DISTRIBUTIONAL SEMANTICS AND TAGSET OF NOUNS TAXONOMIES: A CORPUS-DRIVEN STUDY OF SARAIKI FOLK SONGS

Abstract

This study investigates the distribution of semantic networks of nouns in Saraiki folk songs through
a corpus-driven approach, utilizing 3A Model for annotation and analysis. Saraiki, spoken by
nearly 20 million people in Pakistan, has limited digital resources, with the iJunoon Saraiki
dictionary (2017) being one of the few available corpora. To address the need for a structured
lexical resource, this study aims to uncover the underlying semantic relationships among Saraiki
nouns, which could contribute to the development of a Saraiki WordNet. A specialized corpus of
0.25 million words was compiled from the folk songs, and nouns were tagged, categorized, and
manually annotated using UAM Corpus Tool. Unlike traditional approaches, a discourse analysis
framework was employed to examine the contextual use of nouns and identify recurring patterns
in semantic relations. Specifically, relational semantics was used to analyze connections such as
hyponymy, meronymy, and part-whole relations. The findings revealed that meronymic
relationships were the most frequent, while antonymy was the least common occurrence. This study
highlights the importance of corpus-driven methods for understanding the semantic structures of
under-resourced languages and suggests that insights gained from the Saraiki noun relations
could be central to creating digital tools for Natural Language Processing (NLP) and
computational linguistics applications.