Hashtag recommendation for hyperlinked tweets

Abstract

Presence of hyperlink in a tweet is a strong indication of tweet being more informative. In this paper, we study the problem of hashtag recommendation for hyperlinked tweets (i.e., tweets containing links to Web pages). By recommending hashtags to hyperlinked tweets, we argue that the functions of hashtags such as providing the right context to interpret the tweets, tweet categorization, and tweet promotion, can be extended to the linked documents. The proposed solution for hashtag recommendation consists of two phases. In the first phase, we select candidate hashtags through five schemes by considering the similar tweets, the similar documents, the named entities contained in the document, and the domain of the link. In the second phase, we formulate the hashtag recommendation problem as a learning to rank problem and adopt RankSVM to aggregate and rank the candidate hashtags. Our experiments on a collection of 24 million tweets show that the proposed solution achieves promising results.

Hashtagger+: Efficient High-Coverage Social Tagging of Streaming News

“…The work in [16] uses pairwise L2R for hashtag recommendation for tweets. This work is tailored for tweets with at least one URL and one hashtag in their body, a very small subset of the overall tweet pool discussing news.…”

“…This work is tailored for tweets with at least one URL and one hashtag in their body, a very small subset of the overall tweet pool discussing news. Training on this small and noisy tweet set can pose serious problems for the recommendation, resulting in low Precision (reported P@1 of 40% in [16]) and low coverage, i.e., few tweets receiving any recommendation at all (reported coverage of 50%). The data collection is seeded by an external set of 135 trending hashtags collected from hashtags.org each day.…”

“…Furthermore, there is no focus on news nor on efficient recommendation which is critical for our setting. In contrast to the approach in [16], we use the actual news articles to drive the selection of tweets and candidate hashtags. In our experiments we compare to the method of [16] and show that our model achieves much better coverage and Precision@1.…”

“…Pointwise approaches were shown to be efficient and effective for binary relevance labels [23], [32], [33], a finding reinforced in our experiments. We compare our method to existing MCC approaches (Naïve Bayes [9], SVM, Neural Networks) and L2R approaches (pointwise and pairwise L2R [6], [16]). …”

“…We investigate a commercial tool for generating queries from articles, as well as an approach used in related literature [16], [19]. We further investigate two efficient heuristics for extracting queries.…”