HUM data lab > Workflows > Analysis of Tale Type 0567A in the Corpus of Seto Folk Tales

Analysis of Tale Type 0567A in the Corpus of Seto Folk Tales

This workflow describes the aim of analysing tale type 567(A), the tale of the magic bird, in a dataset of Seto folk tales, in order to understand how and to what extent the variants of this type differ from one another. Gretta Nikolajeva was responsible for the qualitative analysis, while Aleksander Amos Nigesen was responsible for data processing in R.

Workflow steps

Keywords: Discovering

The aim of the first stage was to become familiar with the dataset. The folk tale database of the Estonian Folklore Archives was established in 1999, initially for collecting Estonian tales of magic and preparing an academic publication series. Later, it was expanded to include other subgenres of folk tales as well. The Estonian folk tale database contains around six thousend tales, of which 2,111 Seto fairy tales were available in the dataset. Familiarisation with the dataset made it possible to formulate research questions and decide which metadata or textual features should be analysed in the following stages.

Keywords: Contextualizing

The aim of this stage was to examine Seto culture and its historical background in order to understand the context in which Seto folk tales developed and the features characteristic of them. Particular attention was paid to the fact that Seto culture remained less influenced by written culture for a long time, which is why the folk tales have preserved more authenticity characteristic of oral tradition.

This stage laid the foundation for the subsequent steps of the workflow by helping to understand the specific features of the dataset under analysis and supporting the formulation of the research questions.

Keywords: Conceptualization

The aim of this stage was to define the research questions that would guide the subsequent analysis and help focus on specific themes in Seto folk tales. The analysis focused on tales of the magic bird due to personal interest and in order to examine the specific features of one particular tale type in greater detail. More specifically, the analysis focused on type 567A and its narrative and thematic features.

Research questions:

  1. In tales of the magic bird, does the animal associated with wisdom always appear as a bird, or are there variations? 
  2. Which categories or motifs recur more frequently, and which occur less frequently?

This stage laid the foundation for the following workflow steps, as the research questions determined which metadata and content-related features would be analysed during the initial processing. They also guided the qualitative analysis, metadata processing, and model-based text processing, later helping to interpret and compare the results.

Keywords: knowledge discovery

The aim of metadata processing was to adapt the existing dataset for analysis: to extract the data relevant to the present study from the rest of the dataset and to sort it. The method used was filtering the data with the programming language R and its libraries, and compiling it into a new table. The aim of this stage was to create the conditions for qualitative content analysis — identifying which texts were specifically relevant to the study — and to facilitate further data processing in R and with topic modelling and Word2Vec models. In addition, this stage included a comparison of the main corpus and the subcorpus: how many words appeared in the texts of the subcorpus, and how these compared to the texts in the rest of the dataset. The entire process was carried out in RStudio using R and various libraries: tidyverse, stringr, readxl, dplyr, ggplot2, and writexl.

Keywords: analyzing

The aim of this stage was to carry out a detailed qualitative analysis of the selected fairy tales in order to identify recurring motifs, character types, and narrative elements. During the analysis, 29 fairy tales were read, and a structured Excel table was created on their basis, making it possible to systematise and compare the data. The table included the title of each fairy tale in both Seto and Estonian, the type of bird appearing in the story in both languages, the magical object and its powers, the protagonist, the antagonist, other characters, the main plotlines, and the ending of the story. The categories were defined before reading, with the aim of ensuring that they could apply to every fairy tale. In analysing the content of the fairy tales, artificial intelligence and a Seto dictionary were also used. Both tools helped to understand specific Seto-language expressions or to interpret the content of the texts as accurately as possible.

This stage was connected to the following steps of the workflow because the data collected and categorised during the qualitative analysis formed the basis for content analysis and the application of topic modelling, in the form of the coding table created as a result of this stage. It also helped to answer the research questions more precisely and supported the later interpretation of the results.

Keywords: topic modelling; machine learning; cluster analysis

The aim of this stage was to derive additional synonyms from the keywords found in the texts, for example variants of the word tsirk (bird) and to carry out topic modelling. The intended tools were Word2Vec and LDA. This stage was connected to the earlier stages by using the results obtained from them, and it was also intended to lead to the research results. This stage was unsuccessful, which meant that the research approach had to be adapted and simpler techniques and tools had to be used, despite the fact that the researcher also experimented with applying the method with the support of artificial intelligence (Microsoft Copilot). The obstacles were as follows:

a) Processing the Seto language: since natural language processing is challenging even with tools specifically created for this purpose — for example, using EstNLTK in the context of Estonian — it was especially difficult to work with Seto. Unfortunately, the researcher did not reach a point where it would have been possible to give an adequate assessment of the extent to which existing Estonian-language processing tools are suitable for processing Seto.

b) The researcher lacked previous experience with these tools.

Keywords: comparing

The aim of this stage was to examine the table produced as a result of the qualitative analysis in order to give it a numerical dimension in the form of frequency tables. At this stage, the research questions also received additional answers beyond the qualitative analysis: it was possible to observe whether and to what extent the marvellous bird was referred to generally as a bird or as a specific species, and whether and to what extent there was variation in the plots and main motifs. The work was carried out in RStudio using R and the libraries mentioned above. During the work, the researcher also used artificial intelligence (Microsoft Copilot) to write, organise, and correct the code. The researcher interpreted the results independently.

Keywords: comparing

After processing the coding table, a natural next step was to indicate possible tale types. Since the plots, characters, magical powers, magical objects, and endings differed to some extent, the researcher experimented in RStudio with creating an internal typology of subtypes. The aim was to obtain a more detailed answer to the second research question. The work was carried out in RStudio using R and the libraries mentioned above. During the work, the researcher also used artificial intelligence (Microsoft Copilot) to write, organise, and correct the code, as well as to formulate the approach itself. The researcher interpreted the results independently.

Keywords: diagramming

As the final stage, the results obtained in R were given a visual layer in the form of figures. This included the comparison between the entire corpus and the subcorpus, the frequency tables based on the coding table, and the possible subtypes. The aim was to make the research results more readable for the reader.

Keywords: theorizing

The aim of this stage was to compare and synthesise the preceding stages of analysis in order to answer the research questions and provide a comprehensive overview of the specific features of Seto tales of the magic bird. This comparative approach highlighted both the shared narrative patterns of Seto folk tales and the distinctive features characteristic of individual stories.

Recurring motifs, such as the trials of a poor young man, magical objects, and a just resolution, showed that the Seto tale of the magic bird follows a particular structure: character → consumption of part or all of the magic bird → acquisition of a benefit. At the same time, there is variation in the plot, characters, and outcomes — for example, whether obtaining the benefit is an unintended consequence or a means of achieving a goal. Processing the results in RStudio also revealed this variation numerically and supported the creation of tentative subtypes.

Keywords: data visualization, publishing

The final stage involves visualising and presenting the results in the form of a slide presentation. The results will be presented as part of the University of Tartu course “Cultural Data Project”.