Intelligent Question Answering Using Knowledge Graph Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing intelligent customer service systems in e-commerce are limited in providing answers that stimulate user desire to make purchases, as they often provide cold and targeted responses without incorporating contextual or external knowledge.
Innovation Solution
An intelligent question answering method and apparatus that uses posterior knowledge information, such as comment information, to polish and rewrite answers by determining a target object and attribute, obtaining answer and external knowledge paths from a knowledge graph, and inputting these paths into a neural network model for generating reply texts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional knowledge graph query methods are used to answer user questions, then the answer accuracy is improved, but the answer lacks natural language fluency and cannot stimulate user purchase desire
Solution Approach 1:
The patent introduces a neural network model as an intermediary between the knowledge graph and the user. The model takes structured knowledge paths as input and transforms them into natural language replies, bridging the gap between precise data retrieval and fluent communication. This mediator converts cold factual answers into warm, persuasive language that maintains accuracy while improving fluency.
Solution Approach 2:
The system changes the parameter of output format from structured data to natural language text. By adjusting the output parameter through neural network processing, the same underlying knowledge can be expressed in different linguistic styles, making the answer both accurate and fluent simultaneously.
2Measurement precision
If only answer knowledge path is provided from knowledge graph, then the response is precise and targeted, but the response lacks richness and cannot stimulate user desire to make purchases
Solution Approach 1:
The patent merges multiple knowledge paths (answer knowledge path and external knowledge path) into a unified response. By combining the direct answer with related external information, the system maintains precision while enriching the content with additional context, making the response both accurate and information-rich.
Solution Approach 2:
The system adds another dimension to the knowledge representation by incorporating external knowledge paths alongside the answer knowledge path. This multi-dimensional approach allows the system to maintain precise answering while simultaneously providing rich contextual information from different knowledge dimensions.
3Productivity
If intelligent customer service system uses standard question answering approach, then the service efficiency is improved, but the user experience and purchase stimulation are reduced
Solution Approach 1:
The patent introduces dynamics into the customer service system by using a neural network model that can adapt its output based on the input knowledge paths. The system maintains efficient automated processing while dynamically generating varied, context-appropriate responses that can stimulate different user reactions, thus improving both efficiency and adaptability.
Data Source
AI summary
An intelligent question answering method includes: determining, based on received question information, a target object and a target attribute corresponding to the question information; obtaining an answer knowledge path and an external knowledge path of the target object other than the answer knowledge path from a pre-established knowledge graph based on the target object and the target attribute, the answer knowledge path including target context information for describing the target attribute, and the external knowledge path including external context information for describing another attribute; inputting the answer knowledge path and the external knowledge path into a trained neural network model to obtain a reply text, a training corpus of the neural network model during training including at least comment information of the target object; and outputting the reply text.