Predicting Supervised Machine Learning Performances for Sentiment Analysis Using Contextual Based Approaches

EasyChair Preprint no. 5162

Abstract

The fundamental thought of our methodology is to inspire client inclinations communicated in text-based audits, an issue known as opinion investigation, and guide such inclinations onto some evaluating scales that can be perceived by existing CF calculations. One significant errand in our rating deduction system is the assurance of wistful directions PSWAM and qualities of assessment words. It is because surmising a rating from an audit is chiefly done by separating assessment words in the survey, and afterward accumulating the PSWAM of such words to decide the predominant or normal notion inferred by the client. We played out some primer examination on film audits to research how PSWAM and qualities of assessment words can be resolved and proposed a relative- a recurrence-based technique for performing such assignments. The proposed technique tends to a significant impediment of existing strategies by permitting comparative words to have distinctive PSWAM. We additionally created and assessed a model of the proposed structure. Fundamental outcomes approved the viability of different assignments in the proposed system and suggested that the process doesn’t rely on a large preparation corpus for working. An accelerated algorithm based on the Naïve Bayes approach is used to solve the PSWAM and a parallel algorithm based on FISTA is incorporated to further improve the efficiency. The result is a graph representing opinion target and opinion word candidates before and after extraction further helping users simplify the task of analysis.

Keyphrases: Fast Iterative Shrinkage-Thresholding Algorithm, Naive Bayes, Opinion Mining, Partially-Supervised Word Alignment Model

@Booklet{EasyChair:5162,
  author = {Venkata S Lakshmi and K Janan and Joshua P S Joseph and Mohammed Sharoz},
  title = {Predicting Supervised Machine Learning Performances for Sentiment Analysis Using Contextual Based Approaches},
  howpublished = {EasyChair Preprint no. 5162},

  year = {EasyChair, 2021}}