Home

Awesome

Summary

Turkish FrameNet consists of 2,700 manually annotated example sentences and 19,221 tokens. Its data consists of the sentences taken from the Turkish FrameNet Project. The annotated sentences can be filtered according to the semantic frame category of the root of the sentence.

Introduction

Turkish FrameNet consists of 2,700 manually annotated sentences and 19,221 tokens. The corpus is taken from the Turkish FrameNet Project where the Turkish verbs were annotated according to their semantic arguments. The sentences are like dictionary examples, they include made-up full sentences created by native speakers of Turkish.

The corpus is divided into 139 semantic frames according to the verb meaning. The number of sentences in frames varies according to how many verbs Turkish language has in that category. Each frame has its own dev, test, and training files.

Acknowledgments

We wish to thank the Starlang Software for funding and supporting this work.

References

*The Turkish FrameNet:

Marşan, B., Kara, N., Özçelik, M., Arıcan, B. N., Cesur, N., Kuzgun, A., ... & Yıldız, O. T. (2021, January). Building the Turkish FrameNet. In Proceedings of the 11th Global Wordnet Conference (pp. 118-125).

Changelog

<pre> === Machine-readable metadata (DO NOT REMOVE!) ================================ Data available since: UD v2.8 License: CC BY-SA 4.0 Includes text: yes Genre: grammar-examples Lemmas: converted from manual UPOS: converted from manual XPOS: converted from manual Features: converted from manual Relations: converted from manual Contributors: Cesur, Neslihan; Kuzgun, Aslı; Yıldız, Olcay Taner; Marşan, Büşra; Kuyrukçu, Oğuzhan; Arıcan, Bilge Nas; Sanıyar, Ezgi; Kara, Neslihan; Özçelik, Merve Contributing: elsewhere Contact: nesli@starlangyazilim.com / olcay.yildiz@ozyegin.edu.tr =============================================================================== </pre>