{"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/combining-discrete-and-neural-features-for","title":"Combining Discrete and Neural Features for Sequence Labeling","arxiv_id":"1708.07279","date":"2017-08-24","proceeding":null,"authors":["Jie Yang","Zhiyang Teng","Meishan Zhang","Yue Zhang"],"abstract":"Neural network models have recently received heated research attention in the\nnatural language processing community. Compared with traditional models with\ndiscrete features, neural models have two main advantages. First, they take\nlow-dimensional, real-valued embedding vectors as inputs, which can be trained\nover large raw data, thereby addressing the issue of feature sparsity in\ndiscrete models. Second, deep neural networks can be used to automatically\ncombine input features, and including non-local features that capture semantic\npatterns that cannot be expressed using discrete indicator features. As a\nresult, neural network models have achieved competitive accuracies compared\nwith the best discrete models for a range of NLP tasks.\n  On the other hand, manual feature templates have been carefully investigated\nfor most NLP tasks over decades and typically cover the most useful indicator\npattern for solving the problems. Such information can be complementary the\nfeatures automatically induced from neural networks, and therefore combining\ndiscrete and neural features can potentially lead to better accuracy compared\nwith models that leverage discrete or neural features only.\n  In this paper, we systematically investigate the effect of discrete and\nneural feature combination for a range of fundamental NLP tasks based on\nsequence labeling, including word segmentation, POS tagging and named entity\nrecognition for Chinese and English, respectively. Our results on standard\nbenchmarks show that state-of-the-art neural models can give accuracies\ncomparable to the best discrete models in the literature for most tasks and\ncombing discrete and neural features unanimously yield better results.","url_abs":"http://arxiv.org/abs/1708.07279v1","url_pdf":"http://arxiv.org/pdf/1708.07279v1.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":"combining-discrete-and-neural-features-for","repo_url":"https://github.com/SUTDNLP/NNSegmentation","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"named-entity-recognition-1","task_name":"Named Entity Recognition"},{"task_slug":"named-entity-recognition-ner","task_name":"Named Entity Recognition (NER)"},{"task_slug":"pos","task_name":"POS"},{"task_slug":"pos-tagging","task_name":"POS Tagging"},{"task_slug":"named-entity-recognition","task_name":"named-entity-recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1708.07279","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}