Search

Commonsense knowledge is a foundational cornerstone of artificial intelligence applications. Whereas information extraction and knowledge base construction for instance-oriented assertions, such as Brad Pitt's birth date, or Angelina Jolie's movie awards, has received much attention, commonsense knowledge on general concepts (politicians, bicycles, printers) and activities (eating pizza, fixing printers) has only been tackled recently. In this tutorial we present state-of-the-art methodologies towards the compilation and consolidation of such commonsense knowledge (CSK). We cover text-extraction-based, multi-modal and Transformer-based techniques, with special focus on the issues of web search and ranking, as of relevance to the WSDM community.

The novel coronavirus (Covid-19) has spread rapidly amongst countries all around the globe. Compared to the rise in cases, there are few Covid-19 testing kits available. Due to the lack of testing kits for the public, it is useful to implement an automated AI-based E-health decision support system as a potential alternative method for Covid-19 detection. As per medical examinations, the symptoms of Covid-19 could be somewhat analogous to those of pneumonia, though certainly not identical. Considering the enormous number of cases of Covid-19 and pneumonia, and the complexity of the related images stored, the data pertaining to this problem of automated detection constitutes big data. With rapid advancements in medical imaging, the development of intelligent predictive and diagnostic tools have also increased at a rapid rate. Data mining and machine learning techniques are widely accepted to aid medical diagnosis. In this paper, a huge data set of X-ray images from patients with common bacterial pneumonia, confirmed Covid-19 disease, and normal healthy cases are utilized for AI-based decision support in detecting the Coronavirus disease. The transfer learning approach, which enables us to learn from a smaller set of samples in a problem and transfer the discovered knowledge to a larger data set, is employed in this study. We consider transfer learning using three different models that are pre-trained on several images from the ImageNet source. The models deployed here are VGG16, VGG19, and ResNet101. The dataset is generated by gathering different classes of images. We present our approach and preliminary evaluation results in this paper. We also discuss applications and open issues.

The advancements in the field of deep learning and emotion recognition has been increasing in the recent past. The work presents a model framework that understands the emotion depicted on the face and from the voice. The primary goal of this work remains to improve human-computer cooperation. Hood frontal face images and various voice cuts are provided by the model system. From the Image database FER2013, 25838 samples were used for training and 90 samples of Amrita Emote database (ADB) for testing. The speech database consists of four different datasets, with a total of 20,000 examples. 3/4 of this information is used to prepare and 1/4 of the information used is for testing. Intermittent neural networks (RNNs) and traditional neural networks (CNNs) are nervous system-based projects that use speech and image management to control emotions: pleasure, sadness, anger, hatred, surprise, and fear.

As big data heads towards big knowledge, data management and machine learning techniques work together to address several interesting problems. In this paper, we address a problem in natural language processing that involves learning by mining from large text databases. More specifically, we deal with the problem of preposition prediction, especially for ESL (English as a second language) learners. Prepositions are function words that typically show a relationship between a noun or a pronoun and other elements of a sentence. They play a key role in determining the meaning of a sentence. Accurate prediction of correct prepositions in a sentence is a challenging job since preposition usage is one of the most subtle aspects of the English grammar, making it difficult for non-native speakers. This paper proposes an approach for preposition prediction called WordPrep based on which we build a tool. WordPrep relies on mining based on the words themselves rather than on their lexical or syntactic connotations. This addresses the challenges of prepositions appearing in idiomatic phrases or in different semantic contexts, due to which the actual words are better than their grammatical positions. Our proposed solution entails a direct data-driven approach to predict the missing preposition in a sentence by learning from matching tokens consisting of ngrams with words before and after the preposition. Using various searches and pattern-matching methods against a large number of database records from big text corpora, this approach predicts the missing preposition(s). We describe our pilot approach, tool implementation and experiments in this paper. This work is particularly helpful for pedagogical applications.

