Domain-Specific NLP Classification for Pharmacovigilance Documents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for classifying and prioritizing documents, particularly in the field of pharmacovigilance, are burdensome, costly, and error-prone due to the complexity of document formats and the inefficiency of existing machine-learning algorithms, which struggle to accurately classify and prioritize individual case safety reports (ICSRs) from various sources.
Innovation Solution
A domain-specific natural language processing (NLP) model utilizing a Convolutional Neural Network (CNN) and Bidirectional Long-Term-Short-Term Memory (BiLSTM) algorithms is employed to tokenize, assign part-of-speech and dependency tags, and apply Named Entity Recognition (NER) labels to documents, enabling accurate classification and prioritization without manual transcription or translation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional machine-learning algorithms are used for document classification, then automation is achieved, but classification accuracy is insufficient
Solution Approach 1:
The patent introduces an intermediary NLP processing layer between document intake and classification. This intermediary layer performs specialized tokenization, part-of-speech tagging, dependency parsing, and NER labeling to transform raw documents into structured representations that conventional ML algorithms can process more effectively, thereby improving classification accuracy while maintaining automation.
Solution Approach 2:
The patent changes the parameter representation of documents from raw text to transformed features including word embeddings, part-of-speech tags, dependency labels, and NER labels. These parameter transformations enable the system to capture domain-specific semantic information that conventional ML models miss, significantly improving classification precision.
2Measurement precision
If manual extraction of text contents is performed, then extraction accuracy is high, but the process is time-intensive and error-prone
Solution Approach 1:
The system performs self-service by automatically extracting text contents through an integrated NLP pipeline that handles tokenization, parsing, and entity recognition without requiring manual intervention. The model processes documents independently, eliminating human labor while maintaining high extraction accuracy through sophisticated algorithms.
Solution Approach 2:
The patent replaces manual mechanical text extraction with automated computational NLP processing. The system uses algorithms for optical character recognition, tokenization, and named entity recognition to automatically extract and structure information from documents, eliminating the need for manual transcription while improving both speed and consistency.
3Productivity
If conventional NLP models are used, then processing speed is maintained, but domain-specific classification capability is insufficient
Solution Approach 1:
The patent applies local quality by implementing domain-specific NER labels and part-of-speech tags tailored to the specific domain (e.g., medical, legal, financial). The system tokenizes and processes text with domain-aware parameters, assigning specialized labels to entities and relationships relevant to the domain, thereby improving classification accuracy without significantly reducing processing speed.
Solution Approach 2:
The system performs preliminary action by pre-processing documents through tokenization, part-of-speech tagging, dependency parsing, and NER labeling before final classification. This preliminary transformation prepares the data in advance, enabling faster and more accurate domain-specific classification by conventional ML algorithms without requiring them to process raw text directly.
Data Source
AI summary
A method includes receiving a request to classify a document corresponding to a specific domain, generating a word embedding, and tokenizing the word embedding into a set of segments. The method also includes assigning a part-of-speech tag, a dependency tag, and a named entity recognition label to each corresponding segment in the set of segments. The method also includes classifying the document based on the named entity recognition labels assigned to the set of segments.