In this short paper we report on experiences gained from bachelor and master theses, and from a series of software projects conducted in cooperation with the Department of Computational Linguistics of the Saarland University. Those bachelor/master theses and software projects were dealing with the application of Natural Language Processing and Semantic Web technologies to the representation and analysis of folktales. Data, codes and results of the software projects have been made available in various repository management services, like GitLab, GitHub or Bitbucket. We think that it will be important to discuss the design of such openly accessible repositories in order to ensure their re-usability and further extensions across various educational institutions.
In the past 3-4 years we proposed in cooperation with the Computational Linguistics (CL) department of the Saarland University a series of bachelor/master theses and software projects, which were dealing with various aspects related to the wider field of folktales and therefore introducing Digital Humanities (DH) topics to students trained primarily to learn and apply computational methods of language technologies.
Our diagnosis was that the approach building on software projects for introducing CL students, and some few students from other departments, to Digital Humanities topics has been very successful. It is also the case that some of the projects we conducted have gained the interest of a broader public, including press coverage1 and a broadcast 1http://derstandard.at/2000004368363/ Wenn-der-Computer-zum-Maerchenonkel-wird programme2. We think that a main aspect of this success story lies in the fact that the students had to work together, building teams for working on modules and meeting for integrating the work done so far.
In all the 4 different software projects conducted until now, we could observe that the folktale topic was a driver calling for participation of a larger group of students (they can choose between different software projects). We describe in the following sections the types of approaches we followed and the results that the students generated and made available on various repository management services, like GitLab, GitHub or Bitbucket. The idea of having software projects as a platform followed the work done by two students in their master and bachelor theses, which were written in the context of their Research Assistant appointments within a larger national project3. We describe briefly the results of all those endeavours in the following sections.
In the context of cooperation between the past DSPIN4 and AMICUS projects5 a master thesis was written by the student Antonia Scheidel on the or http://www.abitur-und-studium.de/ Bilder/Jana-Ott-Christian-Eisenreichund-Christian-Willms-Studenten-vonThierry-Declerck-haben-ein-Programmentwickelt-das-Maerchen-vorlesenkann.aspx
2See http://kulturellebildung.de/fa/ user/Fachbereiche/Literatur_Sprache/ Aktuelles/141121_PRESSE_Erzaehlen.pdf 3We do think that involvement of students as Research Assistant in projects is an important aspect to be considered.
4D-SPIN was a predecessor of CLARIN-D. See https://weblicht.sfs.uni-tuebingen.de/ englisch/index.shtml
5AMICUS: Automated Motif Discovery in Cultural Heritage and Scientific Communication Texts, was a Dutch project dealing partly with the annotation of folktales with recurrent Motifs. See https://ilk.uvt.nl/amicus/ annotation of fairy tales with Propp’s functions6. Vladimir Propp “was a Soviet folklorist and scholar who analysed the basic plot components of Russian folk tales to identify their simplest irreducible narrative elements.”7. Those basic plot elements are called by Propp “functions” and he identified 31 such functions, like “Interdiction”, “Delivery” or “Rescue”, etc. Propp also introduced circa 150 sub-functions that are specialisations of the 31 toplevel functions. Complementary to the functions, Propp identified 7 broad characters, like “the villain”, “the donor” or the “hero”. The “morphology of the tale” described by Vladimir Propp was based on a subset of the so-called Afanasyev collection of Russian Folktales8.
Antonia Scheidel developed a new annotation
scheme according to which fairy tales can be
queried for texts, temporal structures, characters,
dialogues, and Propp’s functions9. The
annotation scheme has been named APftML, standing for
“Augmented Propp fairy tale Mark-up Language”.
Antonia Scheidels’ work is documented in
Syntactic Analysis and a first Ontology
Based on the annotation framework mentioned in
the previous section, Nikolina Koleva has worked
for her bachelor thesis on an automated system
for processing fairy tale texts. She considered for
her work two tales, “The Magic Swan Geese”, an
English version of the Russian fairy tale
“Gusilebedi”, and “Va¨terchen Frost”, a German version
of the Russian fairy tale “Djed Moros”. She has
6See
8See https://en.wikipedia.org/wiki/ Alexander_Afanasyev
9The annotation scheme can be downloaded at http://www.coli.uni-saarland.de/ ˜ascheidel/APftML.xsd
10Examples of such annotated data can be
downloaded at http://www.coli.uni-saarland.de/
˜ascheidel/APftML.xml
written a program that analyzes the text according
to linguistic criteria, with the aim of recognizing
the (main) characters in it, and storing those in a
database. This database is of the “Ontology” type,
on the base of which logical operations can be
performed. The background is a formal description of
what can be found in these fairy tales, including an
ontology about family relations. Thus, the system
can recognize that in the text “the daughter” is the
same person as the “sister” when this is suggested
by the context. This way, recognized characters in
fairy tales are semantically annotated with more
general categories, like “Woman”. And we then
know in which contexts (or situations) a specific
family member (for example the “daughter”) is
involved (see
Once we had those resources, i.e an annotation framework for folktales, based in a first instance of the mark-up of Proppian functions, and an ontology framework in which characters playing a role in folktales are stored as instances of domain-specific classes, the idea was to extend those to a larger framework supporting DH application scenarios. 4
In a first software project which was building on the top of the two resources mentioned in the previous sections, a division of work could be established between the four members of the project team. One task consisted in offering a meaningful segmentation of the tales. The approach for this consisted in automatically segmenting the tales along the lines of the dialogue structure. This had one motivation: to offer a base for the integration of a text-to-speech system supporting the “read aloud” of a tale, in which voices are associated to each contributors to the dialogues (and for sure one voice for the narrator). This application is described in more details in the next section.
