Mapping Snorra-Edda

First published , rewritten and updated June 22, 2020

Trying to understand the narrative grammar of the mythèmes used in the Snorra-Edda, I analysed the 80 most frequent tokens, dismissing of course stopwords, with our mythème-laboratory. I used Rasmus Anderson’s English translation (1879), that does include all major mythological passages. I defined discs of 80 words around each search token with a sample coverage rate of less than 50%. Only those words are retained where the number of words in the work is greater than 20 and the over-representation coefficient in the sample is greater than 1.2. The radius of the discs may be automatically reduced to comply with the limit specified for the recovery rate. See the details here. I imported the data to gephi, interpreting the search token as source node and the tokens significant defining tokens as target. Weight of edges corresponds to the CSRR, that is the normalized importance of the coefficient of over-representation. In order to gain a readable layout, I ran the Force-Atlas-layout-algorithm.

A first map places the used mythèmes quite nicely in the context of other mythemes used in the same context: As in the text, Har and Gangleri are highly connected, as they are dialoging throughout the text as are the place of the hall as the location of telling stories. Loke and Thor are undertaking journeys together, accompanied of Thor’s hammer and liked both to the Aesir and the giants. God, heaven, earth and the world are put together and with these mythemes questions of cosmology. Odin is linked too questions of kingdom and fatherhood. Spots in dark green have a high degree, that is are highly connected to others, spots in light green less.

In the above map, giants seem to take an important place. To verify the centrality of the different mythèmes, I run the algorithms that define centrality of the different nodes assigned colors and font size according to the centrality of the elements in the graph. The degrees of the nodes were weighted. The next three graphs show three different resulting variants. Attention : the orientation of the map has changed as I rerun several algorithms, but the general structure remained unchanged. Dark blue nodes are central.

Map with Closeness Centrality as central theme
Map with Betweeness Centrality as central theme
Map with Eigenvector Centrality as central theme

Data at Mythème-laboratory : https://mythemes.u-strasbg.fr/w/index.php/Ref:1879_SNORRE%E2%80%99S_EDDA#tab=L_27_C5_93uvre.

Thomas Mohnike