Multi-party Conversation Node List for Answer Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing language processing technologies face challenges in accurately answering questions within multi-party online conversations due to the implicit context and complex structure of such interactions, which differ significantly from single-author texts, leading to incomplete or incorrect answers.
Innovation Solution
The method involves parsing multi-party conversations into elements, constructing a conversation node list that identifies relationships between nodes using 'reply-to' relationships, and employing deep learning models to rank potential answers based on contextual information, thereby untangling contextual streams and improving answer accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing language processing technologies are used to answer questions in multi-party online conversations, then the processing is simple, but the answer accuracy is low due to implicit context and complex structure
Solution Approach 1:
The patent segments multi-party conversations into discrete nodes representing individual messages or turns, allowing the system to analyze and track contextual relationships between different participants and messages. This segmentation enables accurate identification of question-answer pairs while maintaining the complex multi-party structure.
Solution Approach 2:
The patent introduces a conversation node list as an intermediary data structure that captures relationships between conversation nodes. This intermediary representation bridges the gap between raw conversation data and question answering, enabling accurate context tracking without directly processing the full complexity of multi-party interactions.
2Measurement precision
If the conversation structure is fully analyzed to improve answer quality, then the answer accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary parsing of the multi-party conversation into structured nodes and constructs the conversation node list before question answering. This preliminary structuring organizes the data in advance, enabling faster and more accurate question answering without reprocessing the full conversation structure during the actual querying phase.
3Measurement precision
If deep learning models are employed to rank potential answers, then the answer quality improves, but the computational complexity increases
Solution Approach 1:
The patent applies deep learning models locally to rank potential answers based on specific contextual features extracted from the conversation node list, rather than processing the entire conversation structure globally. This localized application of complex models reduces overall computational complexity while maintaining high answer quality.
Data Source
AI summary
A computer-implemented method for determining an answer to a question in a multi-party conversation includes receiving a multi-party conversation having multiple nodes of unstructured natural language. Each node is parsed into a plurality of elements. Each element of the plurality of elements that comprises a question is identified. A conversation node list is constructed that identifies relationships between the nodes. At least one answer to the question is produced based on the conversation node list.