SENTENCE SIMILARITY CALCULATION DEVICE, TRAINED MODEL GENERATION DEVICE, AND VARIANCE EXPRESSION MODEL
To improve the accuracy of calculating the degree of similarity with respect to sentences.SOLUTION: A sentence similarity calculation device 20 comprises: a second analysis unit 22 for dividing a plurality of sentences to be compared into grammatically functional phrases each having a prescribed gra...
Gespeichert in:
Hauptverfasser: | , , , |
---|---|
Format: | Patent |
Sprache: | eng ; jpn |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | To improve the accuracy of calculating the degree of similarity with respect to sentences.SOLUTION: A sentence similarity calculation device 20 comprises: a second analysis unit 22 for dividing a plurality of sentences to be compared into grammatically functional phrases each having a prescribed grammatical function; a variance expression acquisition unit 23 for inputting a word included in each of the sentences to be compared to a variance expression model generated for each grammatical function in accordance with the grammatical function which a grammatically functional phase in which the word is included corresponds to, and acquiring as the variance expression of each word; a phrase similarity calculation unit 24 for calculating a phrase similarity which represents the degree of similarity between grammatically functional phases having the same grammatical function in each sentence to be compared, on the basis of the distance between the variance expressions of each word included in each of the grammatically functional phases; a sentence similarity calculation unit 25 for calculating a sentence similarity that represents the degree of similarity between sentences on the basis of the phase similarity between the grammatically functional phases included in each sentence; and an output unit 27 for outputting the sentence similarity between the sentences to be compared.SELECTED DRAWING: Figure 2
【課題】文を対象とした類似度の算出の精度を向上する。【解決手段】文類似度算出装置20は、比較対象の複数の文を、各々が所定の文法的機能を有する文法的機能句に分割する第2解析部22と、比較対象の文のそれぞれに含まれる単語を、単語が含まれる文法的機能句が該当する文法的機能に応じて、文法的機能ごとに生成された分散表現モデルに入力し、各単語の分散表現として取得する分散表現取得部23と、比較対象の各文における、同じ文法的機能を有する文法的機能句の間の類似度である句類似度を、文法的機能句のそれぞれに含まれる各単語の分散表現の間の距離に基づいて算出する句類似度計算部24と、各文に含まれる文法的機能句の間の句類似度に基づいて、文の間の類似度である文類似度を算出する文類似度計算部25と、比較対象の文の間の文類似度を出力する出力部27と、を備える。【選択図】図2 |
---|