Analysis and Prediction of User Editing Patterns in Ontology Development Projects

Area of Science:

• Knowledge Representation and Reasoning
• Human-Computer Interaction
• Data Mining

Background:

• Ontology development is a complex, collaborative process requiring specialized tools.
• Streamlining ontology authoring enhances efficiency and information discovery.
• Predicting user actions can optimize ontology editing interfaces.

Purpose of the Study:

• To investigate the predictability of the next user editing operation in ontology development using data mining.
• To evaluate the effectiveness of association rule mining for predicting user actions based on historical data.
• To assess the impact of various parameters on prediction accuracy.

Main Methods:

• Utilized association rule mining on ontology change logs.
• Employed a sliding window model for continuous time-based prediction.
• Generated patterns from training data and tested on adjacent testing data.
• Evaluated prediction accuracy across different window sizes, user groups, and ontologies.

Main Results:

• Demonstrated the capability to predict the next user editing operation with significant accuracy.
• Confirmed the effectiveness of association rule mining in identifying user editing patterns.
• Showcased that prediction accuracy is influenced by training/testing window sizes, user groups, and specific ontologies.

Conclusions:

• User editing behaviors in ontology development are predictable.
• Discovered editing patterns can inform the creation of recommendation modules for ontology tools.
• Findings support the design of user interfaces tailored to user editing behaviors for enhanced usability.