Sentence modeling via multiple word embeddings and multi-level comparison for semantic textual similarity |
| |
| Institution: | 1. Japan Advanced Institute of Science and Technology (JAIST) Japan;2. Toshiba Research & Development Center, Japan;1. Faculty of Computer Science and Engineering, Shahid Beheshti University, Tehran, Iran;2. Faculty of Computer Engineering, University of Isfahan, Isfahan, Iran;1. College of Computer Science and Electronic Engineering, Hunan University, Lushan Road (S), Yuelu District, Changsha, China;2. Computer Science and Engineering department, State University of New York at Buffalo, NY 14260-2500, USA;1. Center for Studies of Information Resources, Wuhan University, Wuhan, 430072, China;2. School of Information Management, Wuhan University, Wuhan, 430072, China;3. Department of Information and Service Economy, Aalto University School of Business, Helsinki, Finland;4. Department of Industrial and Information Management, Tampere University, Tampere, Finland;1. School of Data and Computer Science, Sun Yat-sen University, China;2. School of Computing Science, University of Glasgow, Glasgow, UK;3. School of Computer Science, The University of Adelaide, Adelaide, Australia;1. University of North Carolina at Chapel Hill, United States;2. Data Science, Facebook 1 Hacker Way Menlo Park, CA 94025, United States;3. User Experience and Customer Insights, NetBrain 15 Network Drive Burlington, MA, 01803, United States |
| |
| Abstract: | Recently, using a pretrained word embedding to represent words achieves success in many natural language processing tasks. According to objective functions, different word embedding models capture different aspects of linguistic properties. However, the Semantic Textual Similarity task, which evaluates similarity/relation between two sentences, requires to take into account of these linguistic aspects. Therefore, this research aims to encode various characteristics from multiple sets of word embeddings into one embedding and then learn similarity/relation between sentences via this novel embedding. Representing each word by multiple word embeddings, the proposed MaxLSTM-CNN encoder generates a novel sentence embedding. We then learn the similarity/relation between our sentence embeddings via Multi-level comparison. Our method M-MaxLSTM-CNN consistently shows strong performances in several tasks (i.e., measure textual similarity, identify paraphrase, recognize textual entailment). Our model does not use hand-crafted features (e.g., alignment features, Ngram overlaps, dependency features) as well as does not require pre-trained word embeddings to have the same dimension. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
