Click on the project card to read more, view the publication and visit the related repository.

#FrameNet #NLPresources

Following frame semantics, FrameNet offers a database consisting of predicates and arguments.

I was the lead researcher and project manager in Turkish FrameNet project. Turkish FrameNet was created following the methdology used by the French FrameNet team and was annotated using a novel interface created for this task (see FrameNet Annotation Interface project for details). Turkish FrameNet contains 139 Frames in 8 domains: Activity, Cause, Change, Motion, Cognition, Perception, Judgement and Commerce.

#PropBank #NLPresources

PropBank is a database that offers argument structure for verbs and verbal idioms.

I was one of the lead researchers in Turkish PropBank project. Turkish PropBank, TRopBank, contains 17.673 predicates in Turkish and marks their argument structure.

#NLPresources #treebank #UD

BOUN Treebank is a manually annotated UD-style treebank.

I was in the team of linguists who contributed to UD style BOUN Treebank. BOUN Treebank is a UD-style treebank created by BOUN TABILAB as part of a research project funded by TÜBİTAK. I was part of the team that re-annotated BOUN Treebank to improve linguistic accuracy and coherency.

#FrameNet #NLPresources

FrameForm is an open-source tool that facilitates frame annotation for FrameNet.

I was the lead researcher and designer in FrameForm project. FrameForm is a freely available open-source annotation tool that allows annotating frames and semantic roles, creating new domains, and adding sample sentences to frames.

#UD #NLPresources

BoAT is a grammar annotation tool designed for agglutinative languages.

I contributed to enhancements made to BoAT v1. BoAT v2 offers advanced search, collaborative properties, autocomplete, improved annotation interface and more to support the annotator and facilitate the annotation process.

#NLI #NLPresources

CNLI-TR is a challenge natural language inference dataset in Turkish.

I was the lead researcher and designer in CNL-TR project, funded by Boğaziçi University Language Research Center. CNL-TR is a natural language inference dataset that consists of sentence triplets: One potentially De Re De Dicto ambiguous sentence, one De Re paraphrase and one De Dicto paraphrase.


Last updated: 8 Dec 2023