{"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/explaining-character-aware-neural-networks","title":"Explaining Character-Aware Neural Networks for Word-Level Prediction: Do They Discover Linguistic Rules?","arxiv_id":"1808.09551","date":"2018-08-28","proceeding":"EMNLP 2018 10","authors":["Fréderic Godin","Kris Demuynck","Joni Dambre","Wesley De Neve","Thomas Demeester"],"abstract":"Character-level features are currently used in different neural network-based\nnatural language processing algorithms. However, little is known about the\ncharacter-level patterns those models learn. Moreover, models are often\ncompared only quantitatively while a qualitative analysis is missing. In this\npaper, we investigate which character-level patterns neural networks learn and\nif those patterns coincide with manually-defined word segmentations and\nannotations. To that end, we extend the contextual decomposition technique\n(Murdoch et al. 2018) to convolutional neural networks which allows us to\ncompare convolutional neural networks and bidirectional long short-term memory\nnetworks. We evaluate and compare these models for the task of morphological\ntagging on three morphologically different languages and show that these models\nimplicitly discover understandable linguistic rules. Our implementation can be\nfound at https://github.com/FredericGodin/ContextualDecomposition-NLP .","url_abs":"http://arxiv.org/abs/1808.09551v1","url_pdf":"http://arxiv.org/pdf/1808.09551v1.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":"explaining-character-aware-neural-networks","repo_url":"https://github.com/FredericGodin/ContextualDecomposition-NLP","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"morphological-tagging","task_name":"Morphological Tagging"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1808.09551","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}