Cascaded Span Extraction and Response Generation for Document-Grounded Dialog
| العنوان: | Cascaded Span Extraction and Response Generation for Document-Grounded Dialog |
|---|---|
| المؤلفون: | Hermann Ney, David Thulke, Nico Daheim, Christian Dugast |
| المصدر: | Stroudsburg, PA : Association for Computational Linguistics (ACL) 57-62 (2021). doi:10.18653/v1/2021.dialdoc-1.8 The 1st Workshop on Document-grounded Dialogue and Conversational Question Answering-proceeding of the workshop : August 5, 2021, Bangkok, Thailand (online) : DialDoc 2021 The 1st Workshop on Document-grounded Dialogue and Conversational Question Answering-proceeding of the workshop : August 5, 2021, Bangkok, Thailand (online) : DialDoc 20211. Workshop on Document-grounded Dialogue and Conversational Question Answering, DialDoc 2021, online, 2021-08-05-2021-08-05 |
| بيانات النشر: | Association for Computational Linguistics, 2021. |
| سنة النشر: | 2021 |
| مصطلحات موضوعية: | FOS: Computer and information sciences, Response generation, Computer Science - Machine Learning, Computer Science - Computation and Language, Computer Science - Artificial Intelligence, business.industry, Computer science, Span (engineering), computer.software_genre, Machine Learning (cs.LG), Task (project management), Set (abstract data type), Artificial Intelligence (cs.AI), restrict, Classifier (linguistics), Artificial intelligence, Dialog box, Baseline (configuration management), business, Computation and Language (cs.CL), computer, Natural language processing |
| الوصف: | This paper summarizes our entries to both subtasks of the first DialDoc shared task which focuses on the agent response prediction task in goal-oriented document-grounded dialogs. The task is split into two subtasks: predicting a span in a document that grounds an agent turn and generating an agent response based on a dialog and grounding document. In the first subtask, we restrict the set of valid spans to the ones defined in the dataset, use a biaffine classifier to model spans, and finally use an ensemble of different models. For the second subtask, we use a cascaded model which grounds the response prediction on the predicted span instead of the full document. With these approaches, we obtain significant improvements in both subtasks compared to the baseline. Accepted by 1st DialDoc Workshop at ACL-IJCNLP 2021 |
| DOI: | 10.18653/v1/2021.dialdoc-1.8 |
| URL الوصول: | https://explore.openaire.eu/search/publication?articleId=doi_dedup___::2c36efdedc0ee05e78cc631b1a70155a https://doi.org/10.18653/v1/2021.dialdoc-1.8 |
| Rights: | OPEN |
| رقم الانضمام: | edsair.doi.dedup.....2c36efdedc0ee05e78cc631b1a70155a |
| قاعدة البيانات: | OpenAIRE |
| DOI: | 10.18653/v1/2021.dialdoc-1.8 |
|---|