There is growing conviction that the future of computing depends on our ability to exploit big data on theWeb to enhance intelligent systems. This includes encyclopedic knowledge for factual details, common sense for human-like reasoning and natural language generation for smarter communication. With recent chatbots conceivably at the verge of passing the Turing Test, there are calls for more common sense oriented alternatives, e.g., the Winograd Schema Challenge. The Aristo QA system demonstrates the lack of common sense in current systems in answering fourth-grade science exam questions. On the language generation front, despite the progress in deep learning, current models are easily confused by subtle distinctions that may require linguistic common sense, e.g.quick food vs. fast food. These issues bear on tasks such as machine translation and should be addressed using common sense acquired from text. Mining common sense from massive amounts of data and applying it in intelligent systems, in several respects, appears to be the next frontier in computing. Our brief overview of the state of Commonsense Knowledge (CSK) in Machine Intelligence provides insights into CSK acquisition, CSK in natural language, applications of CSK and discussion of open issues. This paper provides a report of a tutorial at a recent conference with a brief survey of topics.

Load flow analysis is the initial and essential step for any power system computation. It is required for choosing better options for power system expansion to meet with ever increasing load demand. Implementation of Flexible AC Transmission System (FACTS) device like STATCOM, in the load flow, which is having fast and very flexible control, is one of the important tasks for power system researchers. This paper presents a simple and systematic approach for steady state power flow calculations with FACTS controller, static synchronous compensator (STATCOM) using command line usage of MATLAB tool-power system analysis toolbox (PSAT). The complexity of MATLAB language programming increases due to incorporation of STATCOM in an existing Newton–Raphson load flow algorithm. Thus, the main contribution of this paper is to show how command line usage of user friendly MATLAB tool, PSAT, can extensively be used for quicker and wider interpretation of the results of load flow with STATCOM. The novelty of this paper lies in the method of applying the load increase pattern, where the active and reactive loads have been changed simultaneously at all the load buses under consideration for creating stressed conditions for load flow analysis with STATCOM. The performance have been evaluated on many standard IEEE test systems and the results for standard IEEE-30 bus system, IEEE-57 bus system, and IEEE-118 bus system are presented.

Mobile devices have become an integral part of our daily lives. Most people carry smartphones today almost everywhere; and have other mobile devices such as tablets, often more convenient than full-fledged laptops for work transit, short trips etc. This had led to development of apps for mobile devices, easy to download and access anywhere anytime. An important field improving human experiences on mobile devices is machine learning. This constitutes technqiues involving acquisition of knowledge, skills and understanding by machines from examples, guidance, experience or reflection to learn analogous to humans. Among learning paradigms herein, supervised learning comprises situations where labeled training samples are provided to administer the process, making it more regulated, similar to human instructors providing such examples with notions of correctness to guide human learners. Supervised learning techniques are useful in designing mobile apps as they entail guided examples capturing specific human needs and their reasoning in activities, e.g., classification. This paper gives a comprehensive review of a few useful supervised learning approaches along with their implementation in mobile apps, focusing on Androids as they constitute over 50% of the global smartphone market. It includes description of the approaches and portrays interesting Android apps deploying them, addressing classification and regression problems. We discuss the contributions and critiques of the apps and also present open issues with the potential for further research in related areas. This paper is expected to be useful to students, researchers and developers in mobile computing, human computer interaction, data mining and machine learning.

Can we automatically predict failures of an object detection model on images from a target domain? We characterize errors of a state-of-the-art object detection model on the currently popular smart mobility domain, and find that a large number of errors can be identified using spatial commonsense. We propose øurmodel , a system that automatically identifies a large number of such errors based on commonsense knowledge. Our system does not require any new annotations and can still find object detection errors with high accuracy (more than 80% when measured by humans). This work lays the foundation to answer exciting research questions on domain adaptation including the ability to automatically create adversarial datasets for target domain.

The digital universe is expanding at a very fast pace generating massive datasets. In order to keep up with the processing and storage needs for this big data, and to discover knowledge, we need scalable infrastructure and technologies that can access data from multiple disks simultaneously. Cloud computing provides paradigms for data analytics over such huge datasets. While SQL continues to be popular among database and data mining professionals, in recent years Hive has established itself as a rapidly advancing technology for big data which makes it highly suitable for use over the cloud. In this paper, we present investigatory research on Hive and SQL with a detailed case study between them, considering cloud data management and mining. Our work here constitutes a thorough scrutiny, focusing on processing Hive queries on cloud infrastructure considering three different approaches, and also delving into SQL processing on the cloud with similar approaches. Real datasets are used for conducting various operations using Hive and SQL. This paper conducts performance comparisons of the two technologies and explains the environment in which one is preferred over the other for processing and analyzing data. It provides recommendations for cloud data analytics, based on the case study.

