HUM data lab > Workflows > Analysis of Regional and Temporal Patterns in Estonian Instrumental Folk Music

Analysis of Regional and Temporal Patterns in Estonian Instrumental Folk Music

This workflow enables the analysis of the regional and temporal distribution of instruments and instrumental tunes in a dataset of Estonian instrumental folk music. During the workflow, the metadata of the archival dataset are cleaned, instrument and place names are standardised, and broader analytical instrument categories are created as needed. The frequencies of instruments and folk tunes are then calculated by county and time period. The results are presented as frequency tables, diagrams, and maps, which help make visible patterns that would be difficult to detect by looking at individual records.


The original dataset used in the workflow and the standardised dataset created on its basis are available as csv-files in the code repository of the Estonian Literary Museum. The original dataset is based on the instrumental folk music collections of the Estonian Folklore Archives (EFA) and brings together information on sound recordings, handwritten musical notations, and related metadata. The original dataset contains 13,036 observations and 24 variables, including data on the tune, year, instrument, performer, parish, county, place of origin, collector, data type, and references to sound recordings and musical transcriptions. The data cover the years 1874–1999. In this workflow, only the columns necessary for the research questions were used from the standardised dataset: year, county, and standardised instrument name (instrument_puhas). One record in the dataset was treated as one folk tune, and no more detailed distinction by genre or tune type was made in the context of this work.

Workflow steps

Keywords: Discovering, contextualizing

The aim of this stage was to gain an initial overview of the content, structure, and context of the dataset. The dataset was available in CSV format, so we did not need to create the table ourselves. In familiarising ourselves with the dataset, we examined which fields it contained, how complete the records were, and which variables were most important for our research questions. Since our work focuses on the relationships between instruments, instrumental tunes, time, and regions, it was important to assess at an early stage whether the data were consistent enough to be compared later. At this stage, we also refined the research questions and mapped which characteristics of the dataset might affect the interpretation of the results.

Alongside familiarisation with the dataset, we also began to study its historical and cultural context. We worked through sources dealing with the tradition of Estonian instrumental folk music, regional specificities, and collecting practices, including the introduction of recording devices and the importance of musical notation skills in earlier collecting work. Familiarisation with the context was important preparatory work before the cleaning stage, as it made it possible to identify which different spellings, dialectal names, and possible synonyms referred to the same instrument and therefore needed to be standardised in the subsequent analysis. In addition, contextual familiarisation was important for interpreting the results, because the dataset is not neutral and may not reflect actual musical practice. It primarily reflects what was collected at different times, and the collected records may in turn have been influenced by the development of collecting practices.

Keywords: Preprocessing, discovering

The aim of this stage was to prepare the dataset for further analysis so that names referring to the same instruments would be presented as consistently as possible. Since the dataset was formed through the work of different collectors, regions, and time periods, the metadata are not always uniform. The same instrument could appear under several spellings, dialectal names, or with additional qualifiers. If such differences are not standardised, later statistical analysis may produce misleading results, because occurrences of the same instrument would be artificially divided across several names. To prepare for standardisation, we used the software R and created a frequency table of instrument names in RStudio, which we used together with contextual knowledge as the basis for compiling a correspondence table. This made it possible to carry out the standardisation systematically and in a verifiable way, since for each original name it was possible to trace how it was renamed for the analysis.

We then used R to clean the spelling of instrument names. First, we made a copy of the instrument column in the dataset in order to preserve the original spelling. We then converted the text to lowercase, removed unnecessary qualifiers and punctuation marks, and grouped different names for the same instrument under a common name. Since the final aim of the work was descriptive statistical analysis, in some cases we also grouped substantively related instruments into broader categories. For example, under the category “bellows instrument” (lõõtsaline), we grouped the accordion, bayan, kamorška, lõõts, and similar instruments. This choice helped the later analysis reveal broader temporal and regional patterns rather than only the fragmented distribution of individual names.

Keywords: data visualization

The aim of this stage was to gain an initial overview of the content, structure, and context of the dataset. The dataset was available in CSV format, so we did not need to create the table ourselves. In familiarising ourselves with the dataset, we examined which fields it contained, how complete the records were, and which variables were most important for our research questions. Since our work focuses on the relationships between instruments, instrumental tunes, time, and regions, it was important to assess at an early stage whether the data were consistent enough to be compared later. At this stage, we also refined the research questions and mapped which characteristics of the dataset might affect the interpretation of the results.

In R, we used frequency counting and table creation. We compiled the results into tables that made it possible to compare the occurrence of instruments and instrumental tunes. The initial statistical description helped us see the general distribution of the dataset and was necessary preparatory work for further analysis and visualisation. Based on the results obtained, we decided which variables it would be meaningful to compare.

Keywords: Exploration, data visualization

The aim of this stage was to expand the analysis based on the initial results and to visualise the findings. We grouped the data by decades and counties in order to see temporal and regional patterns more clearly. In the case of this dataset, year-by-year analysis may be too detailed and uneven, since collecting activity did not take place uniformly every year. The use of decades helps reduce random fluctuation and better highlights longer-term changes.

We used R to create line charts and bar charts and to calculate frequencies. In addition, we calculated statistics concerning the dataset metadata, based on the additional questions mentioned earlier, which are potentially necessary for explaining nuances in the results and pointing out limitations. Alongside the temporal dimension, the spatial dimension was also important, so we exported the results from R and used them in QGIS to create map layers. The maps made it possible to show in which regions certain instruments occur more frequently. As the basemap, we used the administrative division of 1917 from the Estonian Land Board, because it suited the historical character of the dataset better than the present-day administrative division. Since Petseri County did not yet exist in the 1917 administrative division, Petseri was added from the 1922 administrative division. The Setomaa records were associated with the Petseri area, because in the historical administrative division used, Setomaa was located in that area.

Keywords: Interpreting, contextualizing

The aim of this stage was to connect the results of the analysis with the broader historical and cultural context of Estonian instrumental folk music. We interpreted what the regional and temporal patterns might indicate about the distribution of instruments and the diversity of instrumental folk music. The aim of this stage was not only to describe where something occurs more or less frequently, but also to consider why such differences may become visible in the dataset.

In interpreting the results, we also took into account the limitations of the dataset. The visible patterns may not always directly reflect historical musical practice, but may also result from the history of collecting, preservation, digitisation, or inconsistencies in metadata. For this reason, we connected the statistical results with the knowledge gathered during the earlier contextualisation stage and treated the conclusions cautiously.