Method and system for building user portrait based on conversation interaction

Abstract

The invention discloses a method and a system for building a user portrait based on conversation interaction. The method comprises the steps of parsing a user input in real time by a semantic parser, and updating weighted distribution of a weight of each attribute according to parsed user semantics, thus generating the accurate and meticulous portrait. The system comprises a speech recognition module, a speech synthesis module, a semantic parsing module, a conversation management module, a natural language generation module and a user portrait generation module, wherein the speech recognition module is connected with the semantic parsing module and transmits text information with the same, the semantic parsing module is connected with the user portrait generation module and transmits the semantics and related time stamp information with the same, meanwhile the semantic parsing module is connected with the conversation management module and transmits semantic information with the same, the conversation management module is connected with the natural language generation module and transmits conservation action information with the same, and the natural language generation module is connected with the speech synthesis module and transmits the text information with the same. According to the method provided by the invention, the accurate user portrait is built, and thus a human-computer interaction process and experience are unified.

Description

technical field [0001] The present invention relates to a technology in the field of information processing, in particular to a method and system for constructing a user portrait based on dialog interaction. Background technique [0002] User portrait, also known as user role (Persona), is an effective tool to outline target users, connect user demands and design direction, and user portrait has been widely used in various fields. In the process of actual operation, we often use the most simple and life-like words to connect the user's attributes, behaviors and expectations. As a virtual representative of actual users, user personas formed by user portraits are not constructed out of products and markets. The formed user roles need to be representative and able to represent the main audience and target groups of the product. At present, although there are examples of applying user portraits in dialogue systems, the general method is still used, and the lack of special consi...

AI Technical Summary

Problems solved by technology

[0004] 1. The first method is the most commonly used, but ignores the nature of user attributes or preferences that change over time, making the portrayal of user portraits more inaccurate in the later stages;

[0005] 2. Although the second method considers the time factor, advancing the time window requires the accumulation of user logs, which makes the update lack real-time performance, and different time window settings will affect the effect of user portraits, making it difficult to adjust parameters;

[0006] 3. Both methods do not pay attention to the differences between different user attributes, and treat each attribute as the same, ignoring whether the attribute is changeable or has a specific time limit;

[0007] 4. Both methods are based on static logs for user portraits, which are difficult to apply in the case of dialogue and interaction, and cannot adapt to the logic and language characteristics of the interaction process;

[0008] 5. The current methods basically discover user attributes by marking and extracting key fields, without considering semantic characteristics, and the accuracy of user information obtained is often not high;

[0009] 6. The current method lacks the consideration of the specificity of user modeling in the dialogue system. Due to the interactive and dynamic characteristics of the dialogue, it is difficult for the traditional user modeling method to accurately process the user information in the dialogue;

[0010] 7. In the user portrait obtained by the current method, each attribute is basically a fixed value. It is impossible to know the historical information of the attribute only from this value, and it is also impossible to know how certain the value of this attribute is.

Images