%{search_type} search results

Computational approaches to inference and information extraction often assume that adjective-noun compounds maintain all the relevant properties of the unmodified noun. A significant portion of nonsubsective adjectives violate this assumption. We present preliminary work towards a classifier for these adjectives. We also compile a comprehensive list of 60 nonsubsective adjectives including those used for training and those found by the classifiers.

We describe Waldo, a scripting language for application networking. Waldo allows programmers to describe complex network interactions between many hosts as transactional operations with atomicity, consistency, and isolation. Waldo is able to provide these transactional semantics starvation-free without assuming global clocks, without centralized scheduling, and under arbitrary transaction conflicts. This allows programmers to write application networking as short transactional scripts that will never starve. Waldo achieves these results with a novel distributed transaction scheduling algorithm that combines the wound-wait algorithm and Lamport clocks with two transaction priority levels. Experimental results show that using the primary algorithm Waldo can perform up to 10,000 transctions per second between two endpoints connected across the wide area network.

We describe TOKENSREGEX, a framework for defining cascaded regular expressions over token sequences.TOKENSREGEX is available as part of the Stanford CoreNLP software package and can be used for various tasks which require reasoning over tokenized text. It has been used to build SUTIME, a state-of-the-art temporal tagger, and can be helpful in a variety of scenarios such as named entity recognition (NER) and information extraction from tokens.

Large-scale acquisition of exterior urban environments is by now a well-established technology, supporting many applications in search, navigation, and commerce. The same is not true for indoorenvironments, however: access is often restricted and the spaces may be cluttered. In addition, such environments typically contain a high density of repeated objects (e.g., tables, chairs, monitors, etc.) in regular or non-regular arrangements with significant pose variations and articulations. In this paper, we exploit the special structure of indoor environments to accelerate their 3D acquisition and recognition with a low-end handheld scanner. Our approach runs in two phases: (i) a learning phase, where we acquire 3D models of frequently occurring objects and capture their variability modes from only a few scans, and (ii) a recognition phase, where from a single scan of new areas, we identify previously seen objects, but in varying poses and locations. This greatly accelerates the capture process (average recognition time of 200ms/model). We demonstrate our framework with the acquisition of typical areas of a university building including cubicle or desk areas, auditoriums, etc., using a Microsoft Kinect sensor.

Researchers have long envisioned a Semantic Web, where unstructured Web content is replaced by documents with rich semantic annotations. Unfortunately, this vision has been hampered by the difficulty of acquiring semantic metadata for Web pages. This paper introduces a method for automatically"semantifying" structural page elements: using machine learning to train classifiers that can be applied in a post-hoc fashion. We focus on one popular class of semantic identifiers: those concerned with the structure—or information architecture—of a page. To determine the set of structural semanticsto learn and to collect training data for the learning, we gather a large corpus of labeled page elements from a set of online workers. We discuss the results from this collection and demonstrate that our classifiers learn structural semantics in a general way

The Web today provides a corpus of design examples unparalleled in human history. However, leveraging existing designs to produce new pages is currently difficult. This paper introduces the Bricolage algorithm for automatically transferring design and content between Web pages. Bricolage introduces a novel structuredprediction technique that learns to create coherent mappings between pages by training on human-generated exemplars. The produced mappings can then be used to automatically transfer the content from one page into the style and layout of another. We show that Bricolage can learn to accurately reproduce human page mappings, and that it provides a general, efficient, and automatic technique for retargeting content between a variety of real Web pages.

Instructive example code is a central part of programming. Web search enables programmers to quickly locate relevant examples. However, existing code editors offer little support for helping users interactively explore examples. This paper proposes that effective use of examples hinges on theprogrammer's ability to quickly identify a small number of relevant lines interleaved among a larger body of boilerplate code. This insight is manifest in Rehearse, a code editing environment with two unique features: First, Rehearse links program execution to source code by highlighting each line of code as it is executed. This enables programmers to quickly determine which lines of code are involved in producing a particular interaction. Second, after a programmer has found a single line applicable to her task, Rehearseautomatically identifies other lines that are also likely to be relevant. In a controlled experiment, participants using visualization and highlighting adapted example code significantly faster than those using an identical editor without these features.

Virtual worlds seek to provide an online setting where users can interact in a shared environment. Popular virtual worlds such as Second Life and World of Warcraft, however, rely on share-nothing data and strict partitioning as much as possible. They translate a large world into many tiny worlds. This partitioning conflicts with the intended goal of a virtual world by greatly limiting interaction and reducing the shared experience. We present Meru, an architecture for scalable, federated virtual worlds. Meru’s key insight is that, compared to traditional distributed object systems, virtual world objects have the additional property of being embedded in a three-dimensional geometry. By leveraging this geometric information in messaging and caching, Meru can allow uncongested virtual world objects to pass messages with 800 times the throughput as Second Life while also gracefully scaling to handle the congestion of ten thousand active senders. Unlike virtual worlds today, Meru achieves this performance without any partitioning, maintaining a single, seamless world.

The ready availability of online source code examples has changed the cost structure of programming by example modification. However, current search tools are wholly separate from editing tools. What benefits might be realized by integrating them? This paper describes the design, implementation,and evaluation of Blueprint, a tool that integrates Web search into the Adobe Flex Builder development environment. Blueprint automatically augments queries with code context, presents an example-centric view of search results, and retains a link between copied code and its source. This paper introducesa technique for retrieving relevant example code, descriptions, and running examples for a user’s query. A between-subjects study found that Blueprint enables participants to search for and select example code significantly faster than with a standard Web browser.

This paper investigates the role of online resources in problem solving. We look specifically at how programmers—an exemplar form of knowledge workers—opportunistically interleave Web foraging, learning, and writing code. We describe two studies of how programmers use online resources.The first study, conducted in the lab, found that programmers leverage the Web with three distinct intentions: They engage in just-in-time learning of new skills and approaches, they extend their skills, and they strategically delegate their memory to online resources. The results also suggest that queries for different purposes have different styles and durations. Do query styles robustly vary with intent, or is thisresult an artifact of the particular lab setting? To address this question, we analyzed a month-long set of Web queries to a commercial programming framework’s online information sources. In this dataset, query style also corresponded to intent. These results contribute to a theory of online resource usage in programming, and suggest opportunities for tools to facilitate opportunistic programming.

We utilize a high-dimensional parametric design space to support a novel and intuitive method for 3D modeling. Users visually explore the design space and pick models using a continuous, map-like interface.We leverage models created by the user community to learn underlying structure in the space via kernel density estimation. This mapping of the space is maintained by a server that synchronizes all the deployed design tools. The tools leverage the mapping to allow users with no prior modeling experience to easily create unique designs by interpolating between and extrapolating from landmark models. The result is a self-reinforcing design system that becomes easier to use as more people participate.Our prototype tree modeling tool was downloaded by over six thousand users from more than eighty countries in the month following its release. Over fifteen hundred trees were voluntarily picked from the roughly hundred dimensional tree space. We report on usage patterns gathered through this deployment and on subsequent user surveys.

We present a new symbolic execution tool, KLEE, capable of automatically generating tests that achievehigh coverage on a diverse set of complex and environmentally-intensive programs. We applied KLEEto all 90 programs in the GNU COREUTILS utility suite, which form the core user-level environment installed on almost all Unix systems and, as such, represent some of the most heavily used and tested open-source programs in existence. For 84% of these utilities, KLEE’s automatically generated tests covered 80–100% of executable statements and, in aggregate, significantly beat the coverage of the developers’ own hand-written test suites. KLEE also found nine serious bugs (including three that had been missed for over 15 years!) and produced concrete inputs that triggered the errors when runon the uninstrumented code. When applied to MINIX’s versions of a small selection of the same applications, KLEE achieved similar coverage (along with two bugs). In addition, we also used KLEE to automatically find numerous incorrect differences between several MINIX and COREUTILS tools. Finally, we checked the kernel of the HISTAR operating system, generating tests that achieved 76.4% (without paging enabled) and 67.1% coverage (with paging) and found one important security bug

Analogy plays an important cognitive role in reasoning and problem solving. One illustration of analogical cognition can be found in design practice, where viewing examples is an established technique for inspiration and learning. While digital information technologies have made it easier for designers to access examples of other designers’ work, significant opportunities exist for selecting and presenting examples in a proactive fashion. In this paper, we introduce techniques for dynamically deriving interfaces for examplebased design tools using decision-theoretic selection, designer specification, and end-user preference as inputs. This paper describes a manifestation of these techniques in theAdaptive Ideas web page builder, an HTML-based display platform for web page designers that leverages content metadata to automatically generate displays of examples. We present an evaluation of these techniques through a first-use study.