Code-Mixed Language Model for Adverse Reaction Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting adverse medication reactions are inefficient, particularly in settings with limited healthcare infrastructure, as they struggle to timely assess and monitor medication use between clinic visits, leading to inadequate data collection and reporting of adverse effects.
Innovation Solution
An automated system and method using a hardware processor to create a language model corpus with multilingual alignment for training a combined language model, which generates code-mixed utterance models to perform turn-by-turn dialogue, analyzing data from online sources to identify potential adverse medication reactions through code-mixed conversations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional adverse reaction detection methods (spontaneous reporting, expert review, cohort monitoring) are used, then data collection can be performed with limited infrastructure, but timely assessment and monitoring of medication use between clinic visits is difficult
Solution Approach 1:
The patent replaces manual mechanical monitoring systems with an automated AI-based conversation system. The language model processes patient responses automatically, eliminating the need for manual review of case reports or cohort monitoring, thereby achieving timely assessment without requiring extensive infrastructure or human resources.
Solution Approach 2:
The system enables patients to self-report adverse reactions through automated conversations. The language model actively engages patients in turn-by-turn dialogue, guiding them to provide necessary information about medication side effects, thereby making the monitoring process self-service oriented and accessible without requiring healthcare provider involvement between visits.
2Productivity
If automated AI conversation systems are implemented, then timely assessment and monitoring capacity is improved, but the system complexity increases
Solution Approach 1:
The language model serves multiple functions within a single unified architecture: it processes patient responses, detects adverse reactions, extracts relevant information, and guides conversations. This multi-functionality reduces the need for separate specialized systems, thereby managing complexity while maintaining high productivity in monitoring capacity.
Solution Approach 2:
The system adapts to different language parameters (code-mixed languages, varying dialects, different expression patterns) through the language model's flexibility. By changing the parameter of language processing to accommodate diverse patient communications, the system maintains simplicity in its core architecture while improving its ability to monitor effectively across different populations.
3Adaptability or versatility
If code-mixed language processing is used to adapt to bilingual contexts, then adaptability to diverse patient populations is improved, but the difficulty of detecting and measuring adverse reactions increases
Solution Approach 1:
The patent segments the language processing into distinct components: the language model handles code-mixed language understanding and adaptation, while separate detection modules analyze for adverse reaction patterns. This segmentation allows the system to maintain high adaptability to bilingual contexts while simplifying the detection process through specialized processing stages that focus on identifying adverse reactions regardless of language mixing.
Solution Approach 2:
The language model acts as an intermediary between the patient's code-mixed language input and the adverse reaction detection system. It translates and processes the mixed language responses into a standardized format that the detection algorithms can accurately analyze, thereby maintaining both adaptability to diverse language patterns and detection accuracy for adverse reactions.
Data Source
AI summary
A computer system may create a language model corpus including multilingual alignment for training a combined language model and train (or pre-train) the combined language model. The computer system may create an adverse medication reaction corpus to include adverse medication reaction utterances and label an N-gram of an utterance in the adverse medication reaction utterances as a response to query, for multiple N-grams. The computer system may generate a code-mixed utterance model to perform code-mixed utterances in a turn by turn dialogue, by at least adding additional output layer including at least a start vector, language vector, and a query vector including at least the labeled N-gram, which are additional to the combined language model's predicted next words.