CONTEXT-AWARE AUTOMATIC TERM EXTRACTION USING USER BEHAVIOR AND TEXTUAL FEATURES

Abstract

Automatic Term Extraction (ATE) is a pivotal task in Natural Language Processing (NLP) that seeks to identify semantically significant words and multi-word expressions from domain-specific corpora. Classical ATE methods, grounded in statistical heuristics such as Term Frequency–Inverse Document Frequency (TF-IDF) or purely linguistic pipelines, often fail to reflect the genuine importance of terms in interactive, user-driven environments. This paper proposes a context-aware ATE framework that synergistically integrates textual features—including term frequency, positional salience, syntactic patterns, and co-occurrence graphs—with implicit user behavior signals such as query history, dwell time, click-through rate, and session-level engagement metrics. A hybrid scoring function merges these heterogeneous signals into a unified relevance score for each candidate term. The system is evaluated on a benchmark corpus drawn from the ACM Digital Library and a proprietary e-learning dataset. Experimental results demonstrate that the proposed approach achieves a Precision of 0.87, Recall of 0.83, and F1-Score of 0.85, outperforming baseline TF-IDF and RAKE-based systems by 9–14% in F1. The framework is lightweight, language-agnostic at the feature level, and readily deployable in real-time digital library, search-engine, and recommendation-system pipelines.