ABSTRACT
In this study, we propose a model for resolving anaphora occurrences in Igbo text using a knowledge-poor approach. The approach does not rely on sophisticated linguistic knowledge, and its core method includes a list of multilingual antecedent indicators (e.g., subject place, recency, frequency) and constraint rules (e.g., gender, person, and number agreement). The model primarily focuses on pronominal anaphora types, specifically third-person pronouns, and deals with both intrasentential and intersentential anaphors. These personal pronouns can be hidden within verbs or independent pronouns. The proposed model involves several subtasks, including text preprocessing, POS tagging, locating independent anaphors in sentences, extracting hidden personal pronouns, identifying potential antecedent candidates within a defined range of preceding sentences, applying eliminative rules (constraint rules), optional rules (preferential rules), and selecting the candidate with the highest aggregate score. The experiment’s datasets were collected from Igbo Holy Bible and fiction texts. The prototype’s evaluation was performed on 330 sentences, and the success rate for different anaphora scenarios was recorded. The study encountered challenges in extracting hidden anaphors from the verb word class due to word ambiguity, requiring consideration of the sentence’s meaning. As a future direction, further research focusing on the pragmatic knowledge of Igbo anaphora resolution is recommended.
Design and Implementation of Anaphora Resolution for Igbo’s Personal Pronoun, GET MORE COMPUTER SCIENCE PROJECT TOPICS AND MATERIALS
