In this paper, we present the re-annotation of the Turkish PUD Treebank and the first annotation of the Turkish National Corpus Universal Dependency (henceforth TNC-UD) Treebank as part of our efforts for unifying and extending the Turkish universal dependency treebanks. In accordance with the Universal Dependencies' guidelines and the necessities of Turkish grammar, both treebanks, the Turkish PUD Treebank and TNC-UD, were revised with regards to their syntactic relations. The TNC-UD is planned to have 10,000 sentences. In this paper, we present the first 500 sentences along with the re-annotation of the PUD Treebank. Moreover, this paper also offers the parsing results of a graph-based neural parser on the previous and re-annotated PUD, as well as the TNC-UD. In light of the comparisons, even though we observe a slight decrease in the attachment scores of the Turkish PUD treebank, we demonstrate that the annotation of the TNC-UD improves the parsing accuracy of Turkish. In addition to the treebanks, we have also constructed a custom annotation software with advanced filtering and morphological editing options. Both of the treebanks, including a full edit-history and the annotation guidelines, as well as the custom software are publicly available online under an open license.

No abstract

This study focuses on a comprehensive analysis and manual re-annotation of the Turkish IMST-UD Treebank, which was automatically converted from the IMST Treebank (Sulubacak et al., 2016b). In accordance with the Universal Dependencies' guidelines and the necessities of Turkish grammar, the existing treebank was revised. The current study presents the revisions that were made alongside the motivations behind the major changes. Moreover, it reports the parsing results of a transition-based dependency parser and a graph-based dependency parser obtained over the previous and updated versions of the treebank. In light of these results, we have observed that the re-annotation of the Turkish IMST-UD treebank improves performance with regards to dependency parsing.

Previous studies have shown that speakers may find sentences violating subject-verb agreement grammatical when the sentence contains a feature-matching noun phrase. This so-called agreement attraction effect has also been found in genitive possessive structures such as 'the teacher's brother' in Turkish (Lago et al., 2019), which is in contrast with its absence in similar constructions in English (Nicol et al., 2016). This discrepancy has been hypothesized to be a result of the association between genitive case marking and subjecthood in Turkish, but not in English. In the present research, we test an alternative explanation in which Turkish number agreement attraction effects are due to a potential confound in Lago et al.'s experiment, as a result of which subject head nouns were locally ambiguous between the possessive and the accusative case. We hypothesized that this ambiguity may have inhibited the availability of the head noun as an agreement controller as the accusative is a non-subject case in Turkish. To test this hypothesis, we conducted a speeded acceptability judgment experiment and our results suggest that case-ambiguity does not play a role in agreement attraction, and thus lends credibility to the claim that genitive noun phrases may function as attractors in Turkish due to the association between genitive case and subjecthood.

In this paper, we describe our contributions and efforts to develop Turkish resources, which include a new treebank (BOUN Treebank) with novel sentences, along with the guidelines we adopted and a new annotation tool we developed (BoAT). The manual annotation process we employed was shaped and implemented by a team of four linguists and five NLP specialists. Decisions regarding the annotation of the BOUN Treebank were made in line with the Universal Dependencies framework, which originated from the works of De Marneffe et al (2014) 1 Our material regarding our treebank and tool as well as our code regarding R and Python scripts are available online. The links are provided within the text.