To enable the interactive exploration of large social media datasets we exploit the temporal distributions of word n-grams within the message stream to discover “interesting” concepts, determine “relatedness” between concepts, and find representative examples for display. We present a new algorithm for context-dependent “interestingness” using the coefficient of variation of the temporal distribution, apply the well-known technique of Pearson's Correlation to tweets using equi-height histogramming to determine correlation, and employ an asymmetric variant for computing “relatedness” to encourage exploration. We further introduce techniques using interestingness, correlation, and relatedness to automatically discover concepts and select preferred word N-grams for display. These techniques are demonstrated on an 800,000 tweet dataset from the Academy Awards.

Even though scripting languages like Pig allow for simpler coding, performing analytics over Big Data using Map-Reduce engines remains challenging. To further assist developers, and support novice users, we offer "The Farm", a catalog of scriptable services supporting creation, discovery, composition, and optimized execution. Each Pig script added to The Farm becomes an executable service, with inputs and outputs defined by relation schemas. Those services are discoverable using natural language search, and composable using a drag-and-drop interface. To support efficient execution, composed services are automatically merged to a single executable script, which can then be run by a growing selection of platform-specific optimizers and interpreters.

Supporting fast access to large RDF stores has been one of key challenges for enabling use of the Semantic Web in real-life applications, more so in sensor-based systems where large amounts of historic data needs to be stored. We propose a semantics-based temporal view mechanism that enables faster access to time-varying data by caching into memory only the required subset of RDF triples. We describe our experience of implementing such a framework in the context of a wide area network monitoring system. Our preliminary results show that our solution significantly improves client access time and scales well for moderate data sets.

The unification of Semantic Web query languages under the SPARQL standard and the development of commercial-quality implementations are encouraging industries to use semantic technologies for managing information. Current implementations, however, lack the performance monitoring and management services that the industry expects. In this paper, we present a performance and management framework interface to a generic SPARQL web server. We leverage existing standards for instrumentation to make the system ready-to-manage through existing monitoring applications, and we provide a performance framework which has the distinct feature of providing measurement results through the same SPARQL interface used to query data, eliminating the need for special interfaces.

Asset management in state of the art data centers is still a manual process. An automated system to track the location of servers and create a real-time inventory would not only improve operations management but also reduce operational costs. This demo shows a novel application of location aware wireless networks, RFID technology and visualization software integrated in end-to-end system to track servers in a data center.

When performing teleoperation in a delayed environment it is recognized that providing the operator with a simulation of the remote environment will improve performance by insulating him or her from the delay. However, while this simulation provides the primary display, the provision of real visual images from the remote site is still desirable to assist the operator in diagnosing and recovering from unexpected events. This paper considers a teleoperation environment in which the delay is too great to permit direct operator control of remote cameras while the bandwidth is too small to permit direct transmission of remote site imagery. The system proposed uses models of the environment, task action and remote cameras to predict that fragment of the available imagery would best assist the operator were an error to occur. The system has been used successfully for performing real world intervention in cases where the operator and remote site were physically separated by 500 km, the communications time delay was in excess of 10 seconds and available bandwidths were software-limited to be on the order of 10 kbits/second.

The objective of this work is to allow subsea teleoperative tasks to be performed efficiently even when all communication between human operator and remote robot is via an acoustic link. This paper describes the integration of remote site cameras into the teleprogramming system. Images from these cameras may be used for calibrating and maintaining the world model as well as for aiding the operator in diagnosing and recovering from unexpected events. The constrained communications channel forces the adoption of an intelligent approach to image transmission. The system automatically selects windowing and subsampling parameters while varying frame rates and task execution speeds to match that which the system can support given available bandwidth and computational constraints. Examples are presented from tests in which an operator in Philadelphia controlled a robot manipulator mounted on an underwater vehicle submerged off the Massachusetts coast.

The teleprogramming system employs a supervisory control approach where operator interaction with a virtual environment directs the actions of a remote manipulator operating semi-autonomously. We focus on the detection, diagnosis, and recovery from unexpected situations frequently arising during normal operation. The teleprogramming system employs a single thread approach to error recovery that is inherent in the system design. We perform two experimental tasks, each highlighting a different role for the operator, using the teleprogramming system with a simulated communication time delay of 10 seconds. In the first task of inserting and extracting bolts, the operator assumes the role of "task level reasoning" or task planning and sequencing, task state reasoning, and diagnostic procedure generation. In the second task, puncturing and slicing duct tape, the operator performs "action level reasoning" or diagnosing and correcting unexpected situations and determining success of actions. Our experimental results indicate that complex tasks can be accomplished when the operator is provided multiple and overlapping forms of feedback.>