Abstract: Opinions of the public and the sentiments originating thereby play a pivotal role in social procedures. Sentiment analysis deals with the resolution of the tone or polarity of the text- how positive or negative it is. When applied to news reports, it provides a wide range of applications.
This study analyses news reports in real-time from reliable sources using a slightly modified Na¨ıve Bayes’ Algorithm. An article is fetched and then pre- processed to get rid of noisy words like English articles. After tokenization, the probability of each word being either positive or negative is determined. This is achieved by training a model using a dataset of brief news headlines, with their sentiment values labelled. The overall probability is summed using the well-known Bayes’ theorem, which gives the name to the algorithm.A slight modification is proposed to this algorithm by calculating sentiment value for the field ‘engineering,’ which separates or calculates how a particular report is related to engineering. Based on the relevance to engineering (defined herewith using the dataset), a system is developed that prompts the head of an organization or any competent authorities with the report through an email
Keywords: text analysis, natural language processing (NLP), machine learning, text polarity, opinion mining, Na¨ıve Bayes’.
|
DOI:
10.17148/IJARCCE.2022.11322
