AMRITA_CEN@SemEval-2015: Paraphrase Detection for Twitter using Unsupervised Feature Learning with Recursive Autoencoders

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 45–50, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics AMRITA¯CEN@SemEval-2015: Paraphrase Detection for Twitter using Unsupervised Feature Learning with Recursive Autoencoders Mahalakshmi Shanmuga Sundaram, Anand Kumar Madasamy and Soman Kotti Padannayil Center for Excellence in Computational Engineering and Networking Amrita Vishwa Vidyapeetham Coimbatore, India mahalakshmisklu@gmail.com m¯anandkumar@cb.amrita.edu kp¯soman@amrita.edu Abstract We explore using recursive autoencoders for SemEval 2015 Task 1: Paraphrase and Seman- tic Similarity in Twitter. Our paraphrase de- tection system makes use of phrase-structure parse tree embeddings that are then provided as input to a conventional supervised classi- fication model. We achieve an F1 score of 0.45 on paraphrase identification and a Pear- son correlation of 0.303 on computing seman- tic similarity. 1 Introduction The process of rewriting text with a different choice of words or using a different sentence structure while preserving meaning is called paraphrasing. Identifying paraphrases can be a difficult task owing to the fact that evaluating surface level similarity is often not enough, but rather systems must take into account the underlying semantics of the content be- ing assessed. Paraphrasing and paraphrase detection are impor- tant and challenging tasks, which find their applica- tion in various subfields of Natural Language Pro- cessing (NLP) such as information retrieval, ques- tion answering (Erwin and Emiel, 2005), plagiarism detection (Paul Clough et al., 2002), text summa- rization and evaluation of machine translation (Chris Callison Burch, 2008). We explore using recursive autoencoders for para- phrase detection and similarity scoring as a part of SemEval 2015 Task 1: Paraphrase and Semantic Similarity in Twitter. Twitter is an online social net- working service with millions of users who casually converse about diverse topics in a continuous and contemporaneous manner (Wei Xu et al., 2014; Wei Xu et al., 2015). Table 1 gives an example of real tweets, some of which are paraphrases of each other. The very casual style of the Twitter corpus makes it more challenging to work with for many NLP tools. We use vector space embeddings, in part, since they are relatively good at dealing with noisy data. 2 Related Work Socher et al. (2011) explored using recursive au- toencoders (RAEs) and dynamic pooling for para- phrase detection. They parse each sentence within a pair, compute embeddings for each node in the parse trees, and then construct a similarity matrix compar- ing the embedding vectors for all nodes within the two parse trees. Using dynamic pooling, they con- vert the variable size similarity matrix for each sen- tence pair to a matrix of fixed size. The resulting fixed size matrix is then given to a softmax classifier to detect whether the sentences are paraphrases. 3 A Deep Learning System The architecture of our system is depicted in Figure 1. The raw Twitter corpus is preprocessed using a phrase-structure parser. The resulting parse trees are then used to train an unfolding RAE model. This model provides us with embedding vectors that are then used to compute …

Content truncated to 500 words