 2,078 (0.08%), from 102 countries mainly in the Arab world (92.75%).
• Saudi Arabia contributes about 41.7% of geolocated content, Kuwait about 9.6%, reflecting early outbreak timelines and country activity.
• URLs in top subset are dominated by news websites from Egypt, Saudi Arabia, and UAE; YouTube videos are among frequently shared media.
• Propagation patterns show many tweets receiving more than 100 retweets, with some exceeding 10k, indicating strong diffusion for top tweets.