Goal-Oriented Dialog Automation with Entity Tagging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
State-based chatbots face scalability issues due to the complexity of managing the universe of possible responses, making it difficult to achieve truly natural language interactions.
Innovation Solution
A method for goal-oriented dialog automation that involves entity tagging, semantic frame extraction, entity interpretation, accessing a database for business schedule and client profile, and using a retrieval engine to generate ranked response templates, with a candidate eliminator to provide recommended responses associated with confidence scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If state-based models are used to build chatbots, then natural language interactions can be achieved, but the system becomes difficult to scale due to complexity of managing the universe of possible responses
Solution Approach 1:
The patent segments the complex response generation task into distinct modules: a retrieval engine that fetches candidate responses from a database, a ranking model that scores candidates based on relevance, and a selection mechanism that chooses the best response. This modular segmentation allows each component to handle a specific aspect of the problem independently, making the overall system more manageable and scalable while maintaining natural language interaction capabilities.
2Adaptability or versatility
If the universe of possible responses is constantly growing, then the system can handle more variety, but the system hits a barrier in terms of functionality manageability
Solution Approach 1:
The patent introduces a ranking model as an intermediary between the large database of possible responses and the final selected response. This intermediary component scores and filters candidates based on relevance to the input, acting as a mediator that manages the complexity of the growing response universe. The ranking model handles the variety of responses systematically, maintaining ease of operation even as the response database expands.
3Measurement precision
If more response templates are generated and ranked, then response relevance improves, but processing time and computational resources increase
Solution Approach 1:
The patent implements partial action by generating and ranking only a limited subset of candidate responses rather than processing the entire universe of possible responses. The retrieval engine fetches a manageable number of candidates, and the ranking model evaluates only these candidates to select the best response. This approach achieves sufficient response relevance accuracy while avoiding the excessive processing time that would result from evaluating all possible responses.
Data Source
AI summary
In one aspect, method useful for goal-oriented dialog automation comprising includes the step of receiving an input message. The method includes the step of implementing an entity tagging operation on the input message. The method includes the step of tagging the message context of the input message to generate a tagged message context. The method includes the step of implementing semantic frame extraction from the tagged message context. The method includes the step of implementing an entity interpretation on the extracted frame. The method includes the step of accessing a database to determine a business schedule and a client profile. The business schedule and the client profile are related to the input message. The method includes the step of implementing a retrieval engine. The retrieval engine obtains one or more response templates. The method includes the step of generating a ranked list of candidate templates from the output of the retrieval engine. Based on the output of the entity interpretation, the business schedule and the client profile, and the ranked list of candidate templates, implementing a candidate eliminator. Based on the output of the candidate eliminator, providing a set of recommended responses. Each recommend response is associated with a confidence score.