Deep Learning & Dialog Systems 2014

Deep Learning


Janus Recognition Toolkit (JRTk) has been used by the Interactive System Lab in many projects for speech recognition, such as BABEL.  The Babel Program (IARPA) is developing agile and robust speech recognition technology that can be rapidly applied to any human language in order to provide effective search capability for analysts to efficiently process massive amounts of real-world recorded speech.

See also:

100 Best Deep Learning Videos | 100 Best GitHub: Deep LearningSkipgram & Deep Learning 2014

A historical perspective of speech recognition X Huang, J Baker, R Reddy – Communications of the ACM, 2014 – … However, many algorithmic improvements have been made, such as how to use distributed algorithms for the deep learning task. … 42. Williams, J. and Young, S. Partially observable Markov decision processes for spoken dialog systems. … Cited by 12 Related articles

Word-based dialog state tracking with recurrent neural networks M Henderson, B Thomson… – 15th Annual Meeting of …, 2014 – … The challenge is based on a large corpus collected using a variety of telephone- based dialog systems in the domain of finding a restaurant in … underlying representations of the input, has been found effective as an initialisation technique in deep learning (Vincent et al., 2008). … Cited by 4 Related articles All 9 versions

Feature enhancement by deep LSTM networks for ASR in reverberant multisource environments F Weninger, J Geiger, M Wöllmer, B Schuller… – Computer Speech & …, 2014 – Elsevier … feature-space de-reverberation. There, we also consider deep learning with pre-training, where the first layers are trained to de-noising and subsequent layer(s) are trained to perform de-reverberation. While the CHiME WSJ … Cited by 9 Related articles All 4 versions

Application of deep belief networks for natural language understanding R Sarikaya, GE Hinton, A Deoras – IEEE/ACM Transactions on Audio, …, 2014 – … 1481–1488. [13] D. Erhan, Y. Bengio, A. Courville, P. Manzagol, and P. Vincent, “Why Does Unsupervised Pre-training Help Deep Learning?,” J. Mach. … Ruhi Sarikaya is a principal scientist and the manager of language understanding and dialog systems group at Microsoft. … Cited by 5 Related articles All 7 versions

Deep learning of knowledge graph embeddings for semantic parsing of twitter dialogs L Heck, H Huang – IEEE Global Conference on Signal and …, 2014 – … Index Terms—deep learning, semantic parsing, Twitter, dialog … Rep. MSR-TR-2014-70, 2014. [16] L. Wang, L. Heck, and D. Hakkani-Tür, “Leveraging semantic web search and browse sessions for multi-turn spoken dialog systems,” in Proceedings of the International … Cited by 1 Related articles All 7 versions

Knowledge Acquisition Strategies for Goal-Oriented Dialog Systems A Pappu, AI Rudnicky – 15th Annual Meeting of the Special …, 2014 – … system, education models, human learning System Predicted Researcher-Interests 4 florian metze dialogue systems, speech recogni- tion … dy- namics System Predicted Researcher-Interests 6 alexander hauptmann discriminatively trained models, deep learning, computer vision … Cited by 2 Related articles All 7 versions

Deriving local relational surface forms from dependency-based entity embeddings for unsupervised spoken language understanding YN Chen, D Hakkani-Tür, G Tur – Proceedings of SLT, 2014 – … surface forms are learned from dependency-based entity embeddings, which encode the contexts of entities from dependency trees in a deep learning model. … Traditional spo- ken dialogue systems (SDS) are trained with annotated examples and support limited domains. … Cited by 2

Dynamically supporting unexplored domains in conversational interactions by enriching semantics with neural word embeddings YN Chen, AI Rudnicky – Proceedings of SLT, 2014 – … Spoken dialogue systems (SDS) are appearing on smart-phones and allow users to launch applications via spontaneous speech. … With the recent advance of deep learning techniques, the continuous-valued word embeddings have further boosted the state-of-the-art results in … Cited by 2

