Amazon.com

About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Service (business) & Service provider. The organization has 13363 authors who have published 17317 publications receiving 266589 citations.

Disambiguation in speech recognition

Abstract: Automatic speech recognition (ASR) processing including a feedback configuration to allow for improved disambiguation between ASR hypotheses. After ASR processing of an incoming utterance where the ASR outputs an N-best list including multiple hypotheses, the multiple hypotheses are passed downstream for further processing. The downstream further processing may include natural language understanding (NLU) or other processing to determine a command result for each hypothesis. The command results are compared to determine if any hypotheses of the N-best list would yield similar command results. If so, the hypothesis(es) with similar results are removed from the N-best list so that only one hypothesis of the similar results remains in the N-best list. The remaining non-similar hypotheses are sent for disambiguation, or, if only one hypothesis remains, it is sent for execution.

Economic Predictions With Big Data: The Illusion of Sparsity

Abstract: We compare sparse and dense representations of predictive models in macroeconomics, microeconomics, and finance. To deal with a large number of possible predictors, we specify a prior that allows for both variable selection and shrinkage. The posterior distribution does not typically concentrate on a single sparse model, but on a wide set of models that often include many predictors.

Extrapolation of behavior-based associations to behavior-deficient items

Abstract: Behavior-based associations are extrapolated to items for which the quantity of collected user activity data is insufficient to create meaningful or reliable behavior-based associations (“behavior-deficient” items). The behavior-based associations are extrapolated based on content-based associations, or another type of “substitutability” association, between the behavior-deficient items and other items. The items can be any type of item for which user behaviors (e.g., purchases, accesses, downloads, etc.) can be monitored and analyzed to detect behavior-based associations, and for which item content or other available information can be used to assess item substitutability. For example, the items can be products represented in an electronic catalog, web pages or other documents accessible on a network, or web sites.

Towards Optimal Off-Policy Evaluation for Reinforcement Learning with Marginalized Importance Sampling

Abstract: Motivated by the many real-world applications of reinforcement learning (RL) that require safe-policy iterations, we consider the problem of off-policy evaluation (OPE) --- the problem of evaluating a new policy using the historical data obtained by different behavior policies --- under the model of nonstationary episodic Markov Decision Processes (MDP) with a long horizon and a large action space. Existing importance sampling (IS) methods often suffer from large variance that depends exponentially on the RL horizon $H$. To solve this problem, we consider a marginalized importance sampling (MIS) estimator that recursively estimates the state marginal distribution for the target policy at every step. MIS achieves a mean-squared error of $$ \frac{1}{n} \sum_{t=1}^H\mathbb{E}_{\mu}\left[\frac{d_t^\pi(s_t)^2}{d_t^\mu(s_t)^2} \Var_{\mu}\left[\frac{\pi_t(a_t|s_t)}{\mu_t(a_t|s_t)}\big( V_{t+1}^\pi(s_{t+1}) + r_t\big) \middle| s_t\right]\right] + \tilde{O}(n^{-1.5}) $$ where $\mu$ and $\pi$ are the logging and target policies, $d_t^{\mu}(s_t)$ and $d_t^{\pi}(s_t)$ are the marginal distribution of the state at $t$th step, $H$ is the horizon, $n$ is the sample size and $V_{t+1}^\pi$ is the value function of the MDP under $\pi$. The result matches the Cramer-Rao lower bound in [Jiang and Li, 2016] up to a multiplicative factor of $H$. To the best of our knowledge, this is the first OPE estimation error bound with a polynomial dependence on $H$. Besides theory, we show empirical superiority of our method in time-varying, partially observable, and long-horizon RL environments.