A method and system for generating dialogue summaries based on comprehensive feature extraction

Abstract

The invention discloses a method for generating a dialogue summary based on comprehensive feature extraction, which relates to the technical field of information processing, comprising: acquiring dialogue data; performing part-of-speech tagging and named entity recognition on the dialogue data to obtain candidate words; using the dialogue data to The mathematical statistics method is used for feature extraction to obtain word features; the word features are spliced ​​to obtain a feature vector; according to the feature vector and the candidate word, an unsupervised algorithm is used to obtain the first keyword; according to the feature vector and the described The candidate words are obtained by a supervised algorithm to obtain a second keyword; and a dialog abstract is generated according to the first keyword and the second keyword. The method and system provided by the present invention can improve the generation quality of dialogue summaries of complex dialogue scenes.

Description

A method and system for generating dialogue summaries based on comprehensive feature extraction technical field The present invention relates to the technical field of information processing, particularly relate to a kind of dialogue summary based on comprehensive feature extraction [0001] method and system. Background technique [0002] At present, for the automatic summarization problem, researchers mainly adopt the methods of machine learning, deep learning and feature engineering. Some researchers use the unilateral features of frequency, emotion, and semantics, and use SVM and decision tree to extract dialogue summaries. Decision trees outperform SVMs in chat summarization tasks; some scholars have proposed a Pointer‑Generator network to solve summarization Some scholars have proposed a SummaRuNNer network model based on RNN to solve the problem of document summarization; The authors proposed a Sentence-Gated model based on dialogue action selection; in addition...

AI Technical Summary

Problems solved by technology

In recent years, research has mainly focused on deep learning methods and achieved good results. However, compared with classic machine learning methods (such as K-nearest neighbors, decision trees, etc.), the training and testing complexity of deep learning methods is higher. For the case of high data complexity, the generation quality of dialogue summarization is not ideal

Images