Title: The art of context classification and recognition of text conversation using CNN

Authors: Sandeep Rathor; Sanket Agrawal

Addresses: Department of Computer Engineering and Applications, GLA University, Mathura-281406, India ' Infosys Ltd, Plot No. 1, Hinjewadi Rajiv Gandhi Infotech Park, Hinjawadi, Pune-411057, India

Abstract: This paper proposes a robust model for recognising the context of a conversation by using a convolutional neural network (CNN). Initially, the pre-processed text is passed to an embedding layer, which gives out a feature matrix. This is passed to a multi-level CNN. In the proposed model, each CNN reduces the input matrix to half of the input size. Thus, the output of the next CNN layer and the pass of the current CNN layer followed by average pooling are added. The output from the last CNN layer and global outputs are concatenated and finally, passed into two fully connected layers FC(512) and FC(8). In the proposed paper, context is classified into categories like spiritual-mythological, healthcare, trade-commerce, politics, personal, academic, sports, and violence. Experimental results show that our proposed model outperforms with an accuracy of 92.97%. We compared the performance of our model to conventional ML algorithms and obtained promising results.

Keywords: context recognition; text conversation; text mining; convolutional neural network; CNN; machine learning.

DOI: 10.1504/IJIDS.2023.131533

International Journal of Information and Decision Sciences, 2023 Vol.15 No.2, pp.185 - 200

Received: 30 May 2020
Accepted: 29 Jan 2021
Published online: 19 Jun 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article