ENHANCING NAMED ENTITY RECOGNITION ON HINER DATASET USING ADVANCED NLP TECHNIQUES

Abstract

Named entity recognition (NER) is a natural language processing (NLP) categorization labelling job where the objective is to assign words to a predefined set of named entity classes. Although several NER models have been proposed thus far, experts have not yet discovered a workable solution for an inflectional language with limited resources, such as Hindi. . Usually, several manual techniques have been used for entity prediction, such as ambiguity, false positives, and reliance on lengthy annotated data. Nevertheless, these manual methods seem to be ineffectual and time-consuming for acquiring optimum outcomes and are unable to predict all the entity types in the Hindi language. To solve this issue, some researchers have concentrated on NER models. Conversely, it lacks speed and accuracy. Therefore, the present research uses advanced NLP models such as bidirectional encoder representations from transformers (BERT), Distil BERT and the robustly optimized BERT approach (RoBERTA) for effective entity prediction performance. It predicts entities such as person, location, organization and event in the Hindi language. The proposed research uses the Hindi named entity recognition (HiNER) dataset for the purpose of evaluation. The effectiveness of the present model is assessed via several evaluation metrics, such as the F1 score, recall, precision and accuracy, to assess the performance. Furthermore, the comparison of the proposed model reveals the effectiveness of the present research. In conclusion, the projected research envisioned contributing to the emerging NER models, thereby offering an effective entity prediction model for the needed users.