Bangla-REX: A distinct dataset for Bangla relation extraction.

Area of Science:

• Natural Language Processing (NLP)
• Computational Linguistics
• Information Extraction

Background:

• Lack of specialized datasets for relation extraction in the Bangla language.
• Need for structured resources to support advanced NLP tasks in Bangla.
• Growing importance of multilingual NLP and information processing.

Purpose of the Study:

• Introduce Bangla REX, a comprehensive dataset for Bangla relation extraction.
• Provide a structured corpus and knowledge base to facilitate NLP research.
• Enable accurate extraction and classification of semantic relations in Bangla text.

Main Methods:

• Compilation of 90,441 text entries from reliable sources like Bangla Wikipedia and Wikidata.
• Development of a Bangla Knowledge Base (KB) with 63,256 entries for automated annotation.
• Annotation of the dataset with various relation tags, including movie actors, locations, and company information.

Main Results:

• Creation of a large-scale, structured dataset (Bangla REX) for relation extraction tasks.
• Development of an accompanying Bangla Knowledge Base to aid in corpus annotation.
• Inclusion of diverse relation categories crucial for semantic understanding.

Conclusions:

• Bangla REX addresses a critical resource gap for Bangla NLP.
• The dataset provides a valuable benchmark for training and evaluating relation extraction models.
• This resource is expected to foster innovation in Bangla NLP and multilingual information processing.