Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics |
| Erscheinungsort | Melbourne |
| Seiten | 241-251 |
| Seitenumfang | 11 |
| Publikationsstatus | Veröffentlicht - Juli 2018 |
| Extern publiziert | Ja |
| Veranstaltung | 56th Annual Meeting of the Association for Computational Linguistics - Melbourne, Australien Dauer: 15 Juli 2018 → 20 Juli 2018 |
Abstract
Given any argument on any controversial topic, how to counter it? This question implies the challenging retrieval task of finding the best counterargument. Since prior knowledge of a topic cannot be expected in general, we hypothesize the best counterargument to invoke the same aspects as the argument while having the opposite stance. To operationalize our hypothesis, we simultaneously model the similarity and dissimilarity of pairs of arguments, based on the words and embeddings of the arguments' premises and conclusions. A salient property of our model is its independence from the topic at hand, i.e., it applies to arbitrary arguments. We evaluate different model variations on millions of argument pairs derived from the web portal idebate.org. Systematic ranking experiments suggest that our hypothesis is true for many arguments: For 7.6 candidates with opposing stance on average, we rank the best counterargument highest with 60% accuracy. Even among all 2801 test set pairs as candidates, we still find the best one about every third time.
ASJC Scopus Sachgebiete
- Informatik (insg.)
- Software
- Informatik (insg.)
- Theoretische Informatik und Mathematik
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics. Melbourne, 2018. S. 241-251.
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - Retrieval of the Best Counterargument without Prior Topic Knowledge
AU - Wachsmuth, Henning
AU - Syed, Shahbaz
AU - Stein, Benno
PY - 2018/7
Y1 - 2018/7
N2 - Given any argument on any controversial topic, how to counter it? This question implies the challenging retrieval task of finding the best counterargument. Since prior knowledge of a topic cannot be expected in general, we hypothesize the best counterargument to invoke the same aspects as the argument while having the opposite stance. To operationalize our hypothesis, we simultaneously model the similarity and dissimilarity of pairs of arguments, based on the words and embeddings of the arguments' premises and conclusions. A salient property of our model is its independence from the topic at hand, i.e., it applies to arbitrary arguments. We evaluate different model variations on millions of argument pairs derived from the web portal idebate.org. Systematic ranking experiments suggest that our hypothesis is true for many arguments: For 7.6 candidates with opposing stance on average, we rank the best counterargument highest with 60% accuracy. Even among all 2801 test set pairs as candidates, we still find the best one about every third time.
AB - Given any argument on any controversial topic, how to counter it? This question implies the challenging retrieval task of finding the best counterargument. Since prior knowledge of a topic cannot be expected in general, we hypothesize the best counterargument to invoke the same aspects as the argument while having the opposite stance. To operationalize our hypothesis, we simultaneously model the similarity and dissimilarity of pairs of arguments, based on the words and embeddings of the arguments' premises and conclusions. A salient property of our model is its independence from the topic at hand, i.e., it applies to arbitrary arguments. We evaluate different model variations on millions of argument pairs derived from the web portal idebate.org. Systematic ranking experiments suggest that our hypothesis is true for many arguments: For 7.6 candidates with opposing stance on average, we rank the best counterargument highest with 60% accuracy. Even among all 2801 test set pairs as candidates, we still find the best one about every third time.
UR - http://www.scopus.com/inward/record.url?scp=85063094915&partnerID=8YFLogxK
U2 - 10.18653/v1/p18-1023
DO - 10.18653/v1/p18-1023
M3 - Conference contribution
AN - SCOPUS:85063094915
SN - 9781948087322
SP - 241
EP - 251
BT - Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics
CY - Melbourne
T2 - 56th Annual Meeting of the Association for Computational Linguistics, ACL 2018
Y2 - 15 July 2018 through 20 July 2018
ER -