Conversation Recommendation System Using Intent-Entity and QA Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current conversation systems based on Intent-Entity models require significant computational resources and training data to derive recommendation candidates, especially when dealing with complex queries, and often struggle to provide accurate outputs due to unpredictable user inputs and limited training data.
Innovation Solution
A recommendation system that integrates an Intent-Entity model with a Question-Answering system, using a QA reliability score to determine the appropriateness of recommendations, combining outputs from both systems to provide robust and accurate suggestions, and incorporating a QA reliability calculator to evaluate the confidence of recommendations based on user inputs and system outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an Intent-Entity model is used to provide recommendation candidates, then the system can handle automatic conversation with users, but the system requires enormous computational resources and training data
Solution Approach 1:
The patent segments the recommendation system into two parts: an Intent-Entity model for basic conversation handling and a Question-Answering system for complex queries. By dividing the system, the Intent-Entity model only needs to handle simple cases, reducing its training data and computational requirements while maintaining conversation capability.
Solution Approach 2:
The patent introduces a Query Complexity Evaluator as an intermediary component that assesses whether a user query is simple or complex. This evaluator mediates between the Intent-Entity model and the Question-Answering system, routing simple queries to the Intent-Entity model and complex queries to the QA system, thereby optimizing resource utilization.
2Measurement precision
If the Intent-Entity model is improved to handle complicated conditions in queries, then recommendation accuracy improves, but the amount of training data required increases hugely
Solution Approach 1:
The patent merges the Intent-Entity model with a Question-Answering system to handle complex queries. The QA system is pre-trained on general knowledge and can handle complicated conditions without requiring the Intent-Entity model to be retrained on extensive additional data, thus maintaining accuracy while avoiding huge training data requirements.
Solution Approach 2:
The Question-Answering system serves as a universal component that can handle various types of complex queries across different domains. By using a pre-trained QA system, the patent avoids the need to collect and train on domain-specific complex query data, reducing the training data burden while maintaining the ability to handle complicated conditions.
3Reliability
If the system uses both Intent-Entity model and Question-Answering system, then recommendation reliability improves, but device complexity increases
Solution Approach 1:
The patent implements a dynamic routing mechanism where the Query Complexity Evaluator determines in real-time whether to use the Intent-Entity model or the Question-Answering system. This dynamic approach allows the system to adapt to each query's complexity, ensuring reliable recommendations while managing complexity through intelligent routing rather than always activating all components.
Solution Approach 2:
The system incorporates feedback through the Query Complexity Evaluator, which assesses each incoming query and provides feedback on the appropriate processing path. This feedback mechanism ensures that the more complex QA system is only activated when necessary, maintaining reliability for complex queries while avoiding unnecessary complexity for simple queries.
Data Source
AI summary
A computer-implemented method, computer program product, and system are provided for providing a recommendation. The method includes receiving, by a conversation system, a query from a user. The method also includes inputting the query into an intent-entity model to obtain a first output. The method further includes inputting the query into a question-answering system to obtain a second output. The method additionally includes analyzing the first output from the intent-entity model to calculate a value of degree of ambiguity of the first output to the query. The method also includes analyzing the query and the second output from the question-answering system to calculate a value of extent of satisfaction to which the second output satisfies the query. The method further includes determining one or more recommendations using the degree of ambiguity and the extent of satisfaction. The method additionally includes outputting the one or more recommendations to the user.