ABSTRACT
User satisfaction in social question-and-answer (Q&A) systems depends on the quality of answers typically measured by a proxy metrics of user votes on the answers. We show that user votes in TurboTax AnswerXchange (AXC) can be predicted with reasonable accuracy based on the attributes of the question alone. This provides an opportunity for "pro-active" detection of potentially high or low quality content in real time while the question is still being formulated. As a result, undesirable content can be prevented by instructing the user to re-phrase the question. We can also optimize the AXC answer queue or tweak the AXC point system to generate higher quality answers.
- AnswerXchange. http://ttlc.intuit.com.Google Scholar
- Liu, Y., Bian, J., and Agichtein, E. Predicting information seeker satisfaction in community question answering. Proc. of SIGIR, (2008), 483--490. Google ScholarDigital Library
- Pal, A., Farzan, R., Konstan, J., and Kraut, R. E. Early detection of potential experts in question answering communities. User Modeling, Adaptation, and Personalization (UMAP), (2011), 231--242. Google ScholarDigital Library
- Podgorny, I. A., and Goodyear, T. Content quality and user ranking in TurboTax AnswerXchange. Proc. Of the European Conference on Social Media, (2014), 758--762.Google Scholar
- Pomerantz, J. A linguistic analysis of question taxonomies. J. Am. Soc. Inf. Sci. Technol. 56, 7, (2005), 715--728. Google ScholarDigital Library
- TurboTax® Tax Preparation Software. http://turbotax.intuit.com.Google Scholar
Index Terms
- Pro-Active Detection of Content Quality in TurboTax AnswerXchange
Recommendations
Real Time Detection and Intervention of Poorly Phrased Questions
CHI EA '15: Proceedings of the 33rd Annual ACM Conference Extended Abstracts on Human Factors in Computing SystemsAnswer helpfulness in Intuit's TurboTax AnswerXchange can be predicted with reasonable accuracy based on the attributes of the question alone. The ability to predict the likelihood of a good answer provides an opportunity to help the asker craft better ...
Self-supervised web search for any-k complete tuples
BEWEB '11: Proceedings of the 2nd International Workshop on Business intelligencE and the WEBA common task of Web users is querying structured information from Web pages. In this paper we propose a novel query processor for systematically discovering any-k relations from Web search results with conjunctive queries. The 'any-k' phrase denotes ...
Predicting the quality of answers with less bias in online health question answering communities
Highlights- We establish an automatic answer quality prediction model that combing semantic features with textual and non-textual features for online health question ...
AbstractExisting approaches in online health question answering (HQA) communities to identify the quality of answers either address it subjectively by human assessment or mainly using textual features. This process may be time-consuming and ...
Comments