Title:
DEPENDENCY ANALYSIS SYSTEM AND DEPENDENCY ANALYSIS METHOD
Document Type and Number:
Japanese Patent JP2020187558
Kind Code:
A
Abstract:
To provide a dependency analysis system that realizes dependency analysis corresponding to a plurality of languages.SOLUTION: A dependency analysis system includes one or more processors and one or more storage devices. One or more processors acquire an array of parts of speech, use a feature extraction machine learning model to extract a plurality of features for each element of the array of parts of speech, use the feature selection machine learning model to select features for each element of the array of parts of speech from a plurality of extracted features, and estimate, based on the selected features, a dependency relation.SELECTED DRAWING: Figure 1
Inventors:
OZAKI HIROAKI
KOREEDA YUTA
SOGABE MASAYOSHI
ABE TAKASHI
KOREEDA YUTA
SOGABE MASAYOSHI
ABE TAKASHI
Application Number:
JP2019091859A
Publication Date:
November 19, 2020
Filing Date:
May 15, 2019
Export Citation:
Assignee:
HITACHI LTD
International Classes:
G06F40/216
Domestic Patent References:
| JP2008123403A | 2008-05-29 |
Other References:
寺西 裕紀 外２名: "マルチタスク学習による遷移型依存構文解析", 言語処理学会第２４回年次大会 発表論文集 ［ＯＮＬＩＮＥ］, JPN6022051534, 5 March 2018 (2018-03-05), JP, pages 1092 - 1095, ISSN: 0004936806
小比田 涼介 外２名: "逐次型依存構造解析における動的特徴量選択", 言語処理学会第２４回年次大会 発表論文集 ［ＯＮＬＩＮＥ］, JPN6022051535, 5 March 2018 (2018-03-05), JP, pages 865 - 868, ISSN: 0004936807
AMMAR, WALEED 外４名: "Many Languages, One Parser", TRANSACTIONS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, vol. 4, JPN6022051533, 31 July 2016 (2016-07-31), pages 431 - 444, XP055756732, ISSN: 0004936808, DOI: 10.1162/tacl_a_00109
小比田 涼介 外２名: "逐次型依存構造解析における動的特徴量選択", 言語処理学会第２４回年次大会 発表論文集 ［ＯＮＬＩＮＥ］, JPN6022051535, 5 March 2018 (2018-03-05), JP, pages 865 - 868, ISSN: 0004936807
AMMAR, WALEED 外４名: "Many Languages, One Parser", TRANSACTIONS OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, vol. 4, JPN6022051533, 31 July 2016 (2016-07-31), pages 431 - 444, XP055756732, ISSN: 0004936808, DOI: 10.1162/tacl_a_00109
Attorney, Agent or Firm:
Fujio Patent Office