Custom Named Entities for Database Natural Language Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional natural language processing (NLP) technologies for database environments are inefficient due to the lack of customized named entities, requiring domain experts and machine learning approaches that yield brittle results across different knowledge domains, and are not effectively linked to the underlying database.
Innovation Solution
The development of custom named entities from a structured search model associated with a database environment, with unique identifiers and tags, allows for the creation of a library that can process tokens from natural language queries to form structured search representations, enabling precise retrieval of records.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If general-purpose named entity libraries are used, then the system can process natural language queries, but the results are unsuitable for customized database environments due to lacking specific terminology and effective database linkage
Solution Approach 1:
The patent segments the named entity library into general-purpose entities and custom database-specific entities. The system extracts custom entities from the database schema (tables, columns, data types) and creates a specialized layer that maps database structure to natural language terminology, allowing the NLP system to adapt to specific database environments while maintaining general processing capabilities
Solution Approach 2:
The patent introduces an intermediary layer consisting of custom named entity libraries that act as mediators between the general NLP processing system and the specific database environment. This intermediary translates database-specific terminology and structure into forms understandable by the NLP system, enabling accurate query processing without requiring direct integration of database schemas into the NLP engine
2Reliability
If domain experts are employed to create custom named entity libraries, then the libraries can be accurately tailored to database environments, but the process becomes tedious and inefficient when replicated for each database installation
Solution Approach 1:
The patent implements self-service by enabling the system to automatically extract custom named entities from the database schema without requiring domain experts. The system autonomously identifies tables, columns, data types, and relationships, and generates the custom named entity library automatically, eliminating manual intervention while maintaining accuracy
Solution Approach 2:
The patent changes the parameters of library creation from manual expert-driven processes to automated schema-driven processes. By using database metadata (schema information, data dictionaries, relationship definitions) as input parameters, the system transforms the library creation process into an automated procedure that can be replicated across different database installations without requiring expert involvement each time
3Productivity
If machine learning approaches are used to create custom named entity libraries, then the libraries can be generated automatically, but they require developing training data for each database installation and yield brittle results across different knowledge domains
Solution Approach 1:
The patent applies preliminary action by pre-extracting custom named entities directly from the database schema before any query processing occurs. The system prepares the custom named entity library in advance by analyzing the database structure, so that when queries are processed, the NLP system already has the appropriate terminology and mappings ready, eliminating the need for training data development
Solution Approach 2:
The patent inverts the conventional approach by instead of training machine learning models on query data to extract entities, it extracts entities from the database schema and uses them to guide query processing. This inversion shifts the source of truth from query examples to database structure, making the system adaptable to different database environments without requiring retraining
4Reliability
If structured search paradigms with precise syntax are used, then queries can be accurately executed, but personnel lacking training cannot effectively implement them
Solution Approach 1:
The patent introduces custom named entity libraries as intermediaries that translate natural language queries into structured search representations. The system acts as a mediator between the user's natural language input and the database's structured query requirements, automatically performing the translation while maintaining accuracy through schema-based entity recognition
Solution Approach 2:
The patent substitutes the mechanical process of manually constructing structured queries with an automated natural language processing system. Instead of requiring users to manually syntax-parse and construct queries according to database-specific rules, the system automatically processes natural language input and generates the appropriate structured search representation through NLP and entity matching
Data Source
AI summary
Methods and apparatus are disclosed for building a library of custom named entities for a database environment and using the library for processing natural language search queries. At configuration time, custom named entities are extracted or derived from a search model or the database environment. Records for the custom entities with associated database unique identifiers and tags are stored in a library. Custom entities can be based on labels of database objects, variants thereof, or domain values. At search time, a natural language query is tokenized and matched with custom entities from the library, and with other predefined named entities, to obtained structured search descriptors. For more efficient search, compound entities can be identified in the search string, comprising a custom entity and a value, or a custom entity and another token sequence. Variations and examples are disclosed.