Date

Author

Derya, Cokal
Zeyrek Bozşahin, Deniz
Sanders, Ted

Metadata

Show full item record

Item Usage Stats

153
views

0
downloads

Cite This

Connectives are the prototypical linguistic markers of coherence relations in discourse. But connectives cannot express just any coherence relation. In fact, connectives can be grouped according to the type of relation expressed, such as additive, temporal, causal or contrastive relations (3; 4). In addition, recent studies show various languages sometimes have subtle distinctions within the same class of connectives. In Dutch, for instance, the causal connectives omdat and want have been shown to express the difference between typical consequence-cause and claimargument relations (5; 6; 7). Such relational differences in causality are explained in different (but related) terms, such as ‘objective vs. subjective relation’ (7) or ‘epistemic vs. content domains’ (8). In example (i) below, because connects units (a) and (b) in a claim-argument relation, which is inferred by readers (i.e., [b] provides the speaker’s subjective reason for [a]). The paraphrase of (i) would be: “The reason I thought the neighbors are not home is their lights are turned off.” In example (ii) the same connector because is used in a consequence-cause relation. Consequently, the paraphrase of (ii) is “The reason the temperature rose is the shining of the sun.” While Ex. (i) conveys an epistemic, or subjective relation, Ex. (ii) conveys real-world causality (i.e., an objective relation). (i) (a) The neighbours are not at home because (b) their lights are off. (ii) (a) The temperature rose because (b) the sun was shining. Using categorization principles, including ‘domains’ and ‘subjectivity’, causal connectives have been studied in Dutch (6, 7), French (9), German (1), and English (5). This has lead to conclusions regarding similarities (e.g., distinctions like these areclearly relevant in Dutch, German and French) and differences (In English because can express any causal relation.). While these corpus studies’ results are interesting, they have at least one serious limitation: they are all concerned with a limited set of closely related European languages. Only recently have typologically different languages like Mandarin been studied seriously. (2) demonstrated different causal connectives signal different degrees of subjectivity. In this paper, we outline our study that involves another typologically different language: Turkish. Coming from an Altaic language family, and with a rich list of causality connectives, borrowed from Persian and Arabic, our study promises to enrich the discussion of types of causality. Currently there is little knowledge about the distribution of Turkish causal connectives (i.e., çünkü, - dığı için, da and zira roughly all mean “because”). To our knowledge, they have not been described in termsrelation categorizations like domains or subjectivity. To fill this gap, we present the results of our first corpus analysis and acceptability judgment task to categorize various Turkish causal connectives. The main questions of this paper are: (1) Can Turkish causality markers used in written discourse be categorized in terms of domains or subjectivity? and (2) Does Turkish have specialized connectives, which are really specific for objective/epistemic or subjective/ content domains? Our initial logistic regression analysis from an acceptability judgment task shows that while “çünkü” and “zira” are mostly preferred when expressing a subjective relation, “-dığı için” and “da” can be used to express both subjective and objective relations. The current paper will contribute to the ongoing work on Discourse Relational Devices in three key ways: (a) Carrying out a functional categoration of Turkish causality connectives; (b) Highlighting difficulties experienced while annotating discourse relations marked by Turkish causality markers; and (c) Discussing whether in Turkish there are strong conceptual contrasts between ‘objective’ and ‘subjective’ relations or between ‘epistemic’ and ‘content’ domains that have been found in European languages.

URI

http://lpts2016.blogs.uv.es/files/2015/12/LPTS-2016_abstracts.pdf
https://hdl.handle.net/11511/84992

Collections

Graduate School of Informatics, Conference / Seminar