Text classification using TensorFlow.js

Prokhorov Andrey

The topic of this paper is the use of machine learning on the example of the TensorFlow software package in its implementation for the JavaScript programming language for the task of natural language recognition. The relevance of the research and subsequent development of the software system is due to the rapid development of artificial intelligence and the increasingly frequent need for classification and categorization of large volumes of data (Big Data) in information systems. Objective: to develop an optimal natural language classification system using artificial intelligence tools, such as machine learning. Methods: when working on the study, the analysis method was used to study the principles of neural networks, as well as the features of TensorFlow implementation. When testing the system, methods of comparison and empirical measurement of performance were used. Results: the result of the research is a software system that classifies data sets of the subject area containing words and sentences of the natural language. Also, system performance tests were conducted on central processors and video accelerators to identify ways to increase the speed of operation and improve scalability. Conclusions: in the course of the study, the most suitable type of neural network was determined, conclusions were obtained about the optimal software settings of the model for the tested data, as well as optimal hardware configurations. The study of this material will allow beginners in the field of machine learning to quickly move on to developing models and solving real-world applied problems.

Keywords:TensorFlow; JavaScript; machine learning; ML; neural network; web applications; search; NVIDIA; GPU; CPU.