Scholarlink

<p>Polysemous words are those that can provide multiple meanings depending on the diverse circumstances in which they are employed in a natural language text. <i>Polysemy resolution</i> (PR) is a technique for removing any ambiguity generated using a polysemous term in a text. It is important to find out the proper sense of a polysemous term to identify the correct emotion such as positive, negative, or neutral of a Bengali text. In this research, identification of proper textual information can be accomplished by following emotion tagging after comprehension of PR. PR key, term recurrence, <i>emotion detection</i> (ED) factor, and ED-PR score were taken into consideration when extracting a feature vector by the proposed methodology. Following the creation of the feature set, the training data is processed and fed into models under supervised learning approaches. Then, the model performances are assessed to calculate the accuracy of the ten polysemous words based on different window sizes. After analysing the model performances, we found that the iterative dichotomiser 3 or ID3 classifier produced the best result among the tested classifiers. Based on this, the ID3 classifier was modified to fit with Bengali text by introducing reciprocity degree and confidence degree. In this regard, a new tree-based classifier called BEN-ED-PR is proposed to categorise the Bengali text. This model has completed the task with the best accuracy when compared to the tested traditional classification methods for Bengali text. The effectiveness of this strategy has been assessed using a dataset of ten polysemous terms made from current literary works in Bengali. Additionally, a comparative study has been made for measuring the contribution of the respective <i>parts of speech</i> (POS) (i.e., adjective, noun, verb, and preposition) corresponding to each sense of the polysemous words. Thus, it can be said that the use of our proposed BEN-ED-PR model to evaluate its performance on the dataset can offer a new direction on the assessment of PR techniques with emotion detection in Bengali language, throwing some light in this field to encourage the research domain by this work.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

BEN-ED-PR: Emotion detection from Bengali text with polysemy resolution

  • 期刊论文

  • 发表日期：

    2025年7月23日

查看全文
  • Taniya Seal,
  • Debapratim Das Dawn,
  • Shreya Basak,
  • Sreeparna Roy,
  • Sumit Kumar Gond,
  • Dwaipayan Ghosh,
  • Abhinandan Khan,
  • Sanjit Kumar Setua,
  • Rajat Kumar Pal

摘要

Polysemous words are those that can provide multiple meanings depending on the diverse circumstances in which they are employed in a natural language text. Polysemy resolution (PR) is a technique for removing any ambiguity generated using a polysemous term in a text. It is important to find out the proper sense of a polysemous term to identify the correct emotion such as positive, negative, or neutral of a Bengali text. In this research, identification of proper textual information can be accomplished by following emotion tagging after comprehension of PR. PR key, term recurrence, emotion detection (ED) factor, and ED-PR score were taken into consideration when extracting a feature vector by the proposed methodology. Following the creation of the feature set, the training data is processed and fed into models under supervised learning approaches. Then, the model performances are assessed to calculate the accuracy of the ten polysemous words based on different window sizes. After analysing the model performances, we found that the iterative dichotomiser 3 or ID3 classifier produced the best result among the tested classifiers. Based on this, the ID3 classifier was modified to fit with Bengali text by introducing reciprocity degree and confidence degree. In this regard, a new tree-based classifier called BEN-ED-PR is proposed to categorise the Bengali text. This model has completed the task with the best accuracy when compared to the tested traditional classification methods for Bengali text. The effectiveness of this strategy has been assessed using a dataset of ten polysemous terms made from current literary works in Bengali. Additionally, a comparative study has been made for measuring the contribution of the respective parts of speech (POS) (i.e., adjective, noun, verb, and preposition) corresponding to each sense of the polysemous words. Thus, it can be said that the use of our proposed BEN-ED-PR model to evaluate its performance on the dataset can offer a new direction on the assessment of PR techniques with emotion detection in Bengali language, throwing some light in this field to encourage the research domain by this work.