User Action Sequence Labeling for Self-Learning Assistance Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user assistance technologies face challenges due to insufficient training on meaningful sequences or sub-sequences of user actions, limited horizon of recommendations, and lack of self-learning capabilities, leading to inefficient and error-prone learning experiences.
Innovation Solution
A method and system for labelling user action data by providing a pre-trained pattern recognition model to recognize and label sequences of user actions, using natural language processing to generate labels, and storing these in a database for continuous training, enabling improved pattern recognition and sequence understanding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labelling of user action sequences is performed to train assistance models, then training data quality improves, but labelling effort and time cost increase exponentially
Solution Approach 1:
The system performs self-service by using the pattern recognition model to automatically generate labels for user action sequences. The model analyzes logged user actions, identifies patterns, and creates training labels without requiring manual human labelling, thus eliminating the exponential growth of labelling effort while maintaining data quality.
Solution Approach 2:
The pattern recognition model acts as an intermediary between raw user action logs and training data. It processes the raw logged sequences and transforms them into labelled training data, serving as an automated mediator that eliminates the need for direct manual labelling while preserving meaningful patterns.
2Measurement precision
If assistance models are trained on comprehensive user action sequences, then prediction accuracy improves, but computational complexity and processing requirements increase
Solution Approach 1:
The system segments user action sequences into smaller, meaningful patterns that can be independently recognized and labelled. By breaking down comprehensive sequences into identifiable patterns, the model can train on structured data without being overwhelmed by the full complexity of entire user sessions, reducing computational burden while maintaining accuracy.
Solution Approach 2:
The system performs preliminary analysis by logging and storing user actions in structured formats before training the assistance model. This pre-processing and organization of data into labelled sequences prepares the training material in advance, reducing the computational complexity during the actual model training and prediction phases.
3Quantity of substance
If user action data is logged and stored for training purposes, then training data availability improves, but data privacy and security risks increase
Solution Approach 1:
The system extracts only the necessary pattern information from user action data for training purposes, rather than storing or processing complete raw user data. By taking out only the essential pattern sequences needed for training the assistance model, the system reduces data privacy risks while maintaining training data availability.
Data Source
Figure 1(a)~1(g)
AI summary
The invention relates to a Computer-implemented method for labelling of user action (UAC) data in a user environment (ENV), in particular wherein said user environment (ENV) is a computer (CMP) implemented user interactive (UIA) environment (ENV), characterized in comprising the steps of: a) providing a pre-trained user action (UAC) pattern recognition model (PRM), wherein the pattern recognition model (PRM) is designed in such a way that the pat-tern recognition model (PRM), from a part-sequence of user actions (UAC) including a user action (UAC) pattern (APT) known to the pattern recognition model (PRM), recognizes the user action (UAC) pattern (APT) known to the pat-tern recognition model (PRM), and outputs the user action (UAC) pattern (APT) or a label associated with the user action (UAC) pattern (APT), b) logging of user actions (UAC), c) analyzing the logged user actions (UAC) by said user action (UAC) pattern recognition model (PRM) for user action (UAC) patterns (APT) known to the pattern recognition model (PRM), d) generating of a user interaction (UIA) in case of not detecting a known user action (UAC) pattern (APT) in a sequence (SQC) of user actions (UAC) as the user interaction (UIA) requesting information about the past sequence (SQC) of user actions (UAC), e) labelling the past sequence (SQC) of user actions (UAC) using the information received from the information request, f) storing the labelled sequence (SQC) of user actions (UAC) in a user action (UAC) sequence (SQC) database (DBS), g) training the pattern recognition model (PRM) using the user action (UAC) sequence (SQC) and corresponding labels of the user action (UAC) sequence (SQC) database (DBS).