Learning Situated Knowledge Bases through Dialog A Pappu, AI Rudnicky – Fifteenth Annual Conference of …, 2014 – … The study took place in the context of EventSpeak Dialog System that informs people about upcoming talks/events of their interest and ongo- ing work of other researchers on a university campus. … search, learning to rank 13. crowdsourcing, deep learning 14. … Cited by 3 Related articles All 5 versions

Eye gaze for spoken language understanding in multi-modal conversational interactions D Hakkani-Tür, M Slaney, A Celikyilmaz… – Proceedings of the 16th …, 2014 – … [5] L. Deng, J. Li, J.-T. Huang, K. Yao, D. Yu, F. Seide, M. Seltzer, G. Zweig, X. He, and J. Williams. Recent advances in deep learning for speech research at microsoft. … [9] T. Misu, A. Raux, I. Lane, J. Devassy, and R. Gupta. Situated multi-modal dialog system in vehicles. … Cited by 3 Related articles All 4 versions

Improving deep neural network acoustic modeling for audio corpus indexing under the IARPA Babel program X Cui, B Kingsbury, J Cui… – submitted to …, 2014 – … M. Heldner, “An instantaneous vec- tor representation of delta pitch for speaker-change prediction in conversational dialogue systems,” ICASSP 2008 … GE Hinton, “Vocal tract length perturbation (VTLP) improves speech recognition, in ICML Workshop on Deep Learning for Audio … Cited by 1 Related articles All 5 versions

A generalized rule based tracker for dialogue state tracking K Sun, L Chen, S Zhu, K Yu – submitted to IEEE SLT 2014, 2014 – … [1] Jason D Williams, “Challenges and opportunities for state tracking in statistical spoken dialog systems: Re- sults … Huang, Kaisheng Yao, Dong Yu, Frank Seide, Michael Seltzer, Geoff Zweig, Xi- aodong He, Jason Williams, et al., “Recent advances in deep learning for speech … Cited by 1 Related articles

Deep Neural Networks For Spoken Dialog Systems C MAIN – 2014 – … 21 2.3 Deep Learning for State Tracking in Spoken Dialogue Systems . . . . . 23 … 2.3 Deep Learning for State Tracking in Spoken Dialogue Systems State tracking in SDSs is a complex task with issues that it needs to deal with such as noisy conditions and ambiguity. … Related articles

Leveraging Frame Semantics and Distributional Semantics for Unsupervised Semantic Slot Induction in Spoken Dialogue Systems YN Chen, WY Wang, AI Rudnicky – … success that implements this distributional theory is Latent Semantic Analy- sis [3]. Recently, with the advance of deep learning techniques, the … frame-semantic parsing to automatically induce and adapt the se- mantic ontology for designing spoken dialogue systems (SDS) in an … Cited by 4 Related articles All 3 versions

Evaluating a Spoken Dialogue System that Detects and Adapts to User Affective States D Litman, K Forbes-Riley – 15th Annual Meeting of the Special Interest …, 2014 – … is and isn’t effective at promoting deep learning. In Intelligent Tutoring Systems Conference, pages 245– 254, Pittsburgh, PA, USA. K. Forbes-Riley and D. Litman. 2009. A user modeling-based performance analysis of a wizarded uncertainty-adaptive dialogue system corpus. … Related articles All 6 versions

Deep Generative and Discriminative Models for Speech Recognition L Deng – … Page 25. Li Deng, Dong Yu (Microsoft Research) Geoffrey Hinton (University of Toronto) Deep Learning for Speech Recognition and Related Applications … Page 53. Outline • Advances in deep learning for features/ representations • Advances in deep learning for models/ … Related articles

FIONA: a platform for embodied cognitive agents C Alvarez, L Fernández Cossío – Proceedings of the second …, 2014 – … Author Keywords FIONA, artificial intelligence, virtual robots, virtual agents, artificial vision, cognitive architectures, dialog systems, speech recognition system, deep learning, neural networks ACM Classification Keywords D.1.7 Programming Techniques: Visual Programming …

