{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/match-srnn-modeling-the-recursive-matching","title":"Match-SRNN: Modeling the Recursive Matching Structure with Spatial RNN","arxiv_id":"1604.04378","date":"2016-04-15","proceeding":null,"authors":["Shengxian Wan","Yanyan Lan","Jun Xu","Jiafeng Guo","Liang Pang","Xue-Qi Cheng"],"abstract":"Semantic matching, which aims to determine the matching degree between two\ntexts, is a fundamental problem for many NLP applications. Recently, deep\nlearning approach has been applied to this problem and significant improvements\nhave been achieved. In this paper, we propose to view the generation of the\nglobal interaction between two texts as a recursive process: i.e. the\ninteraction of two texts at each position is a composition of the interactions\nbetween their prefixes as well as the word level interaction at the current\nposition. Based on this idea, we propose a novel deep architecture, namely\nMatch-SRNN, to model the recursive matching structure. Firstly, a tensor is\nconstructed to capture the word level interactions. Then a spatial RNN is\napplied to integrate the local interactions recursively, with importance\ndetermined by four types of gates. Finally, the matching score is calculated\nbased on the global interaction. We show that, after degenerated to the exact\nmatching scenario, Match-SRNN can approximate the dynamic programming process\nof longest common subsequence. Thus, there exists a clear interpretation for\nMatch-SRNN. Our experiments on two semantic matching tasks showed the\neffectiveness of Match-SRNN, and its ability of visualizing the learned\nmatching structure.","url_abs":"http://arxiv.org/abs/1604.04378v1","url_pdf":"http://arxiv.org/pdf/1604.04378v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"match-srnn-modeling-the-recursive-matching","repo_url":"https://github.com/T-Almeida/tensorflow-keras-multidimensional-rnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":null,"task_name":"Position"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1604.04378","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}