Abstract:

Microblogs, such as tweets, are short messages in which users are able to share any opinion and information. Microblogs are mostly related to real-life events reported in news articles. Finding event-related microblogs is important to analyze online social networks and understand public opinion on events. However, finding such microblogs is a challenging task due to the dynamic nature of microblogs and their limited length. In this study, assuming that news articles are given as queries and microblogs as documents, we find event-related microblogs in Turkish. In order to represent news articles and microblogs, we examine encoding methods, namely traditional bag-of-words and word embeddings provided by BERT and FastText pretrained language models based on deep learning. We find the distance between the encoded news article and microblog to measure text similarity or relatedness between them. We then rank microblogs according to their relatedness to the input query. The experimental results show that (i) BERT-based model outperforms other encoding methods in Turkish, though bag-of-words with Dice similarity has a challenging performance in short text; (ii) news title is successful to represent event as query, and (iii) preprocessing Turkish microblogs has positive impact in bag-of-words and also FastText embeddings, while BERT embeddings are robust to noise in Turkish.

Keywords: