Sign In

Communications of the ACM

91 - 100 of 3,299 for bentley

Managing Diversity in Airbnb Search

One of the long-standing questions in search systems is the role of diversity in results. From a product perspective, showing diverse results provides the user with more choice and should lead to an improved experience. However, this intuition is at odds with common machine learning approaches to ranking which directly optimize the relevance of each individual item without a holistic view of the result set. In this paper, we describe our journey in tackling the problem of diversity for Airbnb search, starting from heuristic based approaches and concluding with a novel deep learning solution that produces an embedding of the entire query context by leveraging Recurrent Neural Networks (RNNs). We hope our lessons learned will prove useful to others and motivate further research in this area.


Map Generation from Large Scale Incomplete and Inaccurate Data Labels

Accurately and globally mapping human infrastructure is an important and challenging task with applications in routing, regulation compliance monitoring, and natural disaster response management etc.. In this paper we present progress in developing an algorithmic pipeline and distributed compute system that automates the process of map creation using high resolution aerial images. Unlike previous studies, most of which use datasets that are available only in a few cities across the world, we utilizes publicly available imagery and map data, both of which cover the contiguous United States (CONUS). We approach the technical challenge of inaccurate and incomplete training data adopting state-of-the-art convolutional neural network architectures such as the U-Net and the CycleGAN to incrementally generate maps with increasingly more accurate and more complete labels of man-made infrastructure such as roads and houses. Since scaling the mapping task to CONUS calls for parallelization, we then adopted an asynchronous distributed stochastic parallel gradient descent training scheme to distribute the computational workload onto a cluster of GPUs with nearly linear speed-up.


Organizing and compressing collections of files using differences

A collection of related files often exhibits strong similarities among its constituents. These similarities, and the dual differences, may be used for both compressing the collection and for organizing it in a manner that reveals human-readable structure and relationships. This paper motivates and studies methods for such organizing and compression of file collections using inter-file differences. It presents an algorithm based on computing a minimum-weight spanning tree of a graph that has vertices corresponding to files and edges with weights corresponding to the size of the difference between the documents of its incident vertices. It describes the design and implementation of a prototype system called diboc (for difference-based organization and compression) that uses these methods to enable both compression and graphical organization and interactive exploration of a file collection. It illustrates the benefits of this system by presenting examples of its operation on a widely deployed and publicly available corpus of file collections (collections of PPD files used to configure the CUPS printing system as packaged by the Debian GNU/Linux distribution). In addition to these qualitative measures, some quantitative experimental results of applying the methods to the same corpus are also presented.


A Research of AHP Approach in Evaluating Sustainable Tourism Development

Sustainable tourism development is a key factor for tourism destination development, and evaluating the sustainable tourism development level can reflect the current situation, existing problems and deficiencies of tourism destinations. Based on constructing the evaluation index system of sustainable tourism development from four aspects, they are sustainability of tourism resources, economic feasibility, social acceptability and environmental rationality, this work applied AHP method to calculate the weight of each index, and takes Luojiang District, Deyang, Sichuan Province, China as research area. On the basis of data collection and collation, the sustainable tourism development level of Luojiang District between 2002 and 2016 was evaluated, the results show that from 2002 to 2016, the sustainable tourism development level of Luojiang District showed an overall upward trend, and can be divided into four stages: the first is the preparation stage of 2002-2011, the second is the preliminary stage of 2012-2013, the third is the basic stage of 2014-2015 and the fourth is the sustainable stage from 2016.


Bear: Cyberinfrastructure for Long-Tail Researchers at the Federal Reserve Bank of Kansas City

       The Federal Reserve Bank of Kansas City has been developing environments uniquely tailored for long-tail researchers in and around the field of economics. Due to the unique requirements of researchers in the Federal Reserve System, the Bear high performance computing environment was developed to meet our researchers’ computational and data needs, which have grown dramatically over the past few decades. In addition to increased accessibility, the environment meets an increasing number of usage modalities, demanding research cycles while decreasing the barrier of entry for researchers who have little to no experience using advanced research computing environments. We took the lessons learned from our previous environment, Bull, and improved on the pain points experienced with our diverse researcher base. We have found that Bear has improved user experience and has led to better usability while finding unique solutions to delivering resources. We have also updated and facilitated better training to give researchers in the Federal Reserve System the tools necessary to take advantage of not only Bear but also other cyberinfrastructure in the larger research ecosystem.


Alexa, How Do I Build a VUI Curriculum?

As Voice User Interfaces (VUI) become more prominent in everyday life, there is an ever growing need to ensure that designers have the appropriate knowledge and tools to be able to build usable VUIs. To do this, we need to develop a curriculum for VUI design. However, there are few resources in academia and in industry to help ground the development of a new VUI-specific HCI curriculum. We discuss the limited approaches to VUI design, and particularly to preparing designers for this emerging interaction paradigm - both academia and industry training. Grounded in this, we then describe potential avenues for understanding what a VUI curriculum may entail, what would HCI educators need from it, and how such a curriculum would be validated.


Tired of Wake Words?: Moving Towards Seamless Conversations with Intelligent Personal Assistants

In this paper, we aim to draw attention towards wake words. Wake words are an integral part of every request addressed to Intelligent Personal Assistants (IPAs). Currently, a request made to an IPA is led by wake words, making a conversation with an IPA more tiresome than a conversation with a human being. The main question we pose in this paper is, whether we can eliminate the use of wake words at least in specific contexts. Based on our experience with IPAs we propose three less burdensome alternatives that avoid the need for speaking wake words in some cases. Based on these approaches we discuss how to design seamless conversations with IPAs.


Toward Voice-Assisted Browsers: A Preliminary Study with Firefox Voice

Web browsers allow people to find, organize, and manage information on the web. While voice interaction research has evaluated the support of web search, the broader role of voice interactions within the browser have yet to be explored at depth. We report findings from a preliminary exploration of the challenges, opportunities, and directions of voice assistants embedded in modern web browsers. We drive our inquiry with Firefox Voice, a browser extension that implements a voice assistant into the Firefox desktop browser. Through a think-aloud study (n = 5), we explore the strengths and shortcomings of Firefox Voice to better understand the role that voice interaction can play in supporting people both in the browser and beyond it.


What Can I Say?: Effects of Discoverability in VUIs on Task Performance and User Experience

Discoverability, the ability for users to find and execute features through a user interface, is a recurrent problem with Voice User Interface (VUI) design that makes it difficult for users to understand what commands are supported by a newly encountered system. We studied the effects of two different discoverability strategies proposed in literature, one which provides informational prompts automatically and one which provides help only when the user requests it by asking 'What Can I Say?'. Our study adopted a Wizard of Oz approach that allowed users to order food delivery by voice. Through statistical analysis, we confirmed the beneficial nature of both strategies, with significantly better task performance and higher usability scores in comparison to a baseline. This suggests designers should consider the use of a discoverability strategy in the design of VUIs. While no significant differences were found between the strategies, a majority of the participants highlighted their preference for the 'What Can I Say?' strategy if they were to use the VUI more frequently. Finally, we reflect on the implications for the design of VUIs, highlighting the need to distinguish between initial use and longer-term use in the selection of a strategy.


Don't Believe The Hype!: White Lies of Conversational User Interface Creation Tools

Following the initial hype and high expectations of conversational user interfaces (CUIs), a number of creation tools have emerged to simplify development of these complex systems. These have the potential to democratise and expand application development to those without programming skills. However, while such tools allow end-user developers to build language understanding and dialog management capability into a CUI application, actually fulfilling or executing an action still requires programmatic API integration. In this paper, we look at how CUI builders that claim to be "no code required" struggle to yield more than toy examples, with an aim to provoke the community to develop better tools for CUI creation.