Summary

Turkish-ITCC contains a conversion of coreference annotations of 24 Turkish documents [1, 2]. The annotations include overt and implicit mentions (e.g., dropped pronouns and null subjects) with their coreferential relations. The first end-to-end baseline CR results on Turkish_ITCC v1.2 were provided in [2].

Update history of the dataset is as follows:

2023 - The annotations for dropped pronouns and null subjects were added and the data was also reannotated to eliminate missing annotations [1, 2]. 2022 - The data from [3] was converted to CorefUD format and included in the CorefUD v1.1 release by ITU NLP Group. 2018 - [3] provided the data in CoNLL format. Only 24 documents from [4] could have been continued to be used due to alignment issues between the annotations provided in [4] and the original text in [5]. 2017 - [4] provided the first coreference annotations in XML format on 33 documents selected from METU Turkish Corpus (MC) [5].

References

[1] @inproceedings{pamay-arslan-eryigit-2023-incorporating,
    title = {Incorporating Dropped Pronouns into Coreference Resolution: The case for {T}urkish},
    author = {Pamay Arslan, Tu{\u{g}}ba and Eryi{\u{g}}it, G{\"u}l{\c{s}}en},
    editor = {Bassignana, Elisa and Lindemann, Matthias and Petit, Alban},
    booktitle = {Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics: Student Research Workshop},
    month = may,
    year = {2023},
    address = {Dubrovnik, Croatia},
    publisher = {Association for Computational Linguistics},
    url = {https://aclanthology.org/2023.eacl-srw.2},
    doi = {10.18653/v1/2023.eacl-srw.2},
    pages = {14--25},
}

[2] @article{pamaybaseline,
    title={Exploring Deep Learning Methods for Turkish Coreference Resolution},
    author={Pamay Arslan, Tu{\u{g}}ba and Eryi{\u{g}}it, G{\"u}l{\c{s}}en},
    journal={Available at SSRN 4588716},
    doi = {10.2139/ssrn.4588716},
    year = {2023},
}

[3] @inproceedings{pamay2018coref,
    title={Turkish Coreference Resolution},
    author={Pamay, Tu{\u{g}}ba and Eryi{\u{g}}it, G{\"u}l{\c{s}}en},
    booktitle={2018 Innovations in Intelligent Systems and Applications (INISTA)},
    pages={1--7},
    year={2018},
    organization={IEEE}
}

[4] @article{schuller2017marmara,
    title={Marmara {T}urkish coreference corpus and coreference resolution baseline},
    author={Sch{\"u}ller, Peter and C{\i}ng{\i}ll{\i}, K{\"u}bra and Tun{\c{c}}er, Ferit and S{\"u}rmeli, Bar{\i}{\c{s}} G{\"u}n and Pekel, Ay{\c{s}}eg{\"u}l and Karatay, Ay{\c{s}}e Hande and Karaka{\c{s}}, Hacer Ezgi},
    journal={arXiv preprint arXiv:1706.01863},
    year={2017}
}

[5] @conference{odtuCorpus,
    title={Development of a corpus and a treebank for present-day written Turkish},
    author={Say, Bilge and Zeyrek, Deniz and Oflazer, Kemal and {\"O}zge, Umut},
    booktitle={Proceedings of the 11th International Conference of Turkish Linguistics},
    pages={183--192},
    address = {Northern Cyprus},
    year={2002}
}

Changelog

2024-03-28 v1.2

issues and gaps in the coreference annotation fixed
coreference of zeros annotated
in the automatically parsed sentences, morpho-syntactic attributes updated using the turkish-imst-ud-2.12-230717 model

2023-02-24 v1.1

initial conversion
some of the paragraphs completely lack the annotation of coreference
coreference in some of the paragraphs is annotated only partially
coreference of zeros is missing
there is an overlap between sentences from Turkish-ITCC and UD_Turkish-IMST. The gold-standard morpho-syntactic annotations of these sentences were directly taken from the UD_Turkish-IMST. For the rest, this annotation was automatically obtained by running UDPipe 2 with the turkish-imst-ud-2.10-220711 model.

=== Machine-readable metadata (DO NOT REMOVE!) ================================
Data available since: CorefUD 1.1
License: CC BY-NC-SA 4.0
Includes text: yes
Genre: essay, news, article, story, novel
Lemmas: partially automatic
UPOS: partially automatic
XPOS: partially automatic
Features: partially automatic
Relations: partially automatic
CorefUD contributors: Pamay Arslan, Tuğba (1); Acar, Kutay (2); Eryiğit, Gülşen (1); Novák, Michal (3)
Contributors' affiliations: (1) Istanbul Technical University, Department of Artificial Intelligence and Data Engineering, Istanbul, Turkey
                            (2) Istanbul Technical University, Department of Computer Engineering, Istanbul, Turkey
                            (3) Charles University, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics, Prague, Czechia
Contributing: here
Contact: [email protected]
===============================================================================