Abstract Question Answering via Concrete Batch Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional question and answer systems are inadequate in handling general, abstract questions, often failing to provide a variety of factual responses that satisfy the user's intent, as they are designed to process specific, concrete queries, and struggle to model open-domain knowledge effectively.
Innovation Solution
A computer-implemented method and system that recognizes broad or abstract questions, generates a batch of concrete questions to address the abstract query, and returns a set of scored answers, utilizing a cognitive system with abstraction scoring and concrete alternative generation to provide relevant responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the system processes only single specific questions as conventional systems do, then the processing simplicity is maintained, but the ability to answer general abstract questions is insufficient
Solution Approach 1:
The system segments a general abstract question into multiple specific concrete questions. For example, the abstract question 'What are the beliefs of the organization?' is segmented into specific questions like 'What is the dominant belief system in the organization?' and 'What are the core values of the organization?'. This segmentation allows the system to handle abstract questions by breaking them down into manageable specific questions that can be answered using conventional question-answer processing methods.
Solution Approach 2:
The system performs preliminary action by generating a batch of specific questions before actually answering the abstract question. The question batch generator creates multiple concrete question variants based on the abstract question, and these questions are processed in advance to gather relevant information. This preliminary question generation enables the system to comprehensively address abstract questions by preparing multiple specific inquiry angles before final answer synthesis.
2Measurement precision
If the system generates a batch of specific questions to answer general questions, then the range and accuracy of responses are improved, but the processing time and computational resources increase
Solution Approach 1:
The system applies partial action by generating a limited batch of specific questions rather than exhaustively covering all possible angles. The question batch generator creates a manageable number of specific questions (e.g., 3-5 questions) that are sufficient to capture the essence of the abstract question without overwhelming computational resources. This partial action approach balances answer comprehensiveness with processing efficiency.
Solution Approach 2:
The system uses copying by generating multiple variants of specific questions based on templates and patterns. Instead of creating entirely new questions for each abstract query, the system copies and adapts proven question patterns from its knowledge base. For example, it might copy the pattern 'What is the [attribute] of the [entity]?' and fill in different attributes and entities based on the abstract question context, significantly reducing the computational effort required for question generation.
3Adaptability or versatility
If the system returns multiple scored answers from specific questions, then the user gets varied factual responses satisfying different interpretations, but the answer selection and presentation complexity increases
Solution Approach 1:
The system implements feedback by scoring and ranking multiple candidate answers from the batch of specific questions. Each answer is evaluated based on its relevance, accuracy, and ability to address the original abstract question. The system provides feedback through scored rankings that indicate the quality of each answer, allowing users to see which answers are most reliable. This feedback mechanism helps manage answer complexity by automatically filtering and prioritizing the best responses.
Solution Approach 2:
The system applies parameter changes by transforming multiple specific answers into a unified ranked output based on scoring parameters. Each answer is assigned numerical scores across multiple dimensions (relevance, accuracy, completeness), and these parameters are used to sort and present the answers. This parameter-based approach simplifies the presentation of varied answers by providing a standardized ranking system that makes it easy for users to identify the most valuable responses among multiple options.
Data Source
AI summary
Embodiments are directed to interaction with an open-domain question and answer system by recognizing questions that are highly broad or abstract, and generating and processing a batch of questions expressing alternate, concrete instances of the more general, abstract question. Responses to all of the questions in the batch are considered as candidates, and the strongest general answers are returned. A weighted, ranked answer set, based on weighting individual concrete questions and scaling the answers proportional to the weight of the questions, is provided to the user. The approach, according to embodiments herein, addresses the need for responses to broad questions in which a response to any of a set of more concrete question instances may serve to answer the question.