Training Data Generation via Rule-Based Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning systems face challenges in achieving sufficient accuracy for natural language processing tasks due to the lack of available training data, particularly in restricted domains like medical diagnosis and financial fraud detection, where access to data is limited.
Innovation Solution
A computer-implemented method that generates training data by using hand-coded rules and dictionaries, including general and domain-specific terminology, to create candidate statements that are filtered based on a domain-specific model, ensuring semantic validity and relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If hand-coded rules and dictionaries are used to generate training data, then training data availability is improved, but data quality and semantic validity may deteriorate
Solution Approach 1:
A domain-specific model acts as an intermediary between the generated candidate statements and the final training data. The model filters candidate statements to ensure they meet semantic validity requirements, thereby maintaining data quality while enabling large-scale data generation through rule-based methods.
Solution Approach 2:
The system uses domain-specific models and validation mechanisms that automatically assess and filter generated statements. This self-validation process ensures that only semantically valid statements are included in the training data, maintaining quality without requiring extensive manual review.
2Adaptability or versatility
If stochastic word insertions are used to create varied test cases, then training data diversity is improved, but risk of generating nonsensical combinations increases
Solution Approach 1:
The domain-specific model provides feedback by evaluating generated candidate statements and filtering out nonsensical combinations. This feedback mechanism allows the system to maintain high diversity through stochastic word insertions while ensuring reliability by eliminating invalid statements before they enter the training data.
3Measurement precision
If extensive filtering based on domain-specific models is applied, then training data accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
Domain-specific models and validation rules are prepared in advance before the actual data generation process begins. This preliminary preparation ensures that filtering operations during data generation are efficient and fast, reducing processing time while maintaining high accuracy standards.
Data Source
AI summary
A computer-implemented method includes receiving a rule, wherein the rule includes at least one token, and receiving at least two dictionaries, wherein the at least two dictionaries include at least one general language dictionary and at least one domain-specific dictionary for a domain. The computer-implemented method further includes, for each of the at least one token, selecting at least one word at random from at least one of the at least two dictionaries and adding the at least one word to a test data line, such that the test data line includes a candidate statement conforming to the rule. The computer-implemented method further includes filtering the candidate statement based on a domain-specific model for the domain and including the candidate statement in training data provided to a machine learning model. A corresponding computer program product and computer system are also disclosed.