Optimizing Generative Dialog State Tracker via Cascading Gradient Descent BJ Lee, W Lim, D Kim, KE Kim – 15th Annual Meeting of the …, 2014 – … 1 Introduction Spoken dialog systems, a field rapidly growing with the spread of smart mobile devices, has to deal with challenges to … The model used in this paper can naturally devel- oped by adding hidden layers, and ultimately deep learning techniques could be applicable. … All 8 versions

AutoTutor and Family: A Review of 17 Years of Natural Language Tutoring BD Nye, AC Graesser, X Hu – … Journal of Artificial Intelligence in Education, 2014 – Springer … Using a domain-independent reactive planner to implement a medical dialogue system. In T. Bickmore (Ed.), AAAI fall symposium on systems for health communication (pp. 24–31). … AutoTutor improves deep learning of computer literacy: Is it the dialog or the talking head? … Related articles

Joint Semantic Utterance Classification and Slot Filling with Recursive Neural Networks DZ Guo, G Tur, W Yih, G Zweig – … Performing intent determination and slot filling together in one model simplifies a dialog system, since only one model needs to be … and A Ng, “Grounded com- positional semantics for finding and describing images with sentences,” in NIPS Deep Learning Workshop, 2013. … Related articles All 6 versions

Personal Knowledge Graph Population From User Utterances In Conversational Understanding X Li, G Tur, DHTQ Li – … Intent: Get Price Slots: good: gas; cost relative: cheapest; location: (lat,long) Typically, spoken dialog queries to a dialog system may be … various classification methods such as Boosting [11, 12, 13], support vector machines (SVMs) [14], and more recently deep learning [15, 16]. …

Noisy speech recognition using blind spatial subtraction array technique and deep bottleneck features N Kitaoka, T Hayashi, K Takeda – Asia-Pacific Signal and …, 2014 – … the field of pattern recognition research, and Deep Neural Networks (DNN) which have been trained using deep learning have achieved … Saruwatari, K. Shikano, “Real-time implementation of blind spatial subtraction array for hands-free robot spoken dialogue system,” IEEE/RSJ …

From here to AGI: A roadmap to the realization of human-level artificial general intelligence B Goertzel, G Yu – Neural Networks (IJCNN), 2014 International …, 2014 – … world Animated Agent Carrying out of simple “scientific” experi- mentation in game world Mobile Robot • Building structures from blocks and other simple objects • Supplementation of third-party speech-to-text with deep learning based speech-to-text Dialogue System • Ability to … Related articles

Creating the Intelligent Novice: Supporting Self-Regulated Learning and Metacognition in Educational Technology B Goldberg, R Spain – Design Recommendations for Intelligent …, 2014 – … The goal is for an ITS to facilitate deep learning of a topic while instilling behaviors that support future SRL opportunities. Identifying strategies to promote metacognitive aware- ness requires an understanding of the processes and phases theoretically linked to SRL approaches. … Related articles All 6 versions

Intelligent Systems’ Holistic Evolving Analysis of Real-Life Universal Speaker Characteristics B Schuller, Y Zhang, F Eyben, F Weninger – … spite them being crucial for real-life applications such as retrieval, dialogue systems and computer-mediated human- to-human conversation … For real- ising these ambitious goals, deep learning (Hinton et al., 2012) combined with neuroevolutionary methods and non- parametric … Cited by 1 Related articles All 2 versions

Acoustic-Prosodic Entrainment and Rapport in Collaborative Learning Dialogues N Lubold, H Pon-Barry – Proceedings of the 2014 ACM workshop on …, 2014 – … can serve as a guide for teachers when students are engaged in collaborative activity. In tutorial dialogue systems, de- tecting rapport has implications for improving dialogue suc- cess and quality. … Strong community, deep learning: Exploring the link. … Related articles All 3 versions

Adaptive Tutoring for Self-Regulated Learning: A Tutorial on Tutoring Systems RA Sottilare, AM Sinatra – … training. The ultimate goal for ITSs is to impact learning with effect sizes equivalent to raising average (“C”) students to experts (“A” students) through tailored instruction and reinforcement of deep learning principles. In other …

