Universals of Language
The quest for language universals - rules and principles that are the same for all languages in the world despite their great apparent diversity - has always captured the interest of linguists and of the general public: Do only humans speak? Why are there so many different human languages?
In the twenty-first century, we can study these questions based on very large amounts of digital data and computational models. Based on very large-scale corpora annotated with grammatical information, called treebanks, current work from my group has identified some strong quantitative tendencies across many languages, in modern languages and across time. For example, we have found that Romance languages tend to put adjectives as close as possible to nouns (so they prefer to say la bella casa al mare, adjective before the noun rather than la casa al mare bella) or that they have developed from Latin in a way that these distances have become shorter with time.
On a larger scale, we develop computational models to answer questions such as: are these processes of dependency minimization functionally determined, to make communication efficient and to minimize memory effort, or do they derive from innate grammatical principles universal to all languages?
Course of spring semester 2018: Traitement de langue - approches linguistiques et approches empiriques
Current and Recent Research Projects
Current and Past Research Topics
- Computational Methods for Language Universals
- Function and Semantic Role Parsing
- Corpus-based Techniques for the Acquisition of Verb's Argument Structure applied to Several Languages
- Corpus-based Accounts of PP Attachment Ambiguity Resolution
- Architecture and Frequency in Human Acquisition and Sentence Processing