Imaginary Discourse Trees for Autonomous Agent Question Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for autonomous agents are unable to provide complete and accurate answers to complex, multi-sentence, or convergent questions as they often rely on ontologies that are difficult and expensive to build, and knowledge graph-based approaches that limit expressiveness and coverage.
Innovation Solution
The use of imaginary discourse trees that represent a combination of an initial answer and additional answers from a corpus of text, connecting missing entities in the question with entities in the answer to form a complete answer, thereby improving question-answering recall without the need for ontologies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If knowledge graph-based approaches are used to answer questions, then answer accuracy is improved, but coverage and expressiveness are limited
Solution Approach 1:
The patent segments the answer generation process into multiple independent components: initial answer retrieval from knowledge graphs, discourse tree construction from retrieved answers, imaginary discourse tree generation from corpus text, and rhetorical relation identification. This segmentation allows each component to specialize while the system as a whole achieves both accuracy and coverage
Solution Approach 2:
The patent introduces discourse trees as an intermediary representation between the knowledge graph and the final answer. The discourse tree captures rhetorical relationships and enables the system to bridge gaps between initial answers and complete answers, thereby expanding coverage while maintaining accuracy through structured reasoning
2Loss of information
If ontologies are built to improve question-answering capability, then answer completeness is improved, but system complexity and cost increase
Solution Approach 1:
The patent enables the system to automatically construct discourse trees and identify rhetorical relationships without requiring manual ontology engineering. The system self-services by retrieving text from corpora, automatically parsing discourse structures, and generating imaginary discourse trees, thereby achieving answer completeness without the complexity and cost of manual ontology building
Solution Approach 2:
The patent copies discourse structures from existing text corpora rather than building ontologies from scratch. By extracting and replicating rhetorical relationship patterns from natural language text, the system achieves comprehensive answer coverage while avoiding the complexity of formal ontology construction
3Speed
If single resources are queried to answer questions, then response speed is improved, but answer completeness deteriorates
Solution Approach 1:
The patent performs preliminary actions by first retrieving an initial answer from a knowledge graph or single resource, then using discourse tree analysis to identify missing entities. This preliminary step allows the system to quickly get partial answers while systematically identifying what additional information is needed, balancing speed and completeness
Solution Approach 2:
The patent maintains continuity of useful action by seamlessly transitioning from initial answer retrieval to discourse tree construction to imaginary discourse tree generation. The system continuously refines the answer by identifying rhetorical relationships and retrieving additional text, ensuring the useful action of answer generation continues until completeness is achieved without significant delay
Data Source
AI summary
Systems and methods for improving question-answering recall for complex, multi-sentence, convergent questions. More specifically, an autonomous agent accesses an initial answer that partly answers a question received from a user device. The agent represents the question and the initial answer as discourse trees. From the discourse trees, the agent identifies entities in the question that are not addressed by the answer. The agent forms an additional discourse tree from an additional resource such as a corpus of text. The additional discourse tree rhetorically connects a non-addressed entity with the answer. The agent designates this discourse tree as an imaginary discourse tree. When combined with the initial answer discourse tree, the imaginary discourse tree is used to generate an improved answer relative to existing solutions.