Open AccessAn Improved Sentiment Extraction Model for Social Media Contents using spaCy Based Deep Neural Networks
Author(s) -
S. Thivaharan
Publication year - 2021
Publication title -
international journal for research in applied science and engineering technology
Language(s) - English
Resource type - Journals
ISSN - 2321-9653
DOI - 10.22214/ijraset.2021.36366
Subject(s) - computer science , categorical variable , social media , asynchronous communication , artificial intelligence , data mining , data extraction , machine learning , world wide web , computer network , medline , political science , law
Modern communication devices generate huge amount of data through the manifold usage of various social media applications. Among the entire generated data more than 40% are unstructured in nature. The industry also is reluctant to retain the data with the following characteristics: data containing asynchronous time stamp, replicated data, data which are broken while transmission and data that leads to misclassification. It is high time to drop-out the irrelevant data and considering the synchronous ones. In this article, a sentiment extraction model is proposed that governs the various social media contents. SpaCy is used as the preferred implementation language as it has many readily available libraries for the purpose of content classification. To avoid the over-fitting problem the actuators like “relu” and “sigmoid” are used. Even though many such classifiers are available for content classification, this article with the appropriate setting of Epoch counts, a categorical accuracy of 68% is obtained. The entire model is implemented in the TensorFlow based platform.