Integrating a Verb Lexicon into a Syntactic Treebank Production

The creation of linguistically interpreted corpora is a tedious task and the automation of the annotation process is indispensable. A fully automated annotation is hardly possible to achieve, since it requires very sophisticated and large knowledge bases which are, themselves difficult to create. However, a "machine-aided approach" to the annotating process, using as many as available sources of linguistic information is justified and desirable. The creation of an HPSG-based syntactic treebank of Bulgarian [18] is a process of incremental augmentation of the real-world sentences with linguistic annotation as the result of linguistic analysis at different processing levels. At this stage of the treebank production, there are the following main sources of linguistic knowledge that can be integrated into a grammar for automated chunk connection and assignment of grammatical relations: