Towards automatically generating Questions under Discussion to link information and discourse structure
3854Authors
nan
Abstract
AbstractQuestions under Discussion (QUD; Roberts, 2012) are emerging as a conceptually fruitful approach to spelling out the connection between the information structure of a sentence and the nature of the discourse in which the sentence can function. To make this approach useful for analyzing authentic data, Riester, Brunetti & De Kuthy (2018) presented a discourse annotation framework based on explicit pragmatic principles for determining a QUD for every assertion in a text. De Kuthy et al. (2018) demonstrate that this supports more reliable discourse structure annotation, and Ziai and Meurers (2018) show that based on explicit questions, automatic focus annotation becomes feasible. But both approaches are based on manually specified questions. In this paper, we present an automatic question generation approach to partially automate QUD annotation by generating all potentially relevant questions for a given sentence. While transformation rules can concisely capture the typical question formation process, a rule-based approach is not sufficiently robust for authentic data. We therefore employ the transformation rules to generate a large set of sentence-question-answer triples and train a neural question generation model on them to obtain both systematic question type coverage and robustness.
Keywords
Full Content
Proceedings of the 28th International Conference on Computational Linguistics, pages 5786–5798 Barcelona, Spain (Online), December 8-13, 2020 5786 Towards automatically generating Questions under Discussion to link information and discourse structure Kordula De Kuthy Madeeswaran Kannan Haemanth Santhi Ponnusamy Detmar Meurers SFB 833, Project A4, University of T¨ubingen, Germany {kdk,mkannan,hsp,dm}@sfs.uni-tuebingen.de Abstract Questions under Discussion (QUD; Roberts, 2012) are emerging as a conceptually fruitful ap- proach to spelling out the connection between the information structure of a sentence and the nature of the discourse in which the sentence can function. To make this approach useful for analyzing authentic data, Riester, Brunetti & De Kuthy (2018) presented a discourse annotation framework based on explicit pragmatic principles for determining a QUD for every assertion in a text. De Kuthy et al. (2018) demonstrate that this supports more reliable discourse structure annotation, and Ziai and Meurers (2018) show that based on explicit questions, automatic focus annotation becomes feasible. But both approaches are based on manually specified questions. In this paper, we present an automatic question generation approach to partially automate QUD annotation by generating all potentially relevant questions for a given sentence. While transfor- mation rules can concisely capture the typical question formation process, a rule-based approach is not sufficiently robust for authentic data. We therefore employ the transformation rules to gen- erate a large set of sentence-question-answer triples and train a neural question generation model on them to obtain both systematic question type coverage and robustness. 1 Introduction As the attention in linguistics is shifting from the analysis of isolated sentences to how information is encoded in discourse, the information structure of sentences and how it is expressed in a given language is receiving increasing interest. In order to connect the information structure of sentences to the overall structure of the discourse, so-called Questions under Discussion (QUD) (Van Kuppevelt, 1995; Roberts, 2012; Beaver and Clark, 2009; Velleman and Beaver, 2016) have emerged as a way to spell out the hinge between the properties of the sentence and the nature of the discourse in which the sentence can function. QUDs make explicit, how the meaning expressed by a sentence fits into the functional structure of the evolving discourse. Example (1) illustrates this with the implicit QUD Q2 connecting assertion A2 to the previous discourse A1. Concretely, the sentence A2 answers the question Q2 given discourse A1. (1) A1 You were working until last summer for the NSA Q2 What did he do during his time at the NSA? A2 and during this time you [secretly collected thousands of confidential documents.]F That part of the sentence that directly answers the QUD, here the VP secretly collected thousands of con- fidential documents, is marked as the so-called Focus (Rooth, 1992; Krifka and Musan, 2012), indicating how the alternatives implicitly present in the discourse are narrowed down as the discourse progresses. QUDs can thus be seen as a way of making alternatives explicit in the discourse. This intuitive idea is also mentioned in corpus-based research attempting to analyze …
Content truncated to 500 words
Labeled Text
AbstractQuestions under Discussion (QUD; Roberts, 2012) are emerging as a conceptually fruitful approach to spelling out the connection between the information structure of a sentence and the nature of the discourse in which the sentence can function. To make this approach useful for analyzing authentic data, Riester, Brunetti & De Kuthy (2018) presented a discourse annotation framework based on explicit pragmatic principles for determining a QUD for every assertion in a text. De Kuthy et al. (2018) demonstrate that this supports more reliable discourse structure annotation, and Ziai and Meurers (2018) show that based on explicit questions, automatic focus annotation becomes feasible. But both approaches are based on manually specified questions. In this paper, we present an automatic question generation approach to partially automate QUD annotation by generating all potentially relevant questions for a given sentence. While transformation rules can concisely capture the typical question formation process, a rule-based approach is not sufficiently robust for authentic data. We therefore employ the transformation rules to generate a large set of sentence-question-answer triples and train a neural question generation model on them to obtain both systematic question type coverage and robustness.
Publication Details
- Paper ID:
- 3854
- Year:
- 2020
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan