EmTaggeR: A Word Embedding Based Novel Method for Hashtag Recommendation on Twitter

摘要

The hashtag recommendation problem addresses recommending (suggesting) one ormore hashtags to explicitly tag a post made on a given social network platform,based upon the content and context of the post. In this work, we propose anovel methodology for hashtag recommendation for microblog posts, specificallyTwitter. The methodology, EmTaggeR, is built upon a training-testing frameworkthat builds on the top of the concept of word embedding. The training phasecomprises of learning word vectors associated with each hashtag, and deriving aword embedding for each hashtag. We provide two training procedures, one inwhich each hashtag is trained with a separate word embedding model applicablein the context of that hashtag, and another in which each hashtag obtains itsembedding from a global context. The testing phase constitutes computing theaverage word embedding of the test post, and finding the similarity of thisembedding with the known embeddings of the hashtags. The tweets that containthe most-similar hashtag are extracted, and all the hashtags that appear inthese tweets are ranked in terms of embedding similarity scores. The top-Khashtags that appear in this ranked list, are recommended for the given testpost. Our system produces F1 score of 50.83%, improving over the LDA baselineby around 6.53 times, outperforming the best-performing system known in theliterature that provides a lift of 6.42 times. EmTaggeR is a fast, scalable andlightweight system, which makes it practical to deploy in real-lifeapplications.