{"corpus_id":14635907,"paper_sha":"937d3a404b8870fb3ff3e243e6a0c6024eef491b","doi":"10.1109/TPAMI.2008.137","arxiv_id":null,"pmid":19299860,"pmcid":null,"mag_id":2122585011,"dblp_id":"journals/pami/GravesLFBBS09","acl_id":null,"title":"A Novel Connectionist System for Unconstrained Handwriting Recognition","year":2009,"publication_date":"2009-05-01","venue":"IEEE Transactions on Pattern Analysis and Machine Intelligence","journal":{"name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","pages":"855-868","volume":"31"},"journal_issn":null,"journal_title":null,"publication_types":["JournalArticle"],"pubmed_pub_types":["Journal Article","Research Support, Non-U.S. Gov't"],"s2_fields_of_study":["Medicine","Computer Science"],"reference_count":59,"citation_count":1937,"influential_citation_count":117,"is_open_access":true,"arxiv_categories":null,"arxiv_license":null,"arxiv_journal_ref":null,"mesh_headings":[{"d":"Algorithms","mj":true,"ui":"D000465"},{"d":"Electronic Data Processing","mj":false,"qs":[{"q":"methods","mj":true,"ui":"Q000379"}],"ui":"D001330"},{"d":"Handwriting","mj":true,"ui":"D006236"},{"d":"Image Enhancement","mj":false,"qs":[{"q":"methods","mj":false,"ui":"Q000379"}],"ui":"D007089"},{"d":"Image Interpretation, Computer-Assisted","mj":false,"qs":[{"q":"methods","mj":true,"ui":"Q000379"}],"ui":"D007090"},{"d":"Information Storage and Retrieval","mj":false,"qs":[{"q":"methods","mj":true,"ui":"Q000379"}],"ui":"D016247"},{"d":"Models, Statistical","mj":false,"ui":"D015233"},{"d":"Pattern Recognition, Automated","mj":false,"qs":[{"q":"methods","mj":true,"ui":"Q000379"}],"ui":"D010363"},{"d":"Reading","mj":false,"ui":"D011932"},{"d":"Reproducibility of Results","mj":false,"ui":"D015203"},{"d":"Sensitivity and Specificity","mj":false,"ui":"D012680"},{"d":"Subtraction Technique","mj":false,"ui":"D013382"}],"chemicals":null,"comments_corrections":null,"source_flags":5,"s2_open_access_pdf_url":"http://www.idsia.ch/~juergen/tpami_2008.pdf","s2_open_access_landing_url":"https://www.semanticscholar.org/paper/937d3a404b8870fb3ff3e243e6a0c6024eef491b","s2_open_access_license":null,"s2_open_access_status":"GREEN","pmc_open_access_pdf_url":null,"pmc_open_access_landing_url":null,"pmc_open_access_license":null,"pmc_open_access_status":null,"unpaywall_open_access_pdf_url":null,"unpaywall_open_access_landing_url":null,"unpaywall_open_access_license":null,"unpaywall_open_access_status":null,"abstract":"Recognizing lines of unconstrained handwritten text is a challenging task. The difficulty of segmenting cursive or overlapping characters, combined with the need to exploit surrounding context, has led to low recognition rates for even the best current recognizers. Most recent progress in the field has been made either through improved preprocessing or through advances in language modeling. Relatively little work has been done on the basic recognition algorithms. Indeed, most systems rely on the same hidden Markov models that have been used for decades in speech and handwriting recognition, despite their well-known shortcomings. This paper proposes an alternative approach based on a novel type of recurrent neural network, specifically designed for sequence labeling tasks where the data is hard to segment and contains long-range bidirectional interdependencies. In experiments on two large unconstrained handwriting databases, our approach achieves word recognition accuracies of 79.7 percent on online data and 74.1 percent on offline data, significantly outperforming a state-of-the-art HMM-based system. In addition, we demonstrate the network's robustness to lexicon size, measure the individual influence of its hidden layers, and analyze its use of context. Last, we provide an in-depth discussion of the differences between the network and HMMs, suggesting reasons for the network's superior performance.","claims":[{"public_id":"cl_a8749939efa2efb5f916740580b33835","status":"active","text":"A novel recurrent neural network designed for sequence labeling with long-range bidirectional interdependencies achieves word recognition accuracies of 79.7 percent on online data and 74.1 percent on offline data, significantly outperforming a state-of-the-art HMM-based system.","confidence":0.95,"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/claims/cl_a8749939efa2efb5f916740580b33835"},{"public_id":"cl_f137a74ebad9a8174fca41707cf71206","status":"active","text":"The proposed network demonstrates robustness to lexicon size, and the individual influence of its hidden layers and its use of context are analyzed.","confidence":0.9,"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/claims/cl_f137a74ebad9a8174fca41707cf71206"}],"concepts":[{"public_id":"co_26fc162a62a42e40808f8a20a968601a","status":"active","name":"network's superior performance","description":"The superior word recognition accuracy achieved by the proposed recurrent neural network compared to