In this pandemic situation, everyone are vulnerable to COVID-19. So the necessity for safety measures checking system has become inevitable. Every measure is currently being checked separately using man power, crowded places are more susceptible for the spread of disease. Using infrared sensor and setting up a threshold crowd management is accomplished. The proposed system delivers a non contact solution by integrating sensors in order to monitor the human body temperature and environmental conditions remotely, automatically and quickly. The system completely avoids direct contact with the residents during the process of screening. Along with temperature sensing, proper sanitization is also critical. As a result, the proposed work also employs a pump motor to spray sanitization automatically and without the need for human intervention. Many companies and organizations must adjust to and protect an infected person by spotting someone who does not wear a masked face;however, this is not always possible. The proposed system also investigates the performance of detecting people wearing a face mask in a real-time situation using Machine Learning. The results show the highest precision from all input images and a camera.

In secure group communication for applications such as pay-per-view, teleconferencing, and video conferencing, key management method with logarithmic computation is expected along with storage and bandwidth efficiency. Major aspects with key management are formation of group key with initial group members and updating the group key with any change in number of group members. Rekeying is mandatory due to group join/leave operations or periodic rekeying. Logical Key Hierarchy has logarithmic rekeying cost. This paper explores Logical Key Hierarchy (LKH), LKH variants with centralized and distributed approach are analyzed and presented. This paper explores their way of achieving and persisting logarithmic cost for secure group communication, to decide the suitable method for a current application.

This article presents a decision support system to provide green or energy efficient solutions for data centers that maintain computers and peripherals to serve organizations. Traditionally, data centers catered to all operations using in-house servers. Cloud technology provides alternatives to outsource operations heading towards greenness. However, using cloud services for all data center operations may have its pitfalls. In this paper, the authors analyze various data center parameters such as carbon footprint and power usage effectiveness along with cloud-based and server-based models. They consider data mining techniques of decision trees and case based reasoning in their work. Among other findings, they head towards a hybrid model that meets the demands of productivity, energy efficiency and related factors. These findings lead to the development of the decision support system. The authors describe the research, development and evaluation of the system. They conclude with important outcomes deployed in real-world scenarios in data center management.

Much of the English in text documents today comes from nonnative speakers. Web searches are also conducted very often by non-native speakers. Though highly qualified in their respective fields, these speakers could potentially make errors in collocation, e.g., "dark money" and "stock agora" (instead of the more appropriate English expressions "black money" and "stock market" respectively). These may arise due to literal translation from the respective speaker's native language or other factors. Such errors could cause problems in contexts such as querying over Web pages, correct understanding of text documents and more. This paper proposes a framework called CollOrder to detect such collocation errors and suggest correctly ordered collocated responses for improving the semantics. This framework integrates machine learning approaches with natural language processing techniques, proposing suitable heuristics to provide responses to collocation errors, ranked in the order of correctness. We discuss the proposed framework with algorithms and experimental evaluation in this paper. We claim that it would be useful in semantically enhancing Web querying e.g., financial news, online shopping etc. It would also help in providing automated error correction in machine translated documents and offering assistance to people using ESL tools.

As databases head towards data streams, discovering knowledge from the data poses challenges. The need to process and mine data is affected by the big data wave, advances in Web technology and other factors. The advent of the cloud with related technologies provides further momentum to enhance knowledge discovery. Data mining is also of interest to research communities outside computer science as there is a need to harvest data from various domains incorporating domainspecific factors. These and other issues motivate PhD students to pursue core and applied research in data mining along with stream data management, big data, cloud computing, Web knowledge discovery, domainspecific techniques and more. A PhD forum on data mining provides an excellent platform for doctoral students to present imminent research and get valuable feedback from experts. In addition it gives them the opportunity to disseminate the results of their work among fellow researchers, and publish their novel contributions at an early stage. IEEE ICDM hosts such a PhD forum for doctoral students with a data mining focus. This article describes the content of the work presented at the ICDM 2013 PhD forum. It also gives a brief overview of the organization of this forum. The article thus provides a panorama of recent doctoral student work in data mining that would be of interest to researchers in database management and related areas.