Language Learning via Unsupervised Corpus Analysis B Goertzel, C Pennachin, N Geisweiller – Engineering General Intelligence …, 2014 – Springer … We consider the approach described here as “deep learning” oriented because it is based on hierarchical pattern recognition in linguistic data: identifying patterns, then patterns among these patterns, etc., in a hierarchy that allows “higher level” (more abstract) patterns to feed …

Serious Games in Education Towards the standardization of the teaching-learning process N HAMDAOUI, MK IDRISSI… – ADVANCES in …, 2014 – … to which the learner must be able to customize his/her learning experience to his/her own learning style and be able to try new styles at the same time; the ‘identity principle’which stands for the fact that deep learning calls for an … User Models in Dialog Systems, 1989: p. 35-51. … Related articles All 3 versions

Extraction of Salient Sentences from Labelled Documents M Denil, A Demiraj, N de Freitas – arXiv preprint arXiv:1412.6815, 2014 – … & Blunsom, 2013a; Devlin et al., 2014; Cho et al., 2014; Bahdanau et al., 2014; Sutskever et al., 2014; Hermann & Blunsom, 2014), question answering (Bordes et al., 2014; Weston et al., 2014), dialogue systems (Kalchbrenner & … Deep learning for answer sentence selection. … Related articles All 3 versions

When Two Heads are Better Than One: A Critical Review of Four Collaborative Intelligent Tutoring Systems R Harsley – … work in CSCL that shows collaboration has a positive effect on individual and group learning gains and promotes deep learning. … Basilica is a software architecture and toolkit intended to support the extension of traditional, individual tutoring dialog systems to accommodate … Related articles

Gaze-enhanced speech recognition M Slaney, R Rajan, A Stolcke… – Acoustics, Speech and …, 2014 – … This suggests that dialog systems in- volving selection steps would benefit from gaze. Finally, we expect that users’ confusion and user-interface failure will be all to obvious from the eye-gaze information. … Recent advances in deep learning for speech re- search at Microsoft. … Cited by 4 Related articles All 8 versions

Providing emotion awareness and affective feedback to virtualised collaborative learning scenarios M Feidakis, S Caballé, T Daradoumis… – International Journal of …, 2014 – Inderscience Page 1. Int. J. Cont. Engineering Education and Life-Long Learning, Vol. 24, No. 2, 2014 141 Copyright © 2014 Inderscience Enterprises Ltd. Providing emotion awareness and affective feedback to virtualised collaborative learning scenarios Michalis Feidakis* … Related articles All 2 versions

Speaker adaptation of deep neural network based on discriminant codes S Xue, O Abdel-Hamid, H Jiang, L Dai, Q Liu – 2014 – …,, Abstract—Fast adaptation of deep neural networks (DNN) is an important research topic in deep learning. In this paper, we have proposed a general adaptation scheme … Cited by 2 Related articles

An artificial neural network approach to automatic speech processing SM Siniscalchi, T Svendsen, CH Lee – Neurocomputing, 2014 – Elsevier … data [27], [28], [29], [30] and [31]. These advances triggered interest in developing acoustic models based on DNNs and other deep learning techniques for ASR, eg, [32] and [33]. We used DNNs to boost classification accuracy … Cited by 3 Related articles All 4 versions

Structural information aware deep semi-supervised recurrent neural network for sentiment analysis W Rong, B Peng, Y Ouyang, C Li, Z Xiong – Frontiers of Computer Science – Springer … Experimental studies have been conducted on commonly used datasets and the results have shown its promising potential. Keywords sentiment analysis, recurrent neural network, deep learning, machine learning 1 Introduction … Related articles

A Regression Approach to Speech Enhancement Based on Deep Neural Networks Y Xu, J Du, LR Dai, CH Lee – … Each layer is pre-trained without supervision to learn a high level representation of its input (or the output of its previous layer). For the regression task, deep learning has been used in several speech synthesis tasks [20, 21]. … Cited by 1 Related articles

