Abstract-Deep neural networks (DNNs) are now a central component of nearly all state-of-the-art speech recognition systems. Building neural network acoustic models requires several design decisions including network architecture, size, and training loss function. This paper offers an empirical investigation on which aspects of DNN acoustic model design are most important for speech recognition system performance. We report DNN classifier performance and final speech recognizer word error rates, and compare DNNs using several metrics to quantify factors influencing differences in task performance. Our first set of experiments use the standard Switchboard benchmark corpus, which contains approximately 300 hours of conversational telephone speech. We compare standard DNNs to convolutional networks, and present the first experiments using locally-connected, untied neural networks for acoustic modeling. We additionally build systems on a corpus of 2,100 hours of training data by combining the Switchboard and Fisher corpora. This larger corpus allows us to more thoroughly examine performance of large DNN models -with up to ten times more parameters than those typically used in speech recognition systems. Our results suggest that a relatively simple DNN architecture and optimization technique produces strong results. These findings, along with previous work, help establish a set of best practices for building DNN hybrid speech recognition systems with maximum likelihood training. Our experiments in DNN optimization additionally serve as a case study for training DNNs with discriminative loss functions for speech tasks, as well as DNN classifiers more generally.

We present an approach to speech recognition that uses only a neural network to map acoustic input to characters, a character-level language model, and a beam search decoding procedure. This approach eliminates much of the complex infrastructure of modern speech recognition systems, making it possible to directly train a speech recognizer using errors generated by spoken language understanding tasks. The system naturally handles out of vocabulary words and spoken word fragments. We demonstrate our approach using the challenging Switchboard telephone conversation transcription task, achieving a word error rate competitive with existing baseline systems. To our knowledge, this is the first entirely neural-network-based system to achieve strong speech transcription results on a conversational speech task. We analyze qualitative differences between transcriptions produced by our lexicon-free approach and transcriptions produced by a standard speech recognition system. Finally, we evaluate the impact of large context neural network character language models as compared to standard n-gram models within our framework.

We present PROCAB, an efficient method for Probabilistically Reasoning from Observed Context-Aware Behavior. It models the context-dependent utilities and underlying reasons that people take different actions. The model generalizes to unseen situations and scales to incorporate rich contextual information. We train our model using the route preferences of 25 taxi drivers demonstrated in over 100,000 miles of collected data, and demonstrate the performance of our model by inferring: (1) decision at next intersection, (2) route to known destination, and (3) destination given partially traveled route.

Human emotional states are not independent but rather proceed along systematic paths governed by both internal, cognitive factors and external, social ones. For example, anxiety often transitions to disappointment, which is likely to sink to depression before rising to happiness and relaxation, and these states are conditioned by the states of others in our communities. Modeling these complex dependencies can yield insights into human emotion and support more powerful sentiment technologies.We develop a theory of conditional dependencies between emotional states in which emotions are characterized not only by valence (polarity) and arousal (intensity) but also by the role they play in state transitions and social relationships. We implement this theory using conditional random fields (CRFs) that synthesize textual information with information about previous emotional states and the emotional states of others. To assess the power of affective transitions, we evaluate our model in a collection of 'mood' updates from the Experience Project. To assess the power of social factors, we use a corpus of product reviews from a website in which the community dynamics encourage reviewers to be influenced by each other. In both settings, our models yield improvements of statistical and practical significance over ones that classify each text independently of its emotional or social context.

Coursera has partnered with more than 100 top universities and institutions, among them Princeton, Caltech, Johns Hopkins, University of Pennsylvania, and others, to offer free courses online for anyone. Since announcing our initial partnerships in April 2012, Coursera has enrolled more than 5 million students across every country in the world.