The students worked in this project mainly on the English version of the “Froschko¨nig” tale (The Frog Prince)11. Following those new steps, the initial annotation format has been augmented with detailed dialogue descriptions. And the ontology has also been extended, including now a description of dialogues (questions, answers, monologues etc.), including the encodings of the participants 11See https://en.wikipedia.org/wiki/The_ Frog_Prince and the dialogue turns. In the two most recent and currently still running software projects the students are implementing a strategy on additionally segmenting a tale by the locations in which events are occurring. There is an interesting correlation between the segmentation by dialogues and the one by locations, as in this kind of narratives the participants to a dialogue are often sharing a location. 5
One student had the task to implement a program able to detect emotions. For this the original annotation scheme has been extended, supporting the mark-up of 6 basic emotions (fear, grief, joy, etc.), which are also encoded in the ontology. The automatic processing of the text (based in this case on the NLTK package12) was then marking the emotion detected in one sentence, on the base of a emotion lexicon build from annotated examples that served a as seed that was completed by consulting the WordNet13 module implemented in NLTK14.
A major extension of the past work in this
software project was that synthetic voices also play a
role. Once a character has been recognized, for
example the princess (in the fairy tale “Frog King”)
additional features are coded (for example age,
gender, emotion, etc.). Then a previously defined
synthetic voice is automatically added to the
character. And when the text is processed by the system,
the story can be “told” by the voices. If there is
no detected character in a dialogue situation, it is
assumed that the narrator is the speaker and the
reader the receiver. A demo can be heard in the
corresponding Bitbucket repository15. In this software
project we made use of the “Mary” Text-To-Speech
System16. The overall results of the projects are
described also in
12NLTK stands for “Natural Language Toolkit” and is written in Python, including a lot of corpus processing and statistical libraries. See http://www.nltk.org/
13See https://wordnet.princeton.edu/ for more details.
14See http://www.nltk.org/howto/wordnet. html for more details.
15The data, algorithms and results of the projects are stored in https://bitbucket. org/ceisen/apftml2repo. A demo of the TTS application is available at: https: //bitbucket.org/ceisen/apftml2repo/src/ cbf4d71de7f96146d17c4c84572ceb9a99cd300f/ example%20output/audio_output.mp3?at= master&fileviewer=file-view-default 16See http://mary.dfki.de/ for more details.
We described in sections 4 and 5 how the original
ontology has been enriched with additional features.
In a second software project, work was dedicated
in the ontologisation of classical knowledge –
indexation and classification – resources in the field
of folklore. We were considering in this software
project two such resources: The “Motif-index of
folk-literature”
The second resource builds on former work by
Antti Aarne
Our approach consisted in extracting from those knowledge resources, which are stored in different formats, classification relevant information and to re-organize them in two interrelated ontologies, using for this the W3C standards OWL19, RDF(s)20 and RDF21.
The integrated ontology resulting from the software project, also after curation done in the context of an internship at DFKI, contains 46,950 motifs for the TMI domain and 2802 elements for the ATU domain, most of them interrelated by corresponding properties. Results of this software project are available in a 17https://sites.ualberta.ca/˜urban/ Projects/English/Motif_Index.htm.
18https://en.wikipedia.org/wiki/Motif_ (folkloristics)
19See http://www.w3.org/TR/owlsemantics/.
20See http://www.w3.org/TR/rdf-schema/ formoredetails.
21See https://www.w3.org/RDF/ for more details. GitLab repository: https://gitlab.com/ folktaleclassification/.
An application of this new integrated ontology
for the classification of characters in folktales has
been presented in
We did report on specific teaching activities in the field of the representation and processing of folktales by students (mainly) in the field of computational linguistics. The specificity of the experiences we are reporting is that those activities took place in the context of software projects or internships, thus with a focus on practical implementation and development works. We noticed that this kind of team work, or also compact work done in the context of an internship, is delivering a very large amount of resources that are potentially very relevant for being reused in other type of teaching activities. Maybe also a coordinated action between universities and other educational institutions toward the organization of such software projects could be an idea to discuss and implement. Last but not least, many of the results presented in this short paper have been submitted to and accepted at relevant workshops and conferences, bringing the students thus also closer to this type of academic achievements. tation of characters in folktales. In Kalliopi Zervanou and Antal van den Bosch and, editors, Proceedings of the 6th Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities (LaTeCH 2012), pages 30–35, 209 N. Eighth Street Stroudsburg, PA 18360 USA, 4. Association for Computational Linguistics (ACL), ACL.