Welcome to the online chiasmus detector

Use the default text as example (Sherlock Holmes anthology)
Output XML version (more technical but more steady)

The detector will send you an email with its containing chiasmi.


If you use this demo please cite this paper

In the upper boxes, write your email and write a title for instance "Sherlock"
In the lowest box paste the content of the text you want to explore, for instance Sherlock Holmes novels (or you can just tick the checkbox and use Sherlock Holmes Anthology)
The machine will send you an email containing your chiasmi ranked by level of relevance.


This demo is based on the paper of:
Dubremetz, Marie and Nivre, Joakim (2015). "Rhetorical Figure Detection: the Case of Chiasmus" in Proceedings of the Fourth Workshop on Computational Linguistics for Literature pages 23–31, Denver, Colorado, USA, June. Association for Computational Linguistics.
[PDF], [Bibtex], [Slides]
Make sure you cite it if you use this demo for research.

Have a question?

Feel free to contact the author(s) on the dedicated mailing list: chiasmus@cl.lingfil.uu.se


What is a chiasmus?

See the second paragraph of this article (the paragraph starts by "Chiasmi are a family of figures that consist in re- peating linguistic elements in reverse order."...)

Why my text has to be MORE than 500 words?

There is two reasons. First antimetabole/chiasmus are extremely unfrequent thus small texts are really unlikely to contain any. For instance in his novel River War, Winston Churchill makes only 1 chiasmus even though the novel is 300 pages (= 150 000 words). Second this detector looks at context around the chiasmus (a few words before and a few words after the sentences), that helps calculating the probability that your chiasmus is a real one. Thus if you just put a small text like "Last should be first, first should be last.", the engine might completely crash and output nothing because it expects more words around. So, please, the more the merrier! Put large texts, millions of words if you want!

Does it work for *Italian/Chinese/Esperanto/Other languages...*

Nope. The algorithm has a lot of English dependant features (for instance it downgrades instances of accidental chiasmus due to stopwords). The sentence '*The* cat pursue *a* mouse but *a* trap was on *the* way.' is an instance of accidental chiasmus due to stopwords. Such false positive is extremely frequent in texts and needs to be filtered.

About the Author

Back to Marie Dubremetz's website Flag Counter