Section: New Results

Syntax and treebanking

Participants : Djamé Seddah, Benoît Sagot, Kim Gerdes, Benjamin Muller, Pedro Ortiz Suárez, Marine Courtin.

In 2019 we have introduced the first treebank for a romanized user-generated content of Algerian, a North-African Arabic dialect called Arabizi. It contains 1500 sentences, fully annotated in morpho-syntax and universal dependencies, and is freely available. We complement it with 50k unlabeled sentences that were collected using intensive data-mining techniques from Common Crawl and web-crawled data. Preliminary results show its usefulness for POS tagging and dependency parsing.

We have also developed the first syntactic treebank for spoken Naija, an English pidgincreole, which is rapidly spreading across Nigeria. The syntactic annotation is developed in the Surface-Syntactic Universal Dependency annotation scheme (SUD) [77] and automatically converted into Universal Dependencies (UD). A crucial step in the syntactic analysis of a spoken language consists in manually adding a markup onto the transcription, indicating the segmentation into major syntactic units and their internal structure. We have shown that this so-called “macrosyntactic” markup improves parsing results. We have also studied some iconic syntactic phenomena that clearly distinguish Naija from English. This work is published in [36].

We have carried out two pilot studies in empirical syntax based on UD treebanks. In a first study [38], we investigate the relationship between dependency distance and frequency based on the analysis of an English dependency treebank. The preliminary result shows that there is a non-linear relation between dependency distance and frequency. This relation between them can be further formalised as a power law function which can be used to predict the distribution of dependency distance in a treebank. In a second study [40], we discussed an empirical refoundation of selected Greenbergian word order univer-sals based on a data analysis of the Universal Dependencies project. The nature of the data we worked on allows us to extract rich details for testing well-known typological universals and constitutes therefore a valuable basis for validating Greenberg's universals. Our results show that we can refine some Greenbergian universals in a more empirical and accurate way by means of a data-driven typological analysis.

Finally, we have introduced a new schema to annotate Chinese Treebanks on the character level. The original UD and SUD projects provide token-level resources with rich morphosyntactic language details. However, without any commonly accepted word definition for Chinese, the dependency parsing always faces the dilemma of word segmentation. Therefore we have presented a character-level annotation schema integrated into the existing Universal Dependencies schema as an extension [39]. The different SUD projects were also presented at the Journées scientifiques “Linguistique informatique, formelle et de terrain” (LIFT 2019), Nov 28-29, 2019 at the University of Orléans.