Chatbot based on neural networks and word embedding to increase customer loyalty

In the digital era mobile devices are becoming the main instrument of human social interaction. With the growing popularity of instant messengers, the role of chatbots in the mobile environment appears to be more and more significant. Intelligent interactive chatbots are often used in mobile applications and help improve the interaction between companies and their customers, which ultimately increases customer loyalty to that organization. Chatbots allow companies to communicate with customers on an individual basis, without involving employees and thereby saving time, money, and human resources. The majority of chatbots works with scripted algorithms and they are not universal. This is due to the simplicity and speed of development. However, in this case, there is a risk of missing many choices in the decision tree. Chatbots based on neural networks can solve this problem, but it should be taken into consideration that both of them have a drawback – long processing of messages and feedback. In the context of the scenario approach, this is caused by long branch transitions. For neural networks, complexity arises because of the feedback processing algorithm. In that instance, the application of the service will not be justified, customer loyalty to the organization will deteriorate. In this connection, the article discusses an alternative approach to creating chatbots with the aid of neural network technologies and text representation methods, which avoids the problems described above. As a means of chatbot design, the following technologies were utilized: Python 3.6, genism libraries, sklearn, scipy, pandas, word2vec and doc2vec technology. The article also describes a way to accelerate chatbot feedback and training using KD-Trees.

1. Chakrabarti C., Luger G.F. Artificial conversations for customer service chatter bots: architecture, algorithms, and evaluation metrics. Expert Syst. Appl. 2015;2(20):6878–6897.

2. Xu A., Liu Z., Guo Y., Sinha V., Akkiraju R. A new chatbot for customer service on social media. Proceedings of CHI. 2017:3506–3510.

3. Robert Dale. The Return of the Chatbots. Natural Language Engineering. 2016;22(5):811–817. DOI:10.1017/S1351324916000243.

4. Bliznuk B.O., Vasilieva L.V., Strelnikov I.D., Tkachuk D. S. Modern methods of natural language processing. Bulletin of the Kharkov National University named by V.N. Karazin. 2017;36:14–26.

5. Tomas Mikolov, Kai Chen, Greg Corrado, Jeffrey Dean. Efficient estimation of word representations in vector space. 2013; CoRR,abs/1301.3781

6. Ivanova A.V., Filippova L.B. Word processing and machine learning techniques for creating chat bots. In the collection: New horizons. VIII scientific and practical conference with international participation. Collection of materials and reports. 2021; 289-292. (In Russ.)

7. Mikolov T., Sutskever I., Chen K., Corrado G.S. and Dean J. Distributed representations of words and phrases and their compositionality. In Advances in Neural Information Processing Systems 26: 27th Annual Conference on Neural Information Processing Systems 2013. 2013;3111–3119.

8. Dolgoshey A.V. Application of neural networks for text processing. New mathematical methods and computer technologies in design, production and scientific research. Materials of the XXIII Republican Scientific Conference of Students and Postgraduates. 2020;42–43. (In Russ.)

9. Fengquan Zhang, Yahui Gao, Liuqing Xu. An adaptive image feature matching method using mixed Vocabulary-KD tree Multimedia. Tools and Applications volume 2020. 2020;79:16421–16439.

10. Lebedinskaya N.A., Lebedinsky D.M. Generalized KD-trees and local transformations. Bulletin of St. Petersburg University. Maths. Mechanics. Astronomy. 2012;3:42–43. (In Russ.)