Semantic Vector Lookup for NAICS Code Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face frustration when trying to supply information requested by software applications, such as NAICS codes, due to lack of knowledge about the code system and inadequate lookup methods like manual tables and TF-IDF techniques.
Innovation Solution
A method utilizing a large language model to generate query and lookup vectors, which are then compared using a semantic matching algorithm to find the correct NAICS code from a lookup table, even with indirect references.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual lookup tables are used to find NAICS codes, then the system can provide structured code lookup, but the system fails to return correct codes when users provide indirect or unconventional descriptions
Solution Approach 1:
The patent introduces semantic vectors as an intermediary between user queries and the NAICS code lookup table. Instead of directly matching user input against the lookup table, the system converts both the query and lookup table entries into semantic vectors, then uses vector similarity to find matching codes. This intermediary representation enables the system to handle indirect references and synonyms effectively.
Solution Approach 2:
The patent transforms the lookup process from exact string matching to semantic similarity comparison by changing the parameter representation from discrete text strings to continuous semantic vectors. This parameter transformation allows the system to capture nuanced meanings and relationships between different descriptions of the same business activity.
2Adaptability or versatility
If TF-IDF and semantic matching algorithms are used for lookup, then the system can handle varied user input, but the system fails when training data is unavailable or when specific wording causes confusion
Solution Approach 1:
The patent pre-computes semantic vectors for all NAICS code descriptions and stores them in the lookup table before runtime queries. This preliminary action ensures that the semantic representations are ready when needed, eliminating the need for runtime training or data collection. The system is pre-equipped to handle varied user input through pre-established semantic relationships.
3Measurement precision
If exact matching is used in lookup tables, then the system can ensure precise code retrieval, but the system cannot handle terms that do not appear in the NAICS code system
Solution Approach 1:
The patent replaces the mechanical exact-matching system with a semantic similarity-based system. Instead of requiring exact string matches between user input and lookup table entries, the system uses vector space models to measure semantic proximity. This substitution allows users to input flexible, natural language descriptions while maintaining precise code retrieval through semantic equivalence.
Data Source
AI summary
A method including applying a large language model to a query to generate a query vector. The query vector has a query data structure storing a semantic meaning of the query. The method also includes applying a semantic matching algorithm to both the query vector and a lookup vector. The lookup vector has a lookup data structure storing semantic meanings of entries of a lookup table. The semantic matching algorithm compares the query vector to the lookup vector and returns, as a result of comparing, a found entry in the lookup table. The method also includes looking up, using the found entry in the lookup table, a target entry in the lookup table. The method also includes returning the target entry.