{"links":{},"id":56731,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00056731","sets":["1164:5159:5168:5170"]},"path":["5170"],"owner":"1","recid":"56731","title":["効率的なクロスバリデーション尤度評価に基づく混合ガウス分布の最適化"],"pubdate":{"attribute_name":"公開日","attribute_value":"2007-07-21"},"_buckets":{"deposit":"8c840a31-d380-444a-9d03-4a26f045c93a"},"_deposit":{"id":"56731","pid":{"type":"depid","value":"56731","revision_id":0},"owners":[1],"status":"published","created_by":1},"item_title":"効率的なクロスバリデーション尤度評価に基づく混合ガウス分布の最適化","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"効率的なクロスバリデーション尤度評価に基づく混合ガウス分布の最適化"},{"subitem_title":"Gaussian Mixture Optimization based on Efficient Cross-validation","subitem_title_language":"en"}]},"item_type_id":"4","publish_date":"2007-07-21","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"京都大学学術情報メディアセンター"},{"subitem_text_value":"京都大学学術情報メディアセンター"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Academic Center for Computing and Media Studies Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"Academic Center for Computing and Media Studies Kyoto University","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/56731/files/IPSJ-SLP07067015.pdf"},"date":[{"dateType":"Available","dateValue":"2009-07-21"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP07067015.pdf","filesize":[{"value":"551.3 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"22"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"e4a0a34e-3a50-444e-ab2f-d9cdf5fdac6d","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2007 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"篠崎, 隆宏"},{"creatorName":"河原, 達也"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Takahiro, SHINOZAKI","creatorNameLang":"en"},{"creatorName":"Tatsuya, KAWAHARA","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10442647","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"従来の自己尤度に代えてクロスバリデーション尤度を用いる新しい混合分布最適化アルゴリズムの提案を行い、HMMを用いた音声認識への応用を行う。混合分布の最適化の目的は、過剰な要素を削減することでモデルの一般性を高めることであり、最適化は混合分布要素の対を尤度に従い順次選択・併合することで行う。クロスバリデーション尤度はモデルパラメタのオーバーフィッティングを避ける上で従来の尤度よりも有効であり、また十分統計量を活用することで高速に評価することができる。これにより、従来よりも優れた分布要素対選択を行うことができるとともに、経験的な閾値に頼らない併合停止基準が与えられる利点がある。日本語話し言葉コーパスを用いた大語彙連続音声認識をタスクとし、HMMの学習に対して本手法を適用した実験結果において、本手法が従来手法よりも高い認識率を与えることを示す。","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"A Gaussian mixture optimization method is explored using cross-validation likelihood as an objective function instead of the conventional training set likelihood.The optimization is based on reducing the number of mixture components by selecting and merging a pair of Gaussians step by step based on the objective function so as to remove redundant components and improve the generality of the model. Cross-validation likelihood is more appropriate for avoiding over-fitting than the conventional likelihood and can be efficiently computed using sufficient statistics. It results in a better Gaussian pair selection and provides a termination criterion that does not rely on empirical thresholds. Large-vocabulary speech recognition experiments on oral presentations show that the cross-validation method gives a smaller word error rate with an automatically determined model size than a baseline using the conventional training procedure.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"86","bibliographic_titles":[{"bibliographic_title":"情報処理学会研究報告音声言語情報処理（SLP）"}],"bibliographicPageStart":"81","bibliographicIssueDates":{"bibliographicIssueDate":"2007-07-21","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"75(2007-SLP-067)","bibliographicVolumeNumber":"2007"}]},"relation_version_is_last":true,"weko_creator_id":"1"},"created":"2025-01-18T23:20:05.095065+00:00","updated":"2025-01-22T04:50:40.059829+00:00"}