{"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/the-lifted-matrix-space-model-for-semantic","title":"The Lifted Matrix-Space Model for Semantic Composition","arxiv_id":"1711.03602","date":"2017-11-09","proceeding":"CONLL 2018 10","authors":["WooJin Chung","Sheng-Fu Wang","Samuel R. Bowman"],"abstract":"Tree-structured neural network architectures for sentence encoding draw\ninspiration from the approach to semantic composition generally seen in formal\nlinguistics, and have shown empirical improvements over comparable sequence\nmodels by doing so. Moreover, adding multiplicative interaction terms to the\ncomposition functions in these models can yield significant further\nimprovements. However, existing compositional approaches that adopt such a\npowerful composition function scale poorly, with parameter counts exploding as\nmodel dimension or vocabulary size grows. We introduce the Lifted Matrix-Space\nmodel, which uses a global transformation to map vector word embeddings to\nmatrices, which can then be composed via an operation based on matrix-matrix\nmultiplication. Its composition function effectively transmits a larger number\nof activations across layers with relatively few model parameters. We evaluate\nour model on the Stanford NLI corpus, the Multi-Genre NLI corpus, and the\nStanford Sentiment Treebank and find that it consistently outperforms TreeLSTM\n(Tai et al., 2015), the previous best known composition function for\ntree-structured models.","url_abs":"http://arxiv.org/abs/1711.03602v2","url_pdf":"http://arxiv.org/pdf/1711.03602v2.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":"the-lifted-matrix-space-model-for-semantic","repo_url":"https://github.com/NYU-MLL/spinn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"the-lifted-matrix-space-model-for-semantic","repo_url":"https://github.com/woojinchung/lms","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"semantic-composition","task_name":"Semantic Composition"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"word-embeddings","task_name":"Word Embeddings"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}