The past two days were spent at the premises of Helsinki Collegium of Advanced Studies, where I participated the Big Data Approaches to Intellectual and Linguistic History Symposium. In this blog entry, I will grasp briefly some discussed topics.

Well, was it the first properly organized digital humanities event in Finland? I reckon yes, even though the loosely operating Digital Humanities Finland Network has organized a couple of symposia during 2013 and 2014 in Helsinki. When speaking to the organizers of this symposium, they indicated that several organizations, at least across the Helsinki central campus, had wished something more concrete to happen and take place – in my opinion, this symposium responded to many needs in deed. Thank you for organizers and all the participants, this was a high quality event and I hope the forthcoming events will be equally successful.

Since I am not going to cover here all the talks held in this symposium, you may retrieve the presentations here (Sign is as a quest). The twitter stream of the event can be tracked via hashtag #HDHD.

First of all, the big data in humanities isn’t that big after all. And again: “There’s no big data in humanities!” This notion was repeated several times during the first day of the symposium. It is often fragmented, limited to available content and research questions. Also, the big data projects in humanities tend often to be quite experimental.

A good example of this was the Mapping Notes and Nodes project, which tries to develop methods to analyze/visualize potential meaningful relationships between artists and intellectuals by combining biographical data with relevant contextual information for the history of the creative industry. The project was presented by Charles van den Heuvel and he indicated strongly that full data integration is impossible and there has to be a shift in focus to integration of big and little heterogenous datasets in order to enable the interaction of end-users, who’ll select which data they want to use.

The MNN project has greatly benefitted of network theory and when visualizing the connections of each individual correspondent – more network layers have become visible. The spatial or social connections are made visible in several layers: artefact, person, society, nuclear family, site, city, location… Some examples of the outcomes of this project can be examined here. Something similar was made by a Finnish historian Jarkko Keskinen at the University of Turku, but unfortunately the datasets are not yet open for the public to my understanding.

Maybe I will take another example on the selectivity of data too. Michael Cysouw is rather unique linguist, who utilizes parallel texts in many languages in his current research – 137 languages, if I am not completely mistaken.

Cysouw paid attention to the quality of the data. According to him, many researchers are reluctant to use big data approaches due to skepticism towards the data quality. What if the 80% of the data would be good and 20% of it would contain obvious mistakes?

Cysouw seems to wave a finger here and sees the combination of man-machine approach as fruitful here. For him the bigger picture is clear: he would let the computers to deal with good quality data and let the specialists to work with problematic cases. Only the detail and minutiae that can be told with big data are the real improvement in research. Yes, the large number of cases allow for much detail, but the trick is that the detail must be identified first. The question arises, are we (finally) heading back to the close reading in so-called big data research too?

Yours &c.,
Jussi-Pekka