US 12,437,152 B2
Headword extraction method and apparatus, device, and storage medium
Bolun Cai, Shenzhen (CN); Xiaoyi Jia, Shenzhen (CN); Haoyu Li, Shenzhen (CN); Yugeng Lin, Shenzhen (CN); Dianping Xu, Shenzhen (CN); Huajie Huang, Shenzhen (CN); Chen Ran, Shenzhen (CN); Yike Liu, Shenzhen (CN); Lijian Mei, Shenzhen (CN); Zhikang Tan, Shenzhen (CN); Yanhua Cheng, Shenzhen (CN); Jinchang Xu, Shenzhen (CN); Minhui Wu, Shenzhen (CN); and Mei Jiang, Shenzhen (CN)
Assigned to Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed by Tencent Technology (Shenzhen) Company Limited, Shenzhen (CN)
Filed on Jul. 5, 2022, as Appl. No. 17/857,841.
Application 17/857,841 is a continuation of application No. PCT/CN2021/096762, filed on May 28, 2021.
Claims priority of application No. 202010486516.7 (CN), filed on Jun. 1, 2020.
Prior Publication US 2022/0343074 A1, Oct. 27, 2022
Int. Cl. G06F 40/279 (2020.01); G06F 40/205 (2020.01); G06F 40/30 (2020.01)
CPC G06F 40/279 (2020.01) [G06F 40/205 (2020.01); G06F 40/30 (2020.01)] 18 Claims
OG exemplary drawing
 
1. A headword extraction method performed by a computer device, the method comprising:
obtaining a sentence feature of a target sentence and word features of a plurality of words in the target sentence;
extracting semantics of the sentence feature of the target sentence to obtain a semantic feature of the target sentence, the semantic feature of the target sentence being used for representing global semantics of the target sentence;
extracting semantics of the word feature of each word of the plurality of words to obtain a semantic feature of the word, the semantic feature of the word being used for representing local semantics of the word;
obtaining a respective matching degree between the semantic feature of each word of the plurality of words and the semantic feature of the target sentence, further comprising:
subtracting the semantic feature of the word from the semantic feature of the target sentence to obtain a differential semantic feature corresponding to the word; and
determining the matching degree corresponding to the word according to the differential semantic feature corresponding to the word, wherein the matching degree corresponding to the word is negatively correlated with the differential semantic feature corresponding to the word such that the larger the differential semantic feature corresponding to the word, the smaller a probability that semantics of the word reflect semantics of the target sentence and the smaller the differential semantic feature corresponding to the word, the larger the probability that the semantics of the word reflect the semantics of the target sentence; and
determining, among the plurality of words in the target sentence, a word with a largest corresponding matching degree as a headword of the target sentence.