An attribute detection based approach to automatic speech processing SM Siniscalchi, CH Lee – Loquens, 2014 – … This number, often referred to as a CM, serves as a reference guide for the dialogue system to provide an appropriate response to its users, just as an intelligent human being is expected to do when interacting with other people. … Related articles

Speak Correct: Phonetic Editor Approach H Al-Barhamtoshy, K Jambi, W Al-Jedaibi… – Life Science …, 2014 – Page 1. Life Science Journal 2014;11(8) 626 Speak Correct: Phonetic Editor Approach Hassanin Al-Barhamtoshy1, Kamal Jambi1, Wajdi Al-Jedaibi1, Diaa Motaweh2, Sherif Abdou3, Mohsen Rashwan4 … Related articles

Support in a Framework for Instructional Technology PJ Durlach – Design Recommendations for Intelligent Tutoring … – … Aleven, V., Ogan, A., Popescu, O., Torrey, C. & Koedinger, KR (2004). Evaluating the effectiveness of a tutorial dialogue system for self-explanation. … A time for emoting: When affect-sensitivity is and isn’t effective at promoting deep learning. … Related articles All 5 versions

Informal workplace learning in Austrian banks: the influence of learning approach, leadership style, and organizational learning culture on managers’ learning … D Froehlich, M Segers… – Human Resource …, 2014 – Wiley Online Library … (1997) find deep learning to result … Marsick and Watkins (2003) define organizational learning culture to comprise seven interlinked dimensions: opportunities for learning, dialogue, systems thinking, collaborative learning, knowledge management systems, empowerment, and … Related articles All 5 versions

Unsupervised Induction of Semantic Roles within a Reconstruction-Error Minimization Framework I Titov, E Khoddam – arXiv preprint arXiv:1412.2812, 2014 – … been shown to benefit question answering [47, 29], textual entailment [46], machine translation [57, 36, 56, 21], and dialogue systems [5, 53 … error, for example, the Euclidean distance ?(x, ˜x) = ||x ? ˜x||2. Though currently popular only within the deep learning community, latent … Related articles All 3 versions

A Guide to Instructional Techniques, Strategies and Tactics to Manage Learner Affect, Engagement, and Grit RA Sottilare, JA DeFalco, J Connor – … for Intelligent Tutoring …, 2014 – … Page 42. Design Recommendations for Intelligent Tutoring Systems-Volume 2: Instructional Management historical. Instructional strategies may be tied to a near-term objective (eg, acquire knowledge) or a long-term objective (eg, promote deep learning to enhance retention). … Related articles All 7 versions

DCLA Meet CIDA: Collective Intelligence Deliberation Analytics SB Shum, A De Liddo, M Klein – Page 1. DCLA14 discussion paper: Buckingham Shum, De Liddo & Klein 1 Working Paper for discussion — 2nd International Workshop on Discourse-Centric Learning Analytics, LAK14: 4th International Conference on Learning … Cited by 1 Related articles

Convolutional neural networks for speech recognition O Abdel-Hamid, AR Mohamed, H Jiang… – IEEE/ACM Transactions …, 2014 – … In retrospect, the performance improve- ments of these recent attempts have been ascribed to their use of “deep” learning, a reference both to the number of hidden layers in the neural network as well as to the abstractness and, by some accounts, psychological plausibility of … Cited by 5 Related articles All 7 versions

Fast adaptation of deep neural network based on discriminant codes for speech recognition S Xue, O Abdel-Hamid, H Jiang, L Dai… – IEEE/ACM Transactions on …, 2014 – … Discriminant Codes for Speech Recognition Shaofei Xue, Ossama Abdel-Hamid, Hui Jiang, Lirong Dai, and Qingfeng Liu Abstract—Fast adaptation of deep neural networks (DNN) is an important research topic in deep learning. … Cited by 2 Related articles

Neural network language models for off-line handwriting recognition F Zamora-Martínez, V Frinken, S España-Boquera… – Pattern Recognition, 2014 – Elsevier Unconstrained off-line continuous handwritten text recognition is a very challenging task which has been recently addressed by different promising techniques. T. Cited by 4 Related articles All 4 versions

