Fiction Book Recomendation System Based on Topic Modeling

This paper presents an overview and analysis of the areas of application of the recommended systems in the field of literature, as well as in the Russian-language systems of the requirements of books, identifies their advantages and disadvantages. In particular, they do not give full recommendations based on topics. The analysis made it possible to put forward requirements for the system being developed. The concepts of thematic and probabilistic-thematic models were reviewed, a review and analysis of approaches to the development of thematic models was carried out. The analysis showed that the Dirichlet latent distribution most fully meets the requirements of the development. Ways to work with thematic models in Python were considered. Test and performance data were assembled for training and the thematic model. A thematic model was conducted, conducted by several training and subsequent tests at each iteration. The architecture of the created thematic model was presented. At the second stage of practical work a web service was developed.