Elias Lönnrot's collection of Finnic oral poetry in 1848 when composing the Kalevala epic.

A pipeline for creating a dataset containing the poems Elias Lönnrot, the composer of the Finnic Kalevala epic, presumably had access to in 1848 when creating Kalevala (1849). The pipeline is created as a fork of Maciej Janicki's work on the Filter-project, which enables calculation of poem-similarities and clusters in Finnic oral poetry. The resulting tool, also based on Janicki's work, is accessible online for searching and viewing poems, poem clusters, and poems with similar verses side by side. The corpus currently combines the following collections: Suomen Kansan Vanhat Runot, only including the poems collected before 1849. (Old Poems of the Finnish People; SKVR) The Finnish Literature Society's Corpus of unpublished poems (JR) only including the poems collected before 1849. A Corpus of literary poems (KR) based on public domain sources. Only those published before 1849 with these exceptions: Elias Lönnrot : Kalevala (1849) Elias Lönnrot : Suomen Kansan arwoituksia ynnä 189 Wiron arwoituksen kanssa (1851) Suomen Kansan Muinaisia Loitsurunoja toimittanut Elias Lönnrot (1880) Elias Lönnrot : Lisiä Vanhaan Kalevalaan (Niemi, A. R., 1895) The aim of this project is to harness the tools created by Maciej Janicki as part of the FILTER-project to analyse a subset of the FILTER-project corpus, consisting of poems recorded before the publication of Kalevala in 1849. This tool enables comparision between poems with high similarity side-by-side, in order to reveal how Elias Lönnrot used, altered and merged variants of poems when compiling the Kalevala epic. The number of poems in this dataset is 21919, which equals to roughly 10% of the whole FILTER-corpus. The need for a computational tool to critically assess E. Lönnrots work arouse when Venla Sykäri, researcher at The Finnish Literature Society (SKS), began her work with conducting contemporary research on how the Kalevala epic came to exist. While SKS provides an online tool for viewing and sorting the collected poetry that became Kalevala, there was no tool view similarities and differences among the poetry with this particular data. As it happens, the Runoregi-tool created in the Filter-project does just that, and could with "minor" modifiactions use and present this custom dataset. For details on using the modified pipeline, database-creation, and the web-interface, see https://github.com/jakobytes/elias-1848. A demonstration of the interface and end result can be viewed here: Elias-1848. Preliminary results suggest that the default similarity and clustering thresholds might be too tight, as the dataset is now less heterogenic. Lowering the threshold better reveals what we wish to see: Poems sharing multiple verses with Kalevala, but different variants. It can also be seen how some characters present in Kalevala are composites of many characters present in the source poems. We do know that from earlier research done by close-reading and expertise. This tool seems to make it easier to see what previously was only accessible to those with extensive knowledge of the poems. The jury's still out ont this though, and adjustments will surely be made once there's more feedback.