Invoice Account Code Recommendation with Historical Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face inefficiencies in automatically recommending account codes for non-PO-backed invoices, requiring manual effort and extensive retraining for each use case, which is resource-intensive and impractical.
Innovation Solution
A computer-implemented method using natural language processing and machine learning techniques to analyze invoice descriptions and patterns, comparing them with historical data to recommend accurate account codes through a similarity scoring system, displayed on a graphical user interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a supervised ML model with NLP is used to predict account codes, then account code prediction capability is improved, but the system complexity and resource requirements increase significantly
Solution Approach 1:
The system segments the account code prediction task into multiple independent components: NLP processing module for extracting semantic meaning from invoice descriptions, feature extraction module for converting text to numerical vectors, and classification module for predicting account codes. This segmentation allows each component to be optimized independently and reduces overall system complexity.
Solution Approach 2:
The patent introduces an intermediary feature extraction layer that converts NLP-processed text into numerical feature vectors, which then serve as input to the classification model. This intermediary layer simplifies the connection between NLP and ML components, making the system more manageable and easier to train.
2Measurement precision
If multiple supervised models are trained for different invoice types, then prediction accuracy for specific use cases is improved, but CPU time and infrastructure resources increase significantly
Solution Approach 1:
The patent implements a universal supervised ML model that can handle multiple invoice types and use cases through a single unified architecture. The model uses a shared feature extraction layer and classification head that can be fine-tuned for different invoice types without requiring complete retraining, thus maintaining high prediction accuracy while reducing computational resource requirements.
Solution Approach 2:
The system employs dynamic model fine-tuning where the classification parameters can be adjusted and refined for different invoice types without retraining the entire model from scratch. This dynamic adaptation allows the system to maintain high accuracy for specific use cases while avoiding the computational burden of training multiple separate models.
3Device complexity
If a single supervised model is trained for all invoice types, then infrastructure resources are reduced, but the model cannot be easily customized for new account codes without retraining
Solution Approach 1:
The patent implements a dynamic model where the classification parameters can be adjusted and refined for new account codes without retraining the entire model from scratch. The feature extraction layer remains fixed, and only the classification head is fine-tuned, allowing the system to adapt to new account codes efficiently while maintaining the benefits of a single unified model architecture.
4Reliability
If manual entry of account codes is performed, then accuracy and review capability are improved, but labor intensity and processing time increase
Solution Approach 1:
The system implements self-service automation where the NLP and ML model automatically extracts account code information from invoice descriptions without human intervention. The model processes invoice descriptions, extracts relevant features, and predicts appropriate account codes independently, thereby maintaining high accuracy while significantly reducing labor intensity and processing time.
Data Source
AI summary
A computer implemented method comprising receiving invoice data comprising at least one of invoice descriptions and invoice characters from user computers, each of the invoice descriptions and invoice characters defines contexts and patterns, wherein each of the invoice data comprising a supplier-customer information that corresponds to a supplier-customer transaction; analyzing the at least one of the invoice descriptions and the invoice characters with corresponding contexts and patterns; determining that amount of the invoice characters is more than a threshold number of characters, for performing: matching invoice data, invoice characters with predefined historical invoice data that corresponds to the same supplier-customer information; computing a similarity score for each of the invoice data; and displaying recommendations including first account codes to map the one or more first account codes to each of the one or more invoice data based on the similarity score; determining that amount of the invoice characters is not more than the threshold number of characters, for performing: filtering the invoice descriptions of the invoice data based on predetermined constraints to extract filtered invoice data comprising filtered description lines; identifying categories associated with each of the filtered description lines of the filtered invoice data; matching each of the identified categories, including corresponding contexts and patterns that is associated with a supplier-customer information with one or more predefined historical categories associated with predefined invoice description of the one or more predefined historical invoice data that corresponds to the same supplier-customer information, wherein each of the contexts and the patterns are matched with predefined contexts and patterns of the predefined historical invoice data that corresponds to the same supplier-customer information; computing a categorical similarity score for each of the categories associated with the supplier-customer information; and displaying on the graphical user interface, second recommendations including second account codes based on the computed categorical similarity score of each of the categories to map the second account codes to the invoice data.