Cross-modal bilateral personalized man-machine social conversation generation method and system

Abstract

The invention discloses a cross-modal bilateral personalized man-machine social conversation generation method and system. The method comprises the following steps: performing weighted fusion on a dialogue context code, a robot personalized information code, a user personalized information code and a code of an output result at the previous moment to obtain a weighted fusion code; inputting the weighted fusion code and the code of the output result at the previous moment into a decoder of the bilateral personalized generation model to generate N optimal candidate reply lists, and selecting thecandidate reply with the maximum conditional mutual information abundance value as the final output result. Personalized information is fused in a cross-modal mode, personalized information of figures of two interaction parties is considered, personalized features of the two interaction parties are fully utilized on the premise that reasonable reply content, smooth grammar and coherent logic areguaranteed, and replies which are rich in personality and differ from one another can be generated.

Description

technical field [0001] The invention relates to the technical field based on human-computer interaction, in particular to a method and system for generating a cross-modal bilateral personalized human-computer social dialogue. Background technique [0002] With the advancement of science and technology, human-computer interaction is gradually developing toward intelligence and personalization, and the interaction between humans and robots is getting closer to the interaction between humans in the real world. The traditional human-machine social dialogue generation belongs to the field of natural language processing, which mainly studies the ability of robots to make natural responses according to the user's text input. Different from interpersonal communication, vision is the main sensory source for people to receive external information, and people can make natural and personalized expressions based on external information. Therefore, in order to make the robot more "human-...

AI Technical Summary

Problems solved by technology

[0005] However, in real social interactions between people, both parties can learn each other's personalized information. When replying, the respondent should not only focus on his own personalized expression, but also consider the other party's personalized characteristics and the other party's information. Reply to the questions, ignore the human-computer interaction of the user's personalized information, which will make people feel disgusted and disgusted, and reduce the user experience

Images