Artificial Conversations for Chatter Bots Using Knowledge Representation, Learning, and Pragmatics C Chakrabarti – 2014 – … 22 2.2 The Syntactic Approach . . . . . 26 2.3 The Semantic Approach . . . . . 27 2.4 Dialogue systems . . . . . 27 viii Page 9. Contents 2.5 Limitations of existing approaches . . . . . 29 … Related articles All 7 versions

Triggering effective social support for online groups R Kumar, CP Rosé – … Transactions on Interactive Intelligent Systems (TiiS …, 2014 – … In line with recent developments in data-driven approaches for building dialog systems, we present a novel technique for learning behavior- specific triggering policies, deploying it as part of our efforts to improve a socially capable conversational tutor agent that supports … Cited by 2 Related articles

Vocabulary development and intervention for English learners in the early grades. DL Baker, S Al Otaiba, M Ortiz, V Correa… – ADVANCES IN CHILD …, 2014 – … instruction. Deep learning requires inte- gration of prior knowledge with new information and results in the ability to use this information constructively in new contexts (the formation of a sit- uation model, Kintsch, 1998). Second … Related articles All 3 versions

A Complete Bibliography of ACM Transactions on Asian Language Information Processing NHF Beebe – 2014 – … Page 14. REFERENCES 14 Kim:2003:RRE [37] Harksoo Kim and Jungyun Seo. Resolution of referring expressions in a Korean multimodal dialogue system. ACM Transactions on Asian Lan- guage Information Processing, 2(4):324–337, December 2003. CODEN ???? … Related articles All 9 versions

[BOOK] Engineering General Intelligence, Part 1 B Goertzel, C Pennachin, N Geisweiller – 2014 – Springer Page 1. Engineering General Intelligence, Part 2 The CogPrime Architecture for Integrative, Embodied AGI Ben Goertzel Cassio Pennachin Nil Geisweiller Atlantis Thinking Machines Series Editor:K. – U. Kühnberger Page 2. Atlantis Thinking Machines Volume 6 Series editor … Cited by 9 Related articles All 9 versions

Latent semantic rational kernels for topic spotting on conversational speech C Weng, DL Thomson, P Haffner… – IEEE/ACM Transactions on …, 2014 – Page 1. 1738 IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 22, NO. 12, DECEMBER 2014 Latent Semantic Rational Kernels for Topic Spotting on Conversational Speech Chao Weng … Related articles All 2 versions

Foundations and Trends in Signal Processing L Deng, Y Dong – Signal Processing, 2014 – Page 1. the essence of knowledge FnT SIG 7:3-4 Deep Learning; Methods and Applications Li Deng and Dong Y u Foundations and Trends® in Signal Processing 7:3-4 Deep Learning Methods and Applications Li Deng and Dong Yu now now Page 2. 7.1. … Cited by 1 Related articles All 8 versions

Comparing Four Instructional Techniques for Promoting Robust Knowledge JE Richey, TJ Nokes-Malach – Educational Psychology Review, 2014 – Springer Page 1. REVIEW ARTICLE Comparing Four Instructional Techniques for Promoting Robust Knowledge J. Elizabeth Richey & Timothy J. Nokes-Malach © Springer Science+Business Media New York 2014 Abstract Robust knowledge … Related articles

Human Translation Evaluation and its Coverage by Automatic Scores M Vela, AK Schumann, A Wurm – Automatic and Manual Metrics for … – Page 25. 19 Full Papers Page 26. 20 Human Translation Evaluation and its Coverage by Automatic Scores Mihaela Vela, Anne-Kathrin Schumann, Andrea Wurm Department of Applied Linguistics, Translation and Interpreting … Related articles All 3 versions

Combining visual recognition and computational linguistics: linguistic knowledge for visual recognition and natural language descriptions of visual content M Rohrbach – 2014 – Page 1. Combining Visual Recognition and Computational Linguistics Linguistic Knowledge for Visual Recognition and Natural Language Descriptions of Visual Content Thesis for obtaining the title of Doctor of Engineering Science (Dr.-Ing.) … Related articles All 4 versions