Text as Data in Demography: Russian-language experience

Abstract

We propose to consider our experience in data use of Russian-language texts of social networks, electronic media, and search engines in demographic analysis. Experiments on the automatic classification of opinions have been carried out. Conversational RuBERT has been used in most cases. The following main scientific results on text data will be described: (1) short-term forecasts of fertility dynamics according to Google trend data, (2) automatic measurement of the demographic temperature of various demographic groups (pronatalists and antinatalists) in social networks, (3) sentiment analysis of reproductive behavior, sentiment analysis of vital behavior in pandemic, sentiment analysis of attitudes toward demographic and epidemiological policy according to social network data, (4) analysis of the arguments of social network users, and (5) analysis of media publications on demographic policy. A description of the created open databases of all these studies will be provided. All of the studies described will contain reflections on the advantages and difficulties of using texts as data in demographic analysis.