Personalized Extractive Summarization for a News Dialogue System

In modern society, people's interests and preferences are diversifying. Along with this, the demand for personalized summarization technology is increasing. In this study, we propose a method for generating summaries tailored to each user's interests using profile features obtained from questionnaires administered to users of our spoken-dialogue news delivery system. We propose a method that collects and uses the obtained user profile features to generate a summary tailored to each user's interests, specifically, the sentence features obtained by BERT and user profile features obtained from the questionnaire result. In addition, we propose a method for extracting sentences by solving an integer linear programming problem that considers redundancy and context coherence, using the degree of interest in sentences estimated by the model. The results of our experiments confirmed that summaries generated based on the degree of interest in sentences estimated using user profile information can transmit information more efficiently than summaries based solely on the importance of sentences.

1. INTRODUCTION

With the widespread of conversational robots and smart speakers, the demand for spoken dialogue technology is increasing. One of the uses of these devices is to access and consume information like news through voice interaction. Compared to textual communication allowing the reader to read at his/her pace, auditory communication enables the speaker to control the content and speed of the conversation while the listener can neither skip unnecessary information nor pick up the only necessary information. Still, because auditory information transmission and reception are performed simultaneously, the listener can instantly supplement the information by asking the speaker. Therefore, in order to realize efficient information transmission in auditory communication, it is essential to develop spoken dialogue systems that can pick up and transmit information that users are interested in without excess or deficiency. In this study, we propose a method of personalizing and summarizing text that include a massive amount of information such as news article at the sentence level for each user.