Hybridization and Treebank Enrichment with Constraint-Based Representations

International audience We present in this paper a method for hybridizing constituency treebanks with constraint-based descriptions and enrich them with an evaluation of sentence grammaticality. Such information is calculated thanks to a two-steps technique consisting in : (1) constraint grammar induction from the source treebank and (2) constraint evaluation for all sentences, on top of which a grammaticality index is calculated. This method is theoretically-neutral and language independent. Because of the precision of the encoded information, such enrichment is helpful in different perspectives, for example when designing psycholinguistics experiments such as comprehension or reading difficulty.