Slot Value Extraction Using Trigger Words and Gold Sentences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing slot filling methods are restrictive, requiring explicit mention of target entities in multiple documents, reliance on large amounts of manually tagged training data, and inability to handle descriptive documents where entities are mentioned implicitly or with aliases, limiting their applicability in certain applications.
Innovation Solution
A method that splits descriptive documents into sentences, filters candidate sentences containing relevant information, calculates similarity measures with gold sentences to determine optimal candidate sentence-entity pairs, and assigns slot values, reducing processing time and reliance on manually tagged data by using trigger words and generic gold sentences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional slot filling methods are used requiring explicit entity mentions and multiple documents, then extraction reliability is improved, but applicability to descriptive documents with implicit entities deteriorates
Solution Approach 1:
The method segments the extraction task into distinct phases: candidate sentence identification using trigger words, candidate entity extraction, and similarity-based selection. This segmentation allows handling of implicit entities by breaking down the complex recognition task into manageable steps that can process descriptive text effectively.
Solution Approach 2:
The patent introduces trigger words as intermediary elements that bridge the gap between implicit entity mentions and explicit entity recognition. These trigger words serve as mediators that guide the extraction process by marking relevant sections of descriptive text without requiring the target entities to be explicitly named.
2Measurement precision
If large amounts of manually tagged training data are used, then machine learning model accuracy is improved, but data preparation time and cost deteriorates
Solution Approach 1:
The method performs preliminary actions by pre-defining trigger word lists and gold sentence sets that capture domain-specific relationships. This preliminary preparation reduces the need for extensive manual tagging during actual extraction, as the framework is pre-configured with knowledge about relevant entities and relationships.
Solution Approach 2:
The patent uses gold sentences as templates or copies of ideal extraction patterns. Instead of requiring numerous manually tagged training examples, the system creates representative gold sentences that encode the desired extraction behavior, which can then be reused across multiple extraction tasks.
3Productivity
If sentence filtering is performed to generate candidate sentences, then processing time is reduced, but extraction completeness may deteriorate
Solution Approach 1:
The filtering process applies local quality by focusing computational resources on sentences containing trigger words, which are locally identified as relevant to the extraction task. This localized approach maintains processing efficiency while preserving completeness for relevant information, as trigger words serve as reliable indicators of candidate sentences containing target entities.
Data Source
AI summary
Methods, systems, and computer program products for extracting from a descriptive document the value of a slot associated with a target entity described in the descriptive document are disclosed. The descriptive document is split into a set of sentences, and the sentences are filtered to generate a set of candidate sentences. Candidate sentence-entity pairs are determined from the candidate sentences, where each candidate sentence-entity pair includes a candidate sentence and an associated entity of the same type as the slot. The candidate sentence-entity pairs are compared to a set of gold sentences associated with the slot to calculate a plurality of similarity measures. A candidate sentence-entity pair associated with a maximum similarity measure is determined, and the value of the associated entity of the determined candidate sentence-entity pair is assigned to the slot.