Natural-Language State Transition Extraction for ML Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning methods struggle to accurately determine appropriate features for state transitions in data sets representing event progress, as these transitions can follow diverse patterns and lack formal semantic definitions, leading to inefficiencies and inaccuracies in model generation.
Innovation Solution
An information processing method using a dialogue system with a language model to generate query texts that identify semantically unreasonable state transitions, allowing for the determination of features based on predefined conditions, and categorizing these transitions to reduce the number of features used in machine learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional machine learning methods are used to process state transition data, then the system can process data, but it cannot accurately determine appropriate features due to lack of formal semantic definitions and diverse transition patterns
Solution Approach 1:
A dialogue system with a language model is introduced as an intermediary component between the data processing system and the feature determination process. The language model processes natural language queries about state transitions and generates structured responses that guide feature extraction, enabling accurate feature determination without requiring complex formal semantic definitions in the original system.
2Reliability
If all possible state transitions are considered as features, then comprehensive information is captured, but the number of features becomes too large leading to inefficiency and overfitting
Solution Approach 1:
The dialogue system extracts and filters only the relevant state transition patterns that satisfy predefined conditions from the comprehensive data. By interacting with the language model, the system identifies and extracts only the meaningful features (e.g., transitions that indicate abnormal processes) while discarding redundant or irrelevant transitions, thus reducing the feature set size and improving processing efficiency without losing model accuracy.
3Measurement precision
If manual feature selection is performed, then appropriate features can be identified, but the process is time-consuming and labor-intensive
Solution Approach 1:
The system enables automated feature selection through a dialogue interface where users can naturally query the system about state transitions. The language model automatically processes these queries, analyzes the data, and generates feature extraction results without requiring manual intervention in the feature selection process. This self-service approach maintains high feature selection accuracy while dramatically reducing the time and labor required compared to manual methods.
Data Source
AI summary
An information processing apparatus generates a query text that instructs, in natural language, extraction of a first state transition that satisfies a predetermined condition regarding meanings of states before and after a transition from a data set indicating a series of transitions of states of a plurality of events. Next, the information processing apparatus inputs the query text to a dialogue system that performs a dialogue in natural language using a language model. Then, the information processing apparatus determines that, for each event, information based on whether the first state transition indicated in a response text output by the dialogue system is included in a series of transitions of states of the event is set as a feature of the event.