The Riddle of Literary Quality

Midterm meeting The Riddle

In the last week of January 2014 the Riddle team held a two-day meeting with the members of their Think Tank. During these two days the project members presented the first accomplishments within the overarching project, as well the results achieved within the individual PhD projects, in order to discuss the progress and to be advised on the proposed next steps.

On the first day, project leader Karina van Dalen-Oskam gave a general introduction of the project, followed by brief introductions of the aims of the PhD candidates Corina Koolen, Andreas van Cranenburgh and Kim Jautze. Gertjan Filarski and Hayco de Jong presented the technical set-up of the project in terms of the microservices which are under development. The leading discussion of the afternoon was about the results of the survey Het Nationale Lezersonderzoek, led by Karina van Dalen-Oskam and Erica Nagelhout. The main question is: what is the best methodology for analyzing the rich source of data from the survey? The most challenging question to the Think Tank was in which ways the continuum of possible readers’ roles might be subdivided. Which methodological steps and statistical approaches could be used to establish which respondents, based on their answers to the sixteen statements about their habits as readers, take a distanced or an identifying stance?


The next morning, Sally Wyatt introduced the Computational Humanities Programme of the KNAW, of which The Riddle is one of the four projects. Corina, Andreas and Kim presented the first results on two classification tasks on genres with high- and low-level features. First a set of low-level features was discussed that has been analyzed in the ‘literary thriller-corpus’ with regard to sentiment and surface characteristics, in order to examine whether the literary thriller has textual features that deviate from those of thrillers and literary novels.

Project Riddle -sentiment analysis- smileys

Subsequently, Andreas presented results on automatic classification of these genres with surface characteristics, and showed on the ‘chick-lit corpus’ (which was already parsed, whereas the literary thriller-corpus was not) how high-level (syntactic) features improve classification compared to just low-level features. The meeting was closed with the discussion about the next phase of the project. The PhD candidates and Think Tank members discussed what other features could be analyzed that are less related to writing style, but more to content and narrative, and what methods could be applied. Another interesting topic is how the methodology of measuring readers’ opinions and relating them to textual features could be applied to other languages, to see if there are corresponding findings. One of the first aims is to examine Frisian literature. With the Think Tank members, especially the two Frisian specialists Hanno Brand and Jelle Krol, it was discussed if it would be feasible to set up a Frisian survey, and what kind of fiction would be included in the corpus.
We would like thank the researchers who have made the effort to travel to The Hague (David Hoover and Patrick Juola flew in all the way from the US) for their inspiring contributions to the presentations and discussions. We are happy to say that we have had an incredible amount of input which inspires us for the upcoming second half of the project, and that we are looking forward to inviting them again by the time we have more insight into our Riddle.