What's the trouble

Abstract

Spoken dialogue systems promise efficient and natural access to information services from any phone. Recently, spoken dialogue systems for widely used applications such as email, travel information, and customer care have moved from research labs into commercial use. These applications can receive millions of calls a month. This huge amount of spoken dialogue data has led to a need for fully automatic methods for selecting a subset of caller dialogues that are most likely to be useful for further system improvement, to be stored, transcribed and further analyzed. This paper reports results on automatically training a Problematic Dialogue Identifier to classify problematic human-computer dialogues using a corpus of 1242 DARPA Communicator dialogues in the travel planning domain. We show that using fully automatic features we can identify classes of problematic dialogues with accuracies from 67% to 89%.

Interaction Quality: Assessing the quality of ongoing spoken dialog interaction by experts—And how it relates to user satisfaction

“…Here, a SDS could react 90 by adapting the dialog strategy when critical situations are 91 automatically detected, cf. Langkilde et al (1999), ?, 92 Walker et al (2002), Hastie et al (2002), Levin and 93 Pieraccini (2006), Herm et al (2008), Schmitt et al 94 (2008), Zgorzelski et al (2010), Schmitt et al (2010c), 95 Ultes et al (2011Ultes et al ( , 2012Ultes et al ( , 2014a. Another option would 96 be the escalation to a live agent who finishes the task jointly 97 with the user.…”

The effect of contour type and epistemic modality on the assessment of speaker certainty

“…Downstepped contours are quite common in Standard American English. For example, in the AT&T Communicator Corpus of read speech [2,5], the H* !H* L-contours represent the most frequent pattern of the 2888 intermediate phrases in this corpus, comprising about 40% of all contours. They occur almost twice as often as the standard declarative contours ((H*) H* L-; illustrated in Figure 2) in this corpus [4].…”

“…Moreover, they claimed that the interpretation of sequences of downstepped pitch accents might be characterized as conveying that the Hearer should be able to infer, from the shared beliefs of Hearer and Speaker, the existence of discourse entities realized with such accents. A possibly related observation is that downstep serves as an alternative to deaccenting, when information being expressed represents given information in the discourse [2,11,12]. Thus, to the extent that given information represents information about which the speaker is certain, we can hypothesize that the speaker's use of downstep will convey an epistemic disposition of certainty.…”