Knowledge-based semantic relatedness measure using semantic features

• Main Authors: Ali Muttaleb, Hasan, Noorhuzaimi@Karimah, Mohd Noor, Rassem, Taha H., Ahmed Muttaleb, Hasan
• Format: Article
• Language: English; English
• Published: The World Academy of Research in Science and Engineering 2020
• Subjects: QA75 Electronic computers. Computer science; QA76 Computer software
• Online Access: http://umpir.ump.edu.my/id/eprint/28458/1/IJATCSE-0292--04-2020Ali.M.pdf; http://umpir.ump.edu.my/id/eprint/28458/7/Knowledge-based%20semantic%20relatedness%20measure%20using%20semantic%20features.pdf; http://umpir.ump.edu.my/id/eprint/28458/; https://doi.org/10.30534/ijatcse/2020/02922020; https://doi.org/10.30534/ijatcse/2020/02922020
• Institution: Universiti Malaysia Pahang

Measuring semantic relatedness has received much attention for uses in many fields such as information retrieval and natural language processing. For handling synonymous problem in distributional-based measures, many researchers are investigating how to exploit semantic features in lexical sources to form knowledge-based measures. In the knowledge-based measures, a hierarchy model is used to measure the relatedness between words based on only the taxonomical features extracted from a provided lexical source. In this paper, a new knowledge feature-based measure is proposed to build the semantic vector of a word construct on taxonomical and non-taxonomical feature of relation words. The proposed measure utilised the topological parameters that weight the importance of each element in the semantic vector. One of the gold dataset used to assess the proposed model and compare the findings with other related works. The results demonstrated the effectiveness of the proposed model on measuring semantic relatedness between words. In this paper, the research framework is identified based on the observations made on the previous related works that have been conducted for semantic representation and semantic relatedness measures. The required data in this research includes the semantic knowledge-based approach and the evaluation datasets. The semantic knowledge that will be used throughout of this research is extracted from English WordNet 3.1. On the other hand, the evaluation datasets covers the gold standard benchmarks which have been used for evaluating the semantic relatedness measurements and text mining tasks. Finally, the evaluation is preform to evaluate the proposed method (PM) based on approach in this research, in which obtained the result have been analyzed, to discuss and compare based on different performance measure and finding the strength and weakness in this paper, to alternative the semantic representation correlated to this research, to designing and develop the topical-based on the semantic representation method for text mining from Social media.