Online Supervised Learning of Non-Understanding Recovery Policies Export

@inproceedings{citeulike:1369087, abstract = {Spoken dialog systems typically use a limited number of nonunderstanding recovery strategies and simple heuristic policies1 to engage them (e.g. first ask user to repeat, then give help, then transfer to an operator). We propose a supervised, online method for learning a non-understanding recovery policy over a large set of recovery strategies. The approach consists of two steps: first, we construct runtime estimates for the likelihood of success of each recovery strategy, and then we use these estimates to construct a policy. An experiment with a publicly available spoken dialog system shows that the learned policy produced a 12.5\% relative improvement in the non-understanding recovery rate.}, author = {Bohus, Dan and Langner, Brian and Raux, Antoine and Black, Alan W. and Eskenazi, Maxine and Rudnicky, Alex}, booktitle = {Proceedings of SLT}, citeulike-article-id = {1369087}, citeulike-linkout-0 = {http://www.cs.cmu.edu/~dbohus/docs/SLTPaper.pdf}, keywords = {error, error\_recovery, speech, sui}, posted-at = {2007-06-06 23:11:28}, priority = {4}, title = {Online Supervised Learning of Non-Understanding Recovery Policies}, url = {http://www.cs.cmu.edu/~dbohus/docs/SLTPaper.pdf}, year = {2006} }

TY - CONF ID - citeulike:1369087 L3 - citeulike-article-id:1369087 N2 - Spoken dialog systems typically use a limited number of nonunderstanding recovery strategies and simple heuristic policies1 to engage them (e.g. first ask user to repeat, then give help, then transfer to an operator). We propose a supervised, online method for learning a non-understanding recovery policy over a large set of recovery strategies. The approach consists of two steps: first, we construct runtime estimates for the likelihood of success of each recovery strategy, and then we use these estimates to construct a policy. An experiment with a publicly available spoken dialog system shows that the learned policy produced a 12.5% relative improvement in the non-understanding recovery rate. TI - Online Supervised Learning of Non-Understanding Recovery Policies T2 - Proceedings of SLT KW - error KW - error_recovery KW - speech KW - sui AU - Bohus, Dan AU - Langner, Brian AU - Raux, Antoine AU - Black, Alan AU - Eskenazi, Maxine AU - Rudnicky, Alex PY - 2006/// UR - http://www.cs.cmu.edu/~dbohus/docs/SLTPaper.pdf ER -