HMM-based systems.","types":["result"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_26fc162a62a42e40808f8a20a968601a"},{"public_id":"co_6960b255be97671bee9e0d4d0d425521","status":"active","name":"online data","description":"Unconstrained handwriting data collected online, used in the experiments to evaluate word recognition accuracy.","types":["dataset"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_6960b255be97671bee9e0d4d0d425521"},{"public_id":"co_7d21234b3280caa93b1f073e15de083e","status":"active","name":"HMM-based system","description":"A state-of-the-art hidden Markov model based system used as a baseline for comparison in handwriting recognition.","types":["baseline"],"aliases":["HMM system"],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_7d21234b3280caa93b1f073e15de083e"},{"public_id":"co_a8a662059df35b8537b43dffb9f43a9a","status":"active","name":"lexicon size","description":"The size of the vocabulary or lexicon used in the handwriting recognition task, to which the network shows robustness.","types":["parameter"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_a8a662059df35b8537b43dffb9f43a9a"},{"public_id":"co_ad56ff777a9bae0ec248ae5cacf03f23","status":"active","name":"recurrent neural network","description":"A novel type of recurrent neural network specifically designed for sequence labeling tasks where data is hard to segment and contains long-range bidirectional interdependencies.","types":["method"],"aliases":["RNN"],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_ad56ff777a9bae0ec248ae5cacf03f23"},{"public_id":"co_b8ad36eecf76dfec581a4a37dbcd462b","status":"active","name":"HMMs","description":"Hidden Markov models that have been used for decades in speech and handwriting recognition, compared with the proposed network.","types":["method"],"aliases":["hidden Markov models"],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_b8ad36eecf76dfec581a4a37dbcd462b"},{"public_id":"co_d304af83e5fb881b27db3f90746f4796","status":"active","name":"offline data","description":"Unconstrained handwriting data collected offline, used in the experiments to evaluate word recognition accuracy.","types":["dataset"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_d304af83e5fb881b27db3f90746f4796"},{"public_id":"co_d99145a2ccb2ee1f2d1bcad28cae05de","status":"active","name":"hidden layers","description":"The individual hidden layers of the recurrent neural network whose influence is measured separately.","types":["component"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_d99145a2ccb2ee1f2d1bcad28cae05de"},{"public_id":"co_eb747312288224ad9f30dea12995f711","status":"active","name":"use of context","description":"The way the network exploits surrounding context in sequence labeling, analyzed in the paper.","types":["property"],"aliases":[],"contributors":[{"id":32,"public_id":"7c402c1b98","public_label":"뀨 (7c402c1b98)","roles":["extraction"],"url":"https://sah.borca.ai/u/7c402c1b98"},{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["review"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":170,"public_id":"gsgmdx9r6e","public_label":"pupuri (gsgmdx9r6e)","roles":["review"],"url":"https://sah.borca.ai/u/gsgmdx9r6e"}],"url":"https://sah.borca.ai/concepts/co_eb747312288224ad9f30dea12995f711"}],"external_ids":{"DOI":"10.1109/TPAMI.2008.137","ArXiv":null,"PubMed":19299860,"PubMedCentral":null,"MAG":2122585011,"DBLP":"journals/pami/GravesLFBBS09","ACL":null},"open_access":{"is_open_access":true,"pdf_url":"http://www.idsia.ch/~juergen/tpami_2008.pdf","landing_url":"https://www.semanticscholar.org/paper/937d3a404b8870fb3ff3e243e6a0c6024eef491b","source":"semantic_scholar","pdf_url_source":"semantic_scholar_open_access_pdf","license":null,"status":"GREEN","reason":null},"reference_availability":{"status":"available","references_indexed":true,"full_text_available":false,"full_text_source":null,"count_basis":"semantic_scholar_metadata","extraction_status":"not_applicable","reason":null},"source":{"provider":"episteme2","base_corpus":"semantic_scholar_dump","freshness_mode":"unknown","basis":["semantic_scholar_metadata","postgres_metadata"],"limits":["paper metadata is based on indexed upstream scholarly datasets","claims and concepts are available only for extracted papers","absence of claims or concepts means no extracted graph data is available in this response"],"status":"available","degraded":false,"degraded_reasons":[],"diagnostics":{"status":"available","degraded":false,"degraded_reasons":[],"metadata_status":"available","graph_status":"available","abstract_status":"available"},"source_flags":5},"paper_id":638231,"paper_uid":"f149507b-9161-459d-8dea-41e7930ece80","canonical_identity":{"paper_id":638231,"paper_uid":"f149507b-9161-459d-8dea-41e7930ece80","identity_status":"available","lookup_basis":"semantic_scholar_external_id","compatibility_path":"corpus_id"},"url":"https://sah.borca.ai/papers/14635907"}