Abstract Meaning Representation for Paraphrase Detection

Proceedings of NAACL-HLT 2018, pages 442–452 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Abstract Meaning Representation for Paraphrase Detection Fuad Issa† Marco Damonte† Shay B. Cohen† Xiaohui Yan‡ Yi Chang‡ † School of Informatics, University of Edinburgh, 10 Crichton Street, Edinburgh EH8 9AB, UK ‡ Huawei Technologies, San Jose, CA 95050, USA issa.fuad@gmail.com m.damonte@sms.ed.ac.uk scohen@inf.ed.ac.uk {yanxiaohui2,yi.chang}@huawei.com Abstract Abstract Meaning Representation (AMR) parsing aims at abstracting away from the syn- tactic realization of a sentence, and denoting only its meaning in a canonical form. As such, it is ideal for paraphrase detection, a problem in which one is required to specify whether two sentences have the same meaning. We show that na¨ıve use of AMR in paraphrase de- tection is not necessarily useful, and turn to describe a technique based on latent seman- tic analysis in combination with AMR parsing that significantly advances state-of-the-art re- sults in paraphrase detection for the Microsoft Research Paraphrase Corpus. Our best results in the transductive setting are 86.6% for accu- racy and 90.0% for F1 measure. 1 Introduction Abstract Meaning Representation (AMR) parsing focuses on the conversion of natural language sen- tences into AMR graphs, aimed at abstracting away from the surface realizations of the sentences while preserving their meaning. We make a first step towards showing that AMR can be used in practice for a task that requires identifying the canonicalization of language: para- phrase detection. In a “perfect world” using AMR to test for paraphrasing relation of two sentences should be simple. It would require finding the two AMR parses for each of the sentences, and then checking whether they are identical. Since AMR is aimed at abstracting away from the surface form which is used to express meaning, two sentences should be paraphrases only if they have identical AMRs. For instance, the three sentences: 1. He described her as a curmudgeon, 2. His description of her: curmudgeon, 3. She was a curmudgeon, according to his de- scription. describe-01 he curmudgeon she ARG0 ARG2 ARG1 Figure 1: AMR graph for “He described her as a cur- mudgeon”, “His description of her: curmudgeon” and “She was a curmudgeon, according to his description” should result in the same AMR graph as shown in Figure 1. However, in practice, things are different. First, there are no known AMR parsers that really distil only the meaning in text. For example, predicates which have interchangeable meaning use differ- ent AMR concepts, and there are errors that exist because of the machine learning techniques that are used for learning the parsers from data. Fi- nally, even human annotations do not yield per- fect AMRs, as the interannotator agreement re- ported in the literature for AMR is around 80% (Banarescu et al., 2013). Second, meaning is often contextual, and it is not fully possible to determine the corresponding AMR parse just by looking at a given sentence. Entity mentions denote different entities in differ- ent contexts, and similarly predicates and nouns are ambiguous and …

Content truncated to 500 words