Data mining, the discovery of knowledge from data, bridges several disciplines such as database management, artificial intelligence, statistics, visualization and the domain of the data, e.g., biology or engineering. Knowledge discovered by mining the data can be used for various purposes such as developing decision support systems and intelligent tutors. In this paper we present such a data mining problem in the mechanical engineering domain where knowledge discovery from the data is performed using statistical approaches, to conduct predictive analysis for decision support. More specifically, we focus on the engine health problem which consists of using existing data on the behavior of an engine in order to predict whether the engine is capable of functioning well (i.e., it is healthy) and to offer suggestions on preventive maintenance. The data we use for this predictive analysis consists of graphs that plot process parameters such as the vibration and temperature of the engine with respect to time. In this paper we define the problem in detail, propose a solution based on statistical inference techniques, summarize our experimental evaluation and discuss the applications of this work in various fields from a decision support angle.

In this paper, we explore a data center's performance with a call for energy efficiency through green computing. Some performance metrics we examine in data centers are server energy usage, Power Usage Effectiveness and utilization rate, i.e., the extent to which data center servers are being used. Recent literature indicates that utilization rates at many internal data centers are quite low, resulting in poor usage of resources such as energy and materials. Based on our study, we attribute these low utilization rates to not fully taking advantage of virtualization, and not retiring phantom (unused) servers. This paper describes our initiative corroborated with real data in a university setting. We suggest that future data centers will need to increase their utilization rates for better energy efficiency, and moving towards a cloud provider would help. However, we argue that neither a pure in-house data center or cloud model is the best solution. Instead we recommend, from a decision support perspective, a hybrid model in data center management to lower costs and increase services, while also providing greater energy efficiency.

High accuracy fault diagnosis systems are extremely important for effective condition based maintenance (CBM) of rotating machines. In this work, we develop a fault diagnosis system using time and frequency domain statistical features as input to a backend support vector machine (SVM) classifier. We evaluate the performance of the baseline system for speed dependent and speed independent performance. We show how feature mapping and feature normalization can help in enhancing the speed independent performance of machine fault diagnosis systems. We first perform feature mapping using locality constrained linear coding (LLC) which maps the input features to a higher dimensional feature space to be used as input to an SVM classifier (LLC-SVM). It is seen that there is a significant improvement in the speed independent performance of the fault identification system. We obtain an improvement of 11.81% absolute and 10.53% absolute respectively for time and frequency domain LLC-SVM systems compared to the respective baseline systems. We then explore variance normalization considering the speed specific variations as noise to further improve the performance of the fault diagnosis system. We obtain a performance improvement of 8.20% absolute and 6.71% absolute respectively over the time and frequency domain LLC-SVM systems. It may be noted that that the variance normalized LLC-SVM system outperforms.

The International Conference on Information Retrieval and Knowledge Management (CIKM) brings together three avenues of data-oriented research, namely, Database Management, Information Retrieval and Knowledge Management. The confluence of these avenues becomes evident also in the PhD theses of doctoral students: Stream processing makes use of knowledge representation techniques, linked data is emerging as a research topic that bridges information retrieval and knowledge representation, and new forms of querying draw on techniques from both information retrieval and databases. In this paper, we survey new PhD theses at the meeting point of the three research avenues. Our survey is based on the 5th PhD workshop at the ACM CIKM conference. The topics include themes as diverse as link prediction, source code querying, and video stream processing.

Complexity in autopilot logic design and confusion involved in its mode transition is one of the major reasons for the accidents in highly automated airliner. In this paper we present the usage of a recently proposed array logic based technique for designing the autopilot mode transition logic for a commercial aircraft in the lateral direction. This designing technique helps to reduce the design effort in the development of an autopilot. Ease to understand and very concise way to specify a large number of transitions in simple tabular column is one the highlight of this method. This paper provides some observations about lateral modes and logic concerning lateral mode transition in a less complex way compared to the prevailing methods for autopilot design. Here various mode possibilities of lateral mode transition in an autopilot is mentioned along with specification criteria’s that bound these transition and these possible transitions were given a frame work using MATLAB software

