Converted from an OASIS Open Document
The paper deals with the observations that are made during the work on Dracor.org
(Skorinkin et al 2018). Originally, Dracor.org is a set of isolated corpora in
different languages with a unified TEI markup that allows you to do quantitative
drama studies (Skorinkin / Fischer 2018). But since the platform now includes a
noticeable set of corpora in German, English, Russian and other languages, it is
natural to try to study cross-corpora drama features. These features include
characters that appear in different plays and even in different corpora. Usually
these are historical personalities (e.g. kings), mythological figures (Achilles,
Heracles), popular literary characters (Faust). The idea is to mark all the
characters that are mentioned in the plays more than once. We must not forget that
Euripides' Achilles (Iphigenia in Aulis) is a different
character than Kleist's Achilles (Penthesilea). Technically,
these characters are part of different fictional systems. But in the future we would
like to explore the differences of interaction between these characters and the
other components of the play. This will help to establish the very difference
between fictional systems and the author's style. So we have to mark the heroes with
the same name to be able to extract them automatically from the corpus.
All characters in Dracor have their ids, which depend on their names:
<person xml:id="Othello_Oth">
<persName>Othello</persName>
</person>
But id in the play is not suitable for cross-corpora annotation tasks. For example, the character "messenger" is found in many plays, and often has the same id, but this is not the case when we are going to mark these characters as the same cross-corpora character. In addition, characters in different plays may occasionally have the same names. That means we need to have a separate attribute for this kind of case.
We have chosen the attribute "ana" for the tag "person". Due to the TEI guidelines this attribute "indicates one or more elements containing interpretations of the element on which the ana attribute appears". This means that we interpret the character of the play as some kind of entity that may appear in other plays. The most convenient option would be to associate the value of this attribute with the Linked Open Data. We use Wikidata, which provides the URI for the entity we find in different plays.
<person xml:id="phaedra" sex="FEMALE" ana="http://www.wikidata.org/entity/Q212818">
This approach works very well and allows us to identify the characters in different plays. But there are cases where we can see the inconvenience of Wikidata URI for the purposes of our cross-corpora markup.
There are three types of such cases.
In ancient plays, characters and their shadows are often found separately. Sometimes they are found within the same play. Then they have different dracor ids, but the URI on Wikidata is the same. See Tantalus and Tantali Umbra in
Thyestes by Seneca
<person xml:id="tantalus" sex="MALE" ana="http://www.wikidata.org/entity/Q515531">
<persName>Tantalus</persName>
</person>
<person xml:id="tantaliumbra" sex="MALE" ana="http://www.wikidata.org/entity/Q515531">
<persName>Tantali Umbra</persName>
</person>
and Δαρεῖος and Εἴδωλον Δαρείου in
Persians by Aeschylus:
<person xml:id="eidolon_dareiou" sex="MALE" ana="http://www.wikidata.org/entity/Q44387">
<persName>Εἴδωλον Δαρείου</persName>
</person>
<person xml:id="dareios" sex="MALE" ana="http://www.wikidata.org/entity/Q44387">
<persName>Δαρεῖος</persName>
</person>
At the moment we have to mark these characters with a single id. But it's not the best solution.
The purpose of cross-corpora markup is to tie together characters from different plays and different corpora at the same data object. From this point of view, it is obvious that Zeus and Jupiter are the same character, they perform the same function in the story about Hercules. But in Wikidata we find different URIs for Zeus (https://www.wikidata.org/wiki/Q34201) and Jupiter (https://www.wikidata.org/wiki/Q4649). Thus, our data is separated. The plan is to get all the necessary information about the characters using SPARQL queries and characters ids. Perhaps, this case could be solved by the additional parameter of SPARQL query (P460). But every time this parameter must be different (depends on gods, kings or someone else we deal with).
Talthybius has two relevant objects in Wikidata. First of all, it's a character from a myth (https://www.wikidata.org/wiki/Q1367860). This URI allows us to connect the character of the tragedy
Troades by Seneca and
The Trojan Women by Euripides. But the Euripides tragedy character has his own URI in Wikidata: https://www.wikidata.org/wiki/Q60607554. Technically, its attribute should have this value, but it contradicts the idea of tying the characters of different plays together. Additional SPARQL parameter P1074 ("fictional analog of") could solve this problem. Another solution is to extend Wikidata by ourselves and add Talthybius to the parameter "present in work" (P1441) from Dracor to the certain play objects.