Composite AI Model for Multi-String Text Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analytics techniques face challenges in processing queries that require the extraction of multiple substrings from unstructured text attributes in records, such as those found in product catalogs, due to inconsistencies and the need for complex analysis, which can be time-consuming and resource-intensive.
Innovation Solution
A composite lightweight machine learning model comprising a classifier sub-model and an extractor sub-model is used to determine the unit measurement category and extract relevant substrings from text attributes, employing character-level embeddings and attention layers to improve robustness and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If straightforward text analysis is used for queries, then processing speed is fast, but accuracy of response is insufficient for multiple response-string queries
Solution Approach 1:
The patent segments the text analysis process into multiple specialized components: a classifier model that identifies query types and required substrings, and an extractor model that retrieves specific substrings. This segmentation allows each component to specialize in one task, improving overall accuracy while managing complexity through modular design.
Solution Approach 2:
The patent introduces an intermediary classification layer between the raw query and the text extraction process. The classifier acts as a mediator that analyzes the query, determines what information is needed, and guides the extractor to find the appropriate substrings. This intermediary step improves response accuracy by ensuring the right information is extracted.
2Reliability
If complex analytics techniques are combined to solve problems, then problem-solving capability is improved, but computational resources consumed increase
Solution Approach 1:
The complex analytics task is divided into two separate models with distinct responsibilities: classification and extraction. This segmentation allows each model to be optimized for its specific function, reducing the computational burden on each individual model while maintaining overall problem-solving capability.
Solution Approach 2:
The patent extracts only the necessary substrings from the text based on query requirements, rather than processing or analyzing the entire text corpus. This selective extraction approach significantly reduces computational resource consumption while maintaining the ability to answer complex queries accurately.
3Measurement precision
If multiple substrings are extracted from unstructured text, then query response quality is improved, but processing time increases
Solution Approach 1:
The classifier model performs preliminary analysis of the query to determine what substrings are needed before the extraction process begins. This preliminary action prevents unnecessary searching and extraction operations, reducing processing time while ensuring the right substrings are identified for high-quality responses.
Solution Approach 2:
The patent implements efficient extraction mechanisms that quickly locate and retrieve specific substrings based on classifier guidance. By extracting only the necessary information rather than processing entire text documents, the system maintains high response quality while minimizing processing time.
Data Source
AI summary
Query types for which responses are to be generated with respect to records comprising text attributes are identified, including a text interpretation query type for which records may comprise one or more response-contributor strings. Results of the text interpretation query for a record are based at least partly on an extracted-property class of the records. A machine learning model comprising a first sub-model and a second sub-model is trained to extract results of the text interpretation query. The first sub-model generates an extracted-property class for a record, and the second sub-model predicts positions of response-contributor strings within the record based at least in part on the extracted-property class. A trained version of the model is stored.