Towards understanding spontaneous speech: word accuracy vs. concept accuracy

Boros, Manuela; Eckert, Wieland; Gallwitz, Florian; Görz, Günther; Hanrieder, Gerhard; Niemann, Heinrich

doi:10.1109/icslp.1996.607774

Cited by 68 publications

(47 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This leads to the following definition of concept accuracy. A similar definition can be found, for instance, in [2]. The number of returned dependencies can be greater than the number of expected dependencies, in cases where the gold parse includes fewer secondary edges than the proposed parse.…”

Section: Discussionmentioning

confidence: 94%

“…Note that the problem not only occurs for artificial examples such as (1); here are a few mis-parsed examples actually encountered in a large parsed corpus: (2) a In this paper, we describe an alternative approach in which we employ point-wise mutual information association score in the maximum entropy disambiguation model. The association scores used here are estimated using a very large parsed corpus of 500 million words (27 million sentences), a preliminary version of the Lassy Large automatically annotated corpus of Dutch.…”

Section: Motivationmentioning

confidence: 99%

See 1 more Smart Citation

Self-Trained Bilexical Preferences to Improve Disambiguation Accuracy

Noord

2010

Text, Speech and Language Technology

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 94%

Section: Motivationmentioning

confidence: 99%

Self-Trained Bilexical Preferences to Improve Disambiguation Accuracy

Noord

2010

Text, Speech and Language Technology

View full text Add to dashboard Cite

“…The number of tasks completed (over the group) is reported in terms of task "success-ratio". Table 2 shows the successratio of the task, the average number of turns needed to complete the task, and the system's perutterance concept accuracy (Boros et al, 1996). There were no statistically significant differences between quiet room and public space, (t(38) < 2, p > 0.5, on any metric).…”

Section: End-to-end System Performancementioning

confidence: 99%

Proceedings of the EACL 2014 Workshop on Dialogue in Motion

Janarthanam¹,

Hill²,

Dickinson³

et al. 2014

View full text Add to dashboard Cite

iii Introduction Spoken dialogue systems used in call centers and car dashboards reflect years of technological development. But the smart devices that now accompany people throughout their daily activities and the extensive integration of sensors and actuators into people's environments demand new concepts in dialogue modeling and management in order to provide intuitive, proactive, personalized, context-aware, multi-modal, multi-domain dialogue systems.The past few years have seen the development of many intelligent speech-enabled virtual assistants for mobile users, such as Siri, S Voice, Google Now, SpeakToIt, Vlingo and Iris. These applications use GIS connectivity for navigation and to contextualize tasks such as search. Other multimodal applications (e.g. Wikitude, WikiHood, FieldTrip) can pro-actively present encyclopedic information about the user's surroundings, such as landmarks and points of interest, as the user walks around. Augmented reality and wearable technology such as Google Glass are presenting new opportunities for dialogue systems 'on the go'.In this proliferation of location-aware systems in the industry, together with research efforts in spatial and mobile contexts, we see a convergence of efforts (e.g. the Word2Actions workshop at NAACL 2012, the Computational Models of Spatial Language Interpretation and Generation workshop series and the Vision and Language workshop at NAACL 2013) towards what we call Dialogue In Motion: any form of interaction between a computer/robot and a human in motion -for example a pedestrian or a driver, in the real world or in a simulated environment. Natural language interactions are promoted as a more direct interaction medium, but they raise additional challenges in the context of dynamic spatial environments. This workshop focuses on these challenging issues in language processing for dialogues in motion.We received 20 submissions; all papers received three reviews from our program committee. We accepted seven papers for oral presentation and six for poster and/or demo presentation. Several of the papers are on in-car dialogue systems, which have a long track record of non-trivial implementations combining voice, GUI, haptic, and gestures with additional constraints on user's cognitive load and environment context. Others are on pedestrian navigation and virtual guides, human-robot interaction, and rapid prototyping and statistical dialogue management for dialogue in motion.We wish to thank all those who submitted papers. We also gratefully acknowledge the work of the members of our program committee. Special thanks go to Tiphaine Dalmas (University of Edinburgh) for acting as main contact for the workshop, and to Bonnie Webber (University of Edinburgh) for helpful comments along the way.We hope you enjoy the workshop! AbstractMobile Internet access via smartphones puts demands on in-car infotainment systems, as more and more drivers like to access the Internet while driving. Spoken dialog systems support the user by less distracting interaction than v...

show abstract

“…Semantic accuracy can thus be seen as the semantic analogue of sentence accuracy. In contrast, concept accuracy is based on the number of semantic units that are substituted, inserted and deleted, and can thus be seen as the semantic analogue of word accuracy (Boros et al, 1996).…”

Section: Units Of Measurementmentioning

confidence: 99%

Robust spoken language understanding in a computer game

Boye

Gustafson

Wirén

2006

Speech Communication

View full text Add to dashboard Cite

We present and evaluate a robust method for the interpretation of spoken input to a conversational computer game. The scenario of the game is that of a player interacting with embodied fairy-tale characters in a 3D world via spoken dialogue (supplemented by graphical pointing actions) to solve various problems. The player himself cannot directly perform actions in the world, but interacts with the fairy-tale characters to have them perform various tasks, and to get information about the world and the problems to solve. Hence the role of spoken dialogue as the primary means of control is obvious and natural to the player. Naturally, this means that robust spoken language understanding becomes a critical component. To this end, the paper describes a semantic representation formalism and an accompanying parsing algorithm which works off the output of the speech recogniserÕs statistical language model. The evaluation shows that the parser is robust in the sense of considerably improving on the noisy output of the speech recogniser.

show abstract

Towards understanding spontaneous speech: word accuracy vs. concept accuracy

Cited by 68 publications

References 5 publications

Self-Trained Bilexical Preferences to Improve Disambiguation Accuracy

Self-Trained Bilexical Preferences to Improve Disambiguation Accuracy

Proceedings of the EACL 2014 Workshop on Dialogue in Motion

Robust spoken language understanding in a computer game

Contact Info

Product

Resources

About