DATA PROCESSING METHOD AND DATA PROCESSING APPARATUS

To output an image with the highest similarity to an evaluation word, the word being easily determined even if a difference in similarity between images is small.SOLUTION: In a data processing method, a plurality of candidate images 011 and an evaluation word 013 are input to an input unit 001, and...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: CHEN FANGGE
Format: Patent
Sprache:eng ; jpn
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To output an image with the highest similarity to an evaluation word, the word being easily determined even if a difference in similarity between images is small.SOLUTION: In a data processing method, a plurality of candidate images 011 and an evaluation word 013 are input to an input unit 001, and a candidate image 011 with the highest first similarity to the evaluation word 013, in the input candidate images 011, is output. An extraction unit 002 calculates a first difference in the first similarity to the evaluation word 013 between the candidate images 011. When the first difference is equal to or lower than a first threshold, a related word search unit 003 searches for a related word 031, from a plurality of sentences including the evaluation word 013, the related word 031 appearing in the sentences at a degree of association with the evaluation word 013 equal to or higher than a first degree of association. The extraction unit 002 calculates second similarity between the related word 031 and the candidate images 011, and corrects the first similarity on the basis of the second similarity. An output unit 004 outputs a candidate image 011 with the highest corrected first similarity.SELECTED DRAWING: Figure 1 【課題】評価単語との類似度が最も高い画像を出力する際に、画像間の類似度の差が少なくても、出力する画像を決めやすくする。【解決手段】データ処理方法では、複数の候補画像011と評価単語013とを入力部001に入力し、入力した候補画像011のうち、評価単語013との第1の類似度が最も高い候補画像011を出力する。抽出部002が、候補画像011間での評価単語013との第1の類似度の第1の差分を算出し、第1の差分が第1の閾値以下である場合に、関連単語探索部003が、評価単語013を含む複数の文章から、その文章において、評価単語013に対して第1の関連度以上の関連度で登場する関連単語031を探索する。抽出部002が、複数の候補画像011と関連単語031との第2の類似度を算出し、第2の類似度に基づいて第1の類似度を補正し、出力部004が、補正後の第1の類似度が最も高い候補画像011を出力する。【選択図】図1