Pattern Prediction: A Comprehensive Review of Evaluation Methods

Abstract

Keyphrases are expressions allow to identify the topics, or the ideas addressed in a document. Various natural language processing (NLP) applications have employed these phrases to improve their performance. Several research papers have been published in recent years that focus on keyphrases extraction or generation approaches. Various techniques are used to evaluate the performance of these methods. These techniques can be classified into three main categories: lexical similarity, rank-based, and semantic similarity techniques. This paper presents a comprehensive review of these techniques, with a particular focus on the major challenges encountered during the evaluation process. Also, this review underscores the need for more robust and rigorous evaluation techniques to improve the reliability and effectiveness of keyphrases prediction approaches. This review will contribute to understanding the different keyphrases evaluation techniques, the main challenges they face, and future research directions related to keyphrases prediction evaluation.