The Semantic Web propels ubiquitous computing by driving the evolution of Web technologies for meaningful user-friendly access. It allows users to apply its potential to find, share and combine information. Ontology based annotations can be used in the Semantic Web to provide better support for organizing knowledge exchange between users. We address this issue to extract meta knowledge from the Web, regardless of devices (e.g., desktops, tablets). Accessing and reusing RDF data from many sources of varying credibility and authority is done through the Semantic Web. This makes the tracking, representation and utilization of various meta knowledge aspects such as reliability, provenance and timestamps extremely important. In this paper, we present an ontology based approach deployed to extract meta knowledge with the following tasks: (1) annotate and extract Wikipedia contents (2) create RDF metadata using the annotations (3) query metadata from the RDF files to obtain the required knowledge. Our general purpose is to link documents to Wikipedia fed to DBpedia / Cyc for extracting the shared lightweight ontology and utilize it for SPARQL querying of documents encompassing meta knowledge. In our approach and experiments we focus on academic scenarios. This knowledge extraction process is useful in developing AI tools and mobile apps where semantics is critical. This work enhances ubiquitous computing by providing efficient and accurate Web access for knowledge extraction on many platforms including desktop computers and mobile devices.

Applications like war gaming, pay per view, disaster relief, video and teleconferencing require a key management approach which will minimize time required for group set up, storage requirement for each user, number of transmissions for initialization & rekeying, as well as it should be bandwidth efficient. Two important aspects for key management are initializing the group with a group key and rekeying the group. Rekeying is necessary due to dynamic nature of multicast group (join/leave semantics) or timely rekeying (periodic). Logical Key Hierarchy (LKH) is an efficient and scalable solution [1][5] to handle keying issue in secure group communication. It minimizes number of transmissions in rekeying as well as storage requirements. In this paper LKH and its Variants are evaluated.

Fine particle pollution is related to road traffic conditions. In this work, we analyze Particulate Matter with a diameter less than 2.5 micrometers, calledPM2.5, along with traffic conditions. This is done for multicity data to study the relationships in the context of environmental modeling. The goal behind this modeling is to support prediction of PM2.5 concentration and resulting air quality. We deploy data mining algorithms in association rules, clustering and classification to discover knowledge from the concerned data sets. The results are used to develop a prototype tool for the prediction of PM2.5 and hence air quality for public health and safety. This paper describes our approach and experiments with examples of PM2.5 prediction that would be helpful for decision support to potential users in a smart cities context. These users include city dwellers, environmental scientists and urban planners. Novel aspects of this work are multicity PM2.5 analysis by data mining and the resulting air quality prediction tool, the first of its kind, to the best of our knowledge.

When it is requiring, releasing a new version of software it is helpful to take reference of old version. If developers know in advance which classes or methods of software may be change or will cause faults then developers can focus more on these classes. It is require identifying classes at earlier stages of development which may cause changes or faults in other classes, so that those classes can be given special attention. The technique presented here improves the quality and reliability of the software. To achieve corrective and adaptive maintenance we require making changes during the software evolution. It is important to analyze the frequency of changes in individual classes and also to identify and show related changes in multiple classes as these changes are key components of software. Prediction of change-prone and fault prone classes of a software is an active topic under research in the area of software engineering. Such prediction can be used to predict changes and faults in different classes of a system from one release to the next release of software. Finding the change-prone and fault prone classes from software in advance can help the developers to focus more attention on these classes. The proposed model presenting a technique for finding dependency of software, change-prone classes and fault-prone classes of Object Oriented Software.

Several research areas today overlap between the tracks of databases, information retrieval and knowledge management, such as natural language processing, semantic web, digital libraries, visualization, information quality and data mining. Inter-disciplinary research across these tracks encourages advances in the development of databases, the extraction of information and the discovery of knowledge. This is precisely the focus of our article. We explain the research issues addressed in a Ph.D. workshop recently held at the ACM Conference on Information and Knowledge Management. This workshop had presentations on novel ideas addressing challenges in information and knowledge management. It covered a broad range of topics such as XML architectures, sensor data streams, personal information managers and text pre-processing. In this article, we provide an overview of the research problems and solutions discussed in the Ph.D. workshop. Our article thus describes the latest technological developments in information and knowledge management as seen by academia. This cutting edge technology also finds practical applications in the corporate world.

