Hierarchical Neural Network Entity Classifier for Unstructured Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multi-tenant cloud-based database systems, users face challenges in accurately and efficiently specifying additions or changes to data using structured approaches, particularly due to heterogeneous data stores with different schemas, which limits the effectiveness of named entity recognition (NER) using unstructured input.
Innovation Solution
The implementation of a computing system that supports unstructured input for entity recognition, utilizing neural network models and hybrid structures, including entity classifiers and intent services, to analyze natural language inputs and identify entities and intents within the database, allowing users to interact more naturally and efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If structured approaches are used for data entry in multi-tenant database systems, then data consistency and schema compliance are improved, but user efficiency and ease of operation deteriorate due to the complexity of heterogeneous schemas
Solution Approach 1:
The patent introduces an intermediary layer (the entity recognition system with confidence scoring) between the user's unstructured input and the structured database schema. This intermediary automatically interprets natural language input, identifies entities, and maps them to appropriate schema fields, thereby maintaining data consistency without requiring users to directly interact with complex schemas.
Solution Approach 2:
The system enables self-service by allowing users to input data in natural language without needing to understand or navigate the underlying heterogeneous schemas. The entity recognition system automatically performs the work of parsing, classifying, and structuring the data according to the appropriate tenant-specific schemas, making the system easy to use while maintaining reliability.
2Ease of operation
If entity recognition systems are implemented to improve user interaction, then ease of operation is improved, but system complexity increases due to the need for confidence scoring and multiple entity types
Solution Approach 1:
The patent segments the entity recognition process into distinct, manageable components: entity detection, entity classification, confidence scoring, and threshold-based routing. Each component handles a specific aspect of the processing, making the overall complex system easier to implement, maintain, and tune. The segmentation allows independent optimization of each module without affecting the entire system.
3Measurement precision
If confidence thresholds are set high to improve entity recognition accuracy, then measurement precision is improved, but productivity decreases due to increased user review requirements
Solution Approach 1:
The patent implements dynamic threshold adjustment where the confidence threshold is not fixed but can be adapted based on context, entity type, and tenant-specific requirements. The system can dynamically lower thresholds for well-established entity types while maintaining high thresholds for ambiguous cases, optimizing the balance between accuracy and productivity without requiring manual intervention for every entry.
4Ease of operation
If unstructured input is allowed to improve user efficiency, then ease of operation is improved, but data quality and reliability may worsen without proper entity recognition
Solution Approach 1:
The patent incorporates feedback mechanisms where the entity recognition system provides confidence scores back to the user or system, enabling quality control. When confidence is high, the structured data is automatically committed; when confidence is low or below the threshold, the system can prompt for review or alternative input, ensuring data quality is maintained while preserving the efficiency benefits of unstructured input.
Data Source
AI summary
A method for natural language processing includes receiving, by one or more processors, an unstructured text input. An entity classifier is used to identify entities in the unstructured text input. The identifying the entities includes generating, using a plurality of sub-classifiers of a hierarchical neural network classifier of the entity classifier, a plurality of lower-level entity identifications associated with the unstructured text input. The identifying the entities further includes generating, using a combiner of the hierarchical neural network classifier, a plurality of higher-level entity identifications associated with the unstructured text input based on the plurality of lower-level entity identifications. Identified entities are provided based on the plurality of higher-level entity identifications.