Paper short abstract:

In our paper we introduce the results of application of topic modeling to a regional selection of Estonian fairy tales and folksongs, and compare the results with the folkloristic typologies.

Paper long abstract:

The large volumes of folklore material collected in the tradition archives have required classification and systematisation for the materials to be findable and usable, and also for (and as a part of) the research of folklore. The classification typically takes in account elements of text and context - main characters, content, functions, ways of performance etc.

During the long course of the history of folkloristics voluminous collections of archival materials have been classified and systematised by reading the texts. The typological systems have been of help to get an overview of the material, as well as in classifying the new items. In the current digital age, and with the existence of large text corpora it seems more than natural that we would use the computer power for the classification task.

Nowadays, the classification and analysis of various everyday texts (news, social media feeds, etc) has become an everyday task in discovering the trends, tendencies, attitudes in society necessary for political as well as for business analysis. One of the methods used for computational classification is topic modeling - a method based on the language statistics that discovers the collections of the words that tend to regularly co-occur in various texts. This collection of words, named topic, can be easily comparable with folkloristic term of motif.

In our paper we introduce the results of application of topic modeling to a regional selection of Estonian fairy tales and folksongs, and compare the results with the folkloristic typologies.