Sub-lexical Answer Type Pre-computation for Query Latency Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Question answer systems face inefficiencies in determining candidate answers due to the exponential increase in data, making open-ended database queries to filter candidate answers extremely costly and time-consuming, especially when ruling out candidate-type pairings in large ontology graphs.

Innovation Solution

The system identifies sub-lexical answer types (sub-LATs) corresponding to the initial lexical answer type (LAT) and performs fast node-edge-node queries to rank candidate answers based on type-coercion scoring, reducing the computational burden by grouping similar queries and sharing intermediate results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If open-ended database queries are used to filter candidate answers through ontology graph traversal, then comprehensive type matching is achieved, but query time increases exponentially

Engineering Contradiction:

Improvetype matching accuracyVSAvoidquery time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system pre-computes and stores type-coercion scores for candidate answer-type pairings in a database before they are needed for query processing. This preliminary action transforms the expensive runtime ontology traversal into a pre-computed lookup operation, achieving both comprehensive type matching and fast query response times

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system pre-identifies and stores disjunctive sub-lexical answer types (sub-LATs) for each LAT in the ontology graph before query processing. This allows the system to quickly retrieve relevant sub-LATs without performing expensive ontology traversals during actual query execution, significantly reducing query time while maintaining matching accuracy

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If the number of candidate answers increases to cover more data, then answer comprehensiveness improves, but type-coercion scoring becomes exponentially more expensive

Engineering Contradiction:

Improveanswer coverageVSAvoidscoring efficiency

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The system pre-computes type-coercion scores for all possible candidate answer-type pairings and stores them in a database. This allows the system to handle exponentially increasing numbers of candidate answers without proportionally increasing scoring costs, as the expensive computation is performed once during pre-processing rather than repeatedly during query processing

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates a pre-computed copy of type-coercion scoring results in a database, allowing rapid retrieval during query processing. This copying approach enables the system to maintain high scoring efficiency even as the number of candidate answers grows exponentially

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11132390B2Efficient resolution of type-coercion queries in a question answer system using disjunctive sub-lexical answer types

Publication Date: 2021.09.28 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

An approach is provided in which an information handling system identifies candidate answers from a knowledge base that correspond to a question comprising a lexical answer type (LAT). The information handling system determines sub-LATs corresponding to the LAT in a database that are child LATs of the LAT. The information handling system then identifies a set of the candidate answers that correspond to at least one of the sub-LATs and ranks the set of candidate answers based on one or more ranking criterion. In turn, the information handling system provides the ranked set of candidate answers to a user.