Transaction Entity Prediction via Embedding Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated data entry techniques rely on text string matching, which is ineffective for identifying transaction entities when transaction data is not standardized, leading to errors and inefficiencies in software applications.
Innovation Solution
A machine learning-based method that generates embeddings for transactions and entities, using a global list to predict entities by calculating similarity scores and interaction probabilities, thereby improving entity identification and reducing errors in automated data entry.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If text string matching techniques are used for automated data entry, then the system is simple and easy to implement, but the accuracy of entity identification deteriorates when transaction data is not standardized
Solution Approach 1:
The patent transforms the entity identification problem from text string matching to embedding space similarity measurement. By converting transactions and entities into vector embeddings and computing cosine similarity, the system achieves robust matching that is insensitive to text formatting variations, thereby improving identification accuracy without requiring complex rule-based systems
Solution Approach 2:
The patent replaces the mechanical text string matching mechanism with a machine learning-based embedding similarity approach. This substitution enables the system to capture semantic relationships and contextual information, significantly improving entity identification accuracy for non-standardized transaction data
2Speed
If text string matching is used, then the processing speed is fast, but the reliability of entity identification deteriorates with diverse data formats
Solution Approach 1:
The patent performs preliminary encoding of transactions and entities into embeddings before the matching process. This pre-processing step transforms diverse text formats into a unified vector representation space, enabling fast and reliable similarity computation without requiring complex runtime processing for each data format variation
Solution Approach 2:
By changing the representation parameter from raw text strings to numerical embeddings, the system achieves both fast processing through vector operations and high reliability through semantic similarity measurement, resolving the contradiction between speed and reliability
3Use of energy by moving object
If conventional matching techniques are applied, then the system requires minimal computational resources, but the ability to handle non-standardized data deteriorates
Solution Approach 1:
The patent creates a universal embedding-based matching system that can handle diverse data formats through a single unified approach. The cosine similarity computation works effectively across different text formats, lengths, and styles, providing versatile data format adaptability while maintaining reasonable computational efficiency through vector operations
Data Source
AI summary
Certain aspects of the disclosure pertain to predicting a candidate entity match for a transaction with a machine learning model. A description of a transaction comprising encoded transaction data associated with an organization is received as input. In response, at least one machine learning model can be invoked to infer a transaction embedding based on the description, a first score that captures similarity between the transaction embedding entity embeddings associated with a global list of entities and organizations, a second score that captures a probability of interaction between the first organization and the entities based on organization and entity embeddings that capture profile data associated with the organization and the entities, and at least one candidate entity based on the first score and the second score. Finally, the inferred candidate entity can be output for use by an automated data entry or other process or system.