DeepQA System Unfamiliar Measurement Unit Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep question answering systems often fail to generate accurate answers when encountering unfamiliar measurement units, leading to incomplete or incorrect responses due to their reliance on restrictive type information from lexical answer type detection and answer generation modules.
Innovation Solution
A method is introduced to retrain a deep question answering system to recognize and match unfamiliar measurement units by associating them with known units, allowing the system to include both units in its answer retrieval process, thereby improving answer accuracy and relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the DeepQA system relies on restrictive type information from lexical answer type detection, then the system structure remains simple, but the system fails to recognize unfamiliar measurement units leading to inaccurate answers
Solution Approach 1:
The system performs preliminary action by training a measurement unit matcher before answering questions. This matcher learns to associate unfamiliar measurement units with known units using training data, enabling the system to recognize and handle unfamiliar units when processing questions, thereby improving answer accuracy without increasing operational complexity
Solution Approach 2:
The measurement unit matcher acts as an intermediary component between the lexical answer type detection and the answer generation modules. It bridges the gap by translating unfamiliar measurement units into recognizable formats, allowing the existing system architecture to handle diverse units without requiring complete structural redesign
2Adaptability or versatility
If the system uses only known measurement units from training data, then the system operates reliably within its training distribution, but it cannot handle unfamiliar measurement units encountered in production questions
Solution Approach 1:
The system implements feedback through confidence score weighting. When the measurement unit matcher encounters an unfamiliar unit, it generates confidence scores indicating the likelihood of unit equivalence. This feedback mechanism allows the system to adaptively weigh different answer candidates, maintaining reliability by relying more on high-confidence matches while still considering unfamiliar units
Solution Approach 2:
The system changes parameters by adjusting confidence score weights based on measurement unit match quality. When unfamiliar units are detected, the system modifies the weighting parameters to account for uncertainty, dynamically balancing between sticking to known units and exploring unfamiliar ones, thereby improving adaptability while preserving reliability
3Loss of information
If the system retrieves passages with both known and unfamiliar measurement units, then the system can provide more comprehensive answers, but the confidence scoring becomes more complex
Solution Approach 1:
The system segments the confidence scoring process into distinct components: one for evaluating known measurement unit matches and another for evaluating unfamiliar unit matches. This segmentation allows each component to use appropriate scoring strategies independently, reducing overall complexity while enabling comprehensive answer retrieval that incorporates both familiar and unfamiliar units
Data Source
AI summary
A method utilizes a deep question answering (QA) system to provide an answer, to a certain type of question, that includes an unfamiliar measurement unit. An answer key is utilized to train a DeepQA system to search for passages that answer a certain type of question, where the DeepQA system outputs an answer key value and an answer key measurement unit that is associated with the answer key value. The method identifies a candidate answer that includes a candidate passage containing the answer key value but not the answer key measurement unit, where a candidate passage measurement unit in the candidate passage is associated with the answer key value. The method then matches the answer key measurement unit to the candidate passage measurement unit based on the answer key measurement unit and the candidate passage measurement unit both being associated with the answer key value.