TrWP: Text Relatedness using Word and Phrase Relatedness

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 90–95, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics TrWP: Text Relatedness using Word and Phrase Relatedness Md Rashadul Hasan Rakib Aminul Islam Evangelos Milios Faculty of Computer Science Dalhousie University, Canada {rakib,islam,eem}@cs.dal.ca Abstract Text is composed of words and phrases. In bag-of-word model, phrases in texts are split into words. This may discard the inner seman- tics of phrases which in turn may give incon- sistent relatedness score between two texts. TrWP, the unsupervised text relatedness ap- proach combines both word and phrase relat- edness. The word relatedness is computed using an existing unsupervised co-occurrence based method. The phrase relatedness is com- puted using an unsupervised phrase related- ness function f that adopts Sum-Ratio tech- nique based on the statistics in the Google n- gram corpus of overlapping n-grams associ- ated with the two input phrases. The second run of TrWP ranked 30th out of 73 runs in SemEval-2015 task2a (English STS). 1 Introduction Generally, a phrase is an ordered sequence of mul- tiple words that all together refer to a particular meaning (Zamir and Etzioni, 1999). Phrase related- ness quantifies how two phrases relate to each other. It plays an important role in different Text Mining tasks; for instance, document similarity 1, classifica- tion and clustering are performed on the documents composed of phrases. Several document clustering methods use phrase similarity to determine the simi- larity between documents so as to improve the clus- tering result (Chim and Deng, 2008; Shrivastava et al., 2013). SpamED (Pera and Ng, 2009) uses the 1We use ‘relatedness’ and ‘similarity’ interchangeably in our paper, albeit ‘similarity’ is a special case of ‘relatedness’. bi-gram and tri-gram phrase similarity between an incoming e-mail message and a previously marked spam to enhance the accuracy of spam detection. Most works on text relatedness can be abstracted as a function of word relatedness (Ho et al., 2010). The classical Bag-of-Word (BoW) text relatedness methods split phrases into words; then compute text- pair relatedness by word-pair relatedness (Islam and Inkpen, 2008; Islam et al., 2012; Tsatsaronis et al., 2010). TrWP considers text as Bag-of-Word-and- Phrase (BoWP). It considers a (word, bi-gram) or (bi-gram, bi-gram) pair as a phrase-pair 2 and com- putes text relatedness using both word and phrase relatedness. There are phrase relatedness tasks that use com- positional distributional semantic (CDS) model (An- nesi et al., 2012; Hartung and Frank, 2011). Some use different tools and knowledge-based re- sources (Han et al., 2013; Tsatsaronis et al., 2010). These methods split phrases into words without con- sidering the word order that might change the mean- ing of phrases leading to inconsistent phrase relat- edness score (Turney and Pantel, 2010). For exam- ple, if we split the phrases “boat house” and “house boat” into words, we get the relatedness score one, nonetheless as a whole unit, these two phrases do not refer to exactly the same meaning (Turney and Pan- tel, 2010). To preserve the …

Content truncated to 500 words