Scientific data is often analyzed in the context of domain-specific problems, for example, failure diagnostics, predictive analysis, and computational estimation. These problems can be solved using approaches such as mathematical models or heuristic methods. In this paper we compare a heuristic approach based on mining stored data with a mathematical approach based on applying state-of-the-art formulae to solve an estimation problem. The goal is to estimate results of scientific experiments given their input conditions. We present a comparative study based on sample space, time complexity, and data storage with respect to a real application in materials science. Performance evaluation with real materials science data is also presented, taking into account accuracy and efficiency. We find that both approaches have their pros and cons in computational estimation. Similar arguments can be applied to other scientific problems such as failure diagnostics and predictive analysis. In the estimation problem in this paper, heuristic methods outperform mathematical models.

Scientific experimental results are often depicted as plots of functions to aid their visual analysis and comparison. In computationally comparing these plots using techniques such as similarity search and clustering, the notion of similarity is typically distance. However, it is seldom known which distance metric(s) best preserve(s) semantics in the respective domain. It is thus desirable to learn such domain-specific distance metrics for the comparison of plots. This paper describes a technique called LearnMet proposed to learn such metrics. The input to LearnMet is a training set with actual clusters of plots. These are iteratively compared with clusters over the same plots predicted using an arbitrary but fixed clustering algorithm. Using a guessed initial metric for clustering, adjustments are made to the metric in each epoch based on the error between the predicted and actual clusters until the error is minimal or below a given threshold. The metric giving the lowest error is output as the learned metric. The proposed LearnMet technique and its enhancements are discussed in detail in this paper. The primary application of LearnMet is clustering plots in the Heat Treating domain. Hence it is rigorously evaluated using Heat Treating data. Given distinct test sets for evaluation, clusters of plots predicted using the learned metrics are compared with given actual clusters over the same plots. The extent to which the predicted and actual clusters match each other denotes the accuracy of the learned metrics.

Experimental data in many domains serves as a basis for predicting useful trends. If the data and analysis are available over the Web this promotes E-Business by connecting clientele worldwide. This paper describes such a predictive tool "QuenchMiner™" in the domain "Materials Science". Data mining, more specifically the "Apriori Algorithm", is used to derive association rules that represent relationships between input conditions and results of domain experiments. This enables the tool to answer questions such as "Given cooling medium and agitation during material heat treatment, predict cooling rate". This allows users to perform case studies on the Web and use their results to optimize the involved processes, thus increasing customer satisfaction. Another interesting aspect is predicting material microstructure during heat treatment. Microstructure controls material properties such as hardness. Hence its prediction helps in making decisions about materials selection. Microstructure prediction has similarities to an artificial intelligence process called "Game-of-Life". Some challenges in our work are incorporating domain expert judgement while mining association rules, simulating microstructure evolution under different conditions, and dealing with uncertainty. These challenges and associated research issues are outlined here. To the best of our knowledge, this is the first tool performing Web-based predictive analysis in Materials Science.

In this paper, we propose a decision support system (DSS) for the greening of data centers to help the environment, hence promoting sustainability. As society continues the relentless shift towards electronic communications there is a growing demand for greater storage and processing on data centers. A potential area of improvement is to gain greater server utilization rates since traditionally the phenomenon of "server sprawl" occurs where more servers are added to the data center without seeking greater utilization rates on existing servers first. This implies maintaining more servers than actually needed that translates to greater carbon dioxide emissions causing potential environmental problems. Presently, average server utilization rates in most data centers are rather low, and we make the claim that utilization rates should be increased so that we can lower the number of servers for enhanced sustainability. A shift to the cloud could potentially be useful here. Some servers can be phased out with their operations being hosted on the cloud instead. We propose an approach based on data mining using CBR and decision trees to build a DSS that would help make decisions pertaining to issues such as server sprawl and migration to the cloud in order to promote data center sustainability. We provide recommendations based on our DSS that would be useful to data center operators in academia and industry.

The Web is a steadily evolving resource comprising much more than mere HTML pages. With its ever-growing data sources in a variety of formats, it provides great potential