This article is devoted to the development and approbation of a new approach to the collection, processing and analysis of open data in the Russian language for identification of key technological trends. To solve the problem of formation and subsequent analysis of structured datasets methods of web scraping, natural language processing and analysis of time-series have been developed and implemented via programming. The approach described in the article has been applied for the first time in order to extract and structure information from scientific articles, news resources and patent documentation in the Russian language for the first time. As a result of analyzing the obtained dataset of scientific publications, 30 most frequently mentioned bigrams and the same number of trigrams of technological terms have been identified. Based on the frequency analysis of bigrams and trigrams, key technological terms were identified which then were used for complex filtration on key technologies. Complex filtration enabled to fulfill the search of patents in Russian and their collection for further analysis. As a result of preprocessing of the obtained patent data time series of patent activity have been formed. The programme system of key technological identification has been implemented in JavaScript and Python using Selenium and BeautifulSoup libraries for web scraping, NLTK and Scikit-learn for text data processing and analysis. The study focused on the dynamics of the development of key technologies over time has allowed to identify periods of intensive patent activity and declining interest in this or that kind of technology. The results presented in the article provide a basis for further development of machine learning methods for the purpose of predicting technological development and identifying promising areas of applied research.
