Semantic Measure Based on Features in Lexical Knowledge Sources

Ummi Zakiah Zainodin, Nazlia Omar, Abdulgabbar Saif

Abstract


Semantic measures between concepts require some of cognitive capabilities such as categorization and reasoning to estimate semantic association among concepts. For this reason, this problem has numerous applications in artificial intelligence, natural language processing, information retrieval, text clustering, text categorization and many other related fields. Measuring lexical semantic relatedness generally requires certain background information about the concept or terms. Semantic measures between concepts are divided into two main sources: knowledge based and unstructured corpora. Both resources play important role in the task of measuring lexical semantic relatedness. Knowledge-based semantic measures have been proposed to estimate semantic similarity between two concepts using several approaches such as ontology-based, graph-based and concept's vector approaches. This paper reviews existing semantic similarity measures which depend on the lexical source and discusses the various approaches on semantic measures which include the path-based, information content, gloss-based and feature-based measures. This paper also focuses on semantic measures that are based on features using lexical knowledge sources and discusses some issues that arise in these measures.

Pengukuran semantik antara konsep memerlukan sebahagian daripada keupayaan kognitif, seperti pengkategorian, ingatan, membuat keputusan dan penaakulan serta penggunaan dan penemuan analogi di kalangan yang lain. Masalah tersebut mempunyai aplikasi di dalam kecerdasan buatan, pemprosesan bahasa tabii, capaian maklumat, pengelompokan teks, pengkategorian teks dan pelbagai bidang yang berkaitan. Pengukuran hubungan semantik leksikal secara amnya memerlukan beberapa latar belakang maklumat mengenai konsep atau terma. Pengukuran semantik antara konsep dibahagikan kepada dua asas utama; berasaskan pengetahuan dan berasaskan korpus tidak tersusun. Kedua sumber ini memainkan peranan yang penting dalam tugasan pengukuran hubungan semantik leksikal. Pengukuran semantik berdasarkan sumber pengetahuan telah diperkenalkan bagi menganggarkan persamaan semantik antara dua konsep menggunakan beberapa pendekatan seperti berdasarkan ontologi, berdasarkan graf dan pendekatan konsep vektor. Kertas ini mengulas pengukuran persamaan semantik sedia ada yang bergantung kepada sumber leksikal dan membincangkan pelbagai pendekatan terhadap pengukuran semantik termasuk pengukuran berasaskan laluan, kandungan maklumat, berasaskan glos dan berasaskan ciri. Kertas ini juga memberi fokus terhadap pengukuran semantik berasaskan ciri dengan menggunakan sumber pengetahuan leksikal dan membincangkan beberapa isu yang timbul berkaitan dalam pengukuran tersebut.



Keywords


Semantic Measure; Knowledge Based; Semantic Similarity; Features-Based

Full Text:

PDF

Refbacks

  • There are currently no refbacks.


e-ISSN : 2289-2192

For any inquiry regarding our journal please contact our editorial board by email apjitm@ukm.edu.my