Presented at The Third Workshop on Computational Models of Narrative (CMN'12), in conjunction with the conference on Language Resources and Evaluation (LREC), Istanbul, Turkey, May 26, 2012.

1. TrollFinder: Geo-Semantic
Exploration of a Very Large Corpus
of Danish Folklore
Peter M. Broadwell
Postdoctoral Scholar,
Digital Initiatives & Information
Technology
UCLA Library
Timothy R. Tangherlini
Professor, Scandinavian Section and
Department of Asian Languages
UCLA

2. Target Domain:
The Folklore Collections of Evald Tang Kristensen

3. It was the old counselor from Skårupgård who came
riding with four headless horses to Todbjærg church. He
always drove out of the northern gate, and there by the
gate was a stall. They could never keep that stall door
closed. They had a farmhand who closed it once after it
had sprung open. But one night, after he’d gone to bed,
something came after the farmhand and it lifted his bed
straight up to the rafters and crushed him quite hard. The
farmhand shouted and asked it to stop lifting him up
there. “No, you've tormented us, and now you’ll die…” I
heard that’s how two farmhands were crushed to death.
He wanted to close the door and then they never tried to
close it again.
Evald Tang Kristensen, Danske Sagn vol. 4, no. 650
Told by Ane Margrete Jensdatter in October 1889

4. The Evald Tang Kristensen
Collection of Danish folklore 1.0

5. ETK 1.0: story view

6. Topic and index browsing in the Danish
folklore archive

7. The ETK Danish folklore
collection version 2.0

8. The 2011 Shoah Foundation Institute RIPS Team
Rodrigo Mendoza Smith (ITAM), Margo Smith (Kenyon),
Anna Kuznetsova (Duke), Peter Sugihara (Bard/Columbia)

9. Interface to the USC Shoah
Foundation Visual History Archive

10. • Insert picture from iWitness – how it sucks
Shoah Foundation VHA map interface

11. Desired feature: place-specific topic
keyword suggestion
Topic
sabotage
ghetto escapes
organizations
…
Rank
1
2
3
...

13. Core approach: build a
place-to-topic
co-occurrence matrix
Grinderslev
Nykøbing Mors
...
...
...
witch cooking
10
1 35
621
...
... ... ...
......

14. Topic heat map with bar
graph overlays for “witch”
Grinderslev

15. Normalizing by population density
Population density based on 1901 census Population density circa 2000

16. Story density: story-to-place
mentions per km2
Normalized story density:
places mentioned/person/km2

17. Not normalized Normalized by population density
Effects of population density-based
normalization for keyword “witch”
GrinderslevGrinderslev

18. Grinderslev? Really?

19. From the Danish Folklore nexus…

20. Why Grinderslev??
• Site of a well-known Augustinian monastery, Grinderslev kloster, founded
in the twelfth century.
• The monastery was built near a holy spring, Breum kilde, but was
abandoned in the aftermath of the Reformation.
• The spring at Breum was subsequently associated with witchcraft
• In 1686, Anne Madsdatter and her sister were burned at Breum, the last
witch burning in Denmark (Bruun, 1920).
• Although this episode is well known in the study of Danish witchcraft, the
persistent relationship between the area surrounding Grinderslev and
stories about witchcraft has not been recognized previously, suggesting a
topic for further, in-depth inquiry.
• Only a few of the stories mention Grinderslev (but rather places near
Grinderslev such as Breum)

21. Who you gonna call?
Ministers (blue)
Cunning folk (red)
Ghosts (blue)
Revenants (purple)

22. DigDag: The Digital Atlas of Danish
Historical-Administrative Geography

23. A spatial query tool
etkspace.scandinavian.ucla.edu/maps/spatialquery.html
Flipping it around

24. Term Frequency – Inverse
Document Frequency
(TF-IDF)

25. Locations as Documents

26. RF-IPF: Ranking topics
in a geographic region
RF-IPF = RF * log( |P| / |p ∈ P : t ∈ p| )
RF = region frequency: the number of times topic t co-occurs
with places in the region, normalized by the total number of
place/topic co-occurrences in the region
|P| = total number of places mentioned in the corpus
|p ∈ P : t ∈ p| = total number of places in the corpus that co-
occur in stories with the topic t

27. Place/topic co-occurrences in the
vicinity of Grinderslev
Raw Normalized RF-IPF
1. bande (to curse)
2. hale (a tail, prob of a
snake)
3. sølv (silver)
4. tigge (to beg)
5. vælte (to tip)
6. læse (to read)
7. flyde (to flow)
8. paste (to take care of
animals)
9. øre (ear)
10. herre (lord)
11. lindorm
(supernatural snake)
12. stille (to place)
13. østen (to the east)
1. kusk (carriage driver)
2. reste (remainder)
3. grønning (village green)
4. rådelig (recommended)
5. boel (a large farm)
6. kristenblod (Christian
blood)
7. om kap (race or
competition)
8. indhylle (enshroud)
9. søkke (to sink down)
10. mæt (sated)
11. konfirmation
(confirmation)
12. tjørn (hawthorn)
13. mane (to conjure)
1. paste (to take care of
animals)
2. flyde (to flow)
3. hale (tail)
4. grønning (village green)
5. borggård (fort)
6. sølv (silver)
7. bande (to curse)
8. søkke (to sink down)
9. herre (lord)
10. læse (to read)
11. mane (to conjure down)
12. lindorm (supernatural
snake)
13. vælte (to tip over)

28. TrollFinder: Geo-Semantic
Exploration of a Very Large Corpus
of Danish Folklore
Peter M. Broadwell
Postdoctoral Scholar,
Digital Initiatives & Information
Technology
UCLA Library
Timothy R. Tangherlini
Professor, Scandinavian Section and
Department of Asian Languages
UCLA