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.

Litter production and litter nutrient concentrations in a fragmented amazonian landscape

Abstract: We analyzed the effects of distance to forest edge and soil texture on fine-litter production and on nutrient concentrations in the leaf fall in an experimentally fragmented landscape in Brazilian Amazonia. Production of fine litter (leaves, twigs 250 m) from forest edges, and in clayey or sandy soils. In total, 28 plots were established, with 10 litter traps per plot. Results reveal a significant effect of distance to forest edge on litter production, but no significant effect of soil type or interaction between soil type and edge distance. On average, annual litter production on edge plots exceeded that on the interior plots by 0.68 Mg/ha (9.50 ± 0.23 vs. 8.82 ± 0.14 Mg·ha−1·yr−1, mean ± 1 se, based on a 3-yr period). With regard to nutrient concentrations in the leaf fall, we detected a significant effect of soil type on three of eight nutrients analyzed. Concentrations of N, Mg, and Mn were greater in leaves on clayey than on sandy soils. Distance to forest edge only significantly affected the concentration of Ca, which was greater near than far from edges, perhaps due to strong Ca mobilization by the roots of pioneer trees. Several factors may account for the observed increase in litterfall near forest edges, including the greater prevalence of winds, increased plant desiccation stress, and higher rates of tree recruitment, especially of pioneer trees, near edges. Elevated rates of litterfall are likely to have cascading effects on the ecology of fragmented forests, affecting the invertebrate fauna, increasing seed and seedling mortality, and causing forest fragments to be more vulnerable to destructive surface fires.

Semantic Product Search

Abstract: We study the problem of semantic matching in product search, that is, given a customer query, retrieve all semantically related products from the catalog. Pure lexical matching via an inverted index falls short in this respect due to several factors: a) lack of understanding of hypernyms, synonyms, and antonyms, b) fragility to morphological variants (e.g. "woman" vs. "women"), and c) sensitivity to spelling errors. To address these issues, we train a deep learning model for semantic matching using customer behavior data. Much of the recent work on large-scale semantic search using deep learning focuses on ranking for web search. In contrast, semantic matching for product search presents several novel challenges, which we elucidate in this paper. We address these challenges by a) developing a new loss function that has an inbuilt threshold to differentiate between random negative examples, impressed but not purchased examples, and positive examples (purchased items), b) using average pooling in conjunction with n-grams to capture short-range linguistic patterns, c) using hashing to handle out of vocabulary tokens, and d) using a model parallel training architecture to scale across 8 GPUs. We present compelling offline results that demonstrate at least 4.7% improvement in Recall@100 and 14.5% improvement in mean average precision (MAP) over baseline state-of-the-art semantic search methods using the same tokenization method. Moreover, we present results and discuss learnings from online A/B tests which demonstrate the efficacy of our method.

An Efficient Bandit Algorithm for Realtime Multivariate Optimization

Abstract: Optimization is commonly employed to determine the content of web pages, such as to maximize conversions on landing pages or click-through rates on search engine result pages. Often the layout of these pages can be decoupled into several separate decisions. For example, the composition of a landing page may involve deciding which image to show, which wording to use, what color background to display, etc. Such optimization is a combinatorial problem over an exponentially large decision space. Randomized experiments do not scale well to this setting, and therefore, in practice, one is typically limited to optimizing a single aspect of a web page at a time. This represents a missed opportunity in both the speed of experimentation and the exploitation of possible interactions between layout decisions Here we focus on multivariate optimization of interactive web pages. We formulate an approach where the possible interactions between different components of the page are modeled explicitly. We apply bandit methodology to explore the layout space efficiently and use hill-climbing to select optimal content in realtime. Our algorithm also extends to contextualization and personalization of layout selection. Simulation results show the suitability of our approach to large decision spaces with strong interactions between content. We further apply our algorithm to optimize a message that promotes adoption of an Amazon service. After only a single week of online optimization, we saw a 21% conversion increase compared to the median layout. Our technique is currently being deployed to optimize content across several locations at Amazon.com.

Namespace management in distributed storage systems

Abstract: A directed acyclic graph (DAG) is generated to represent a namespace of a directory. In response to a request to create a new object with a specified name, a hash value bit sequence is computed for the name. A plurality of levels of the DAG are navigated using successive subsequences of the bit sequence to identify a candidate node for storing a new entry corresponding to the specified name. If the candidate node meets a split criterion, the new entry and at least a selected subset of entries of the candidate node's list of entries are distributed among a plurality of DAG nodes, including at least one new DAG node, using respective bit sequences obtained by applying the hash function for each distributed entry.

Image-based document display

Abstract: The present invention is directed to a method and system for continuously displaying image pages of digital content which are available over a network. More specifically, the method and system enables a user to view image pages in a continuous manner while a limited number of image pages are being downloaded at a given time. Several image pages which are adjacent to the image page(s) the user is currently viewing may be stored in temporary memory. The image pages in the temporary memory are utilized so that, within the image pages, the user can move the displayed image pages up and down without experiencing any discontinuation. In order to ensure continuous display throughout the entire digital content, the next possible set of image pages is constantly determined and obtained to update the current set of image pages in the temporary memory.