Section: New Results

Computational morphology

Participants : Benoît Sagot, Géraldine Walther.

Although computational morphology has been a topic of interest for Alpage for several years now, several new research topics have received attention in 2011, often in collaboration with morphologists from the Laboratoire de Linguistique Formelle (University Paris 7).

Inflectional morphology

Non-canonical inflection (suppletion, deponency, heteroclisis...) is extensively studied in theoretical approaches to morphology. However, these studies often lack practical implementations associated with large-scale lexica. Yet these are precisely the requirements for ob- jective comparative studies on the complexity of morphological descriptions. We have shown [16] , [43] how the Parsli model of inflectional morphology [132] , which can represent many non-canonical phenomena, as well as a formalisation and an implementation thereof can be used to evaluate the complexity of competing morphological descriptions. After illustrating the properties of the model with data about French, Latin, Italian, Persian and Sorani Kurdish verbs and about noun classes from Croatian and Slovak we have conducted experiments on the complexity of four competing descriptions of French verbal inflection. The complexity is evaluated using the information-theoretic concept of description length. We show that the new concepts introduced in the model by the Parsli model enable reducing the complexity of morphological descriptions w.r.t. both traditional or more recent models.

Derivational morphology

This year, in relation with the ANR project EDyLex (see section  8.2.2 ), work has started targeted towards the acquisition of lexical information at the level of derivational morphology, both using semi- and non-supervized techniques.

Semi-supervized techniques have been used in a work dedicated to French denominal adjectives, for which we have implemented an automatic technique based on large-scale lexicons and corpora for extracting derivation links between base nouns and derived adjectives based on the same stem [46] . The resulting derivational lexicon, which is freely available, has already been partially manually validated. Future work include a full validation and adding denominal adjectives with a suppletive base.

Unsupervized techniques have been used for extraction of derivational links that appear more systematically, although their definition is less linguistically motivated as such [51] .

Morphological issues concerning loan words

Also in the context of the ANR project EDyLex (see section  8.2.2 ), we have carried out a preliminary study on the morphological issues raised by borrowing phenomena, concerning in particular French nouns and verbs borrowed from English [52] . Using techniques that are similar to those used on derivational morphology, we have extracted a significant amount of loan words from a large raw corpus. We have proposed a model of the borrowing phenomenon, that takes into account graphemic (spelling), phonetic and morphological variability.