Knowledge-graph-based human-machine conversation control system and method
A technology of knowledge graph and man-machine dialogue, applied in the field of intelligent dialogue, can solve problems such as single form of guiding sentences, skipping topics, and inability to guarantee the fluency of topics, etc., and achieve the effect of humanized answers and good interaction
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
example 1
[0066] Example 1: The user's current conversation is "I don't like eating bananas, they are too sweet"; a pair of corpus is marked with Q and A. General corpus: question Q = "chocolate is too sweet, I don't like it", answer A = "yes, dark chocolate is not sweet, it is recommended", traditional text retrieval and text similarity calculation will match the user's current conversation to This Q, so the chatbot will reply this A to the user. Since there is no "knowledge graph library" to support the understanding of "banana" and "chocolate", although they are both in the food category and the word vectors are very close, they belong to different refined categories. Therefore, they will answer irrelevant questions, resulting in a poor dialogue experience.
[0067] After adopting the present invention, text retrieval and text similarity matching can be controlled through the knowledge map, and entry-level classification knowledge can be obtained: "banana belongs to a kind of fruit",...
example 2
[0071] Example 2: User dialogues include "I like the secret that movies can't be told, time-travel type", "Rainbow sounds very good, very lyrical". The intelligent chat robot will record that the user likes the movie "Unspeakable Secret" and the music "Rainbow". However, it cannot be probabilistically inferred that the user will like "Jay Chou", because the above movies and music are all his works, and the preferences derived from active questioning cannot be used for human-like chat, divergence and expansion, and active recommendation cannot be initiated. .
[0072] After adopting the present invention, in the knowledge map, such as figure 2 As shown, there is a relationship: the character "Jay Chou" is the director of the movie "The Unspeakable Secret" and the singer of the music "Rainbow". After querying the relationship between the two through the knowledge map, it is found that both are representative works of the character "Jay Chou". Combined with the calculation of ...
example 3
[0078] Example 3, the user mentions "help me check the weather in Shanghai", the intent recognition is "check the weather", and then trigger the weather function unit, the weather function needs to extract the information as the city name, this sentence can extract "Shanghai city". However, if the user says "check the weather in Xuhui District for me", the knowledge map control module of the present invention can provide the identification of entry category and hierarchical affiliation. In the knowledge map, "Xuhui District" is a location, which belongs to "Shanghai City" is a district, so it is deduced that the extracted city information is "Shanghai".
[0079] In a specific embodiment provided by the present invention, the task engine unit is configured to obtain a task answer statement according to the association information, the characteristic information and the current dialogue, and output the task answer statement to the user.
[0080] Among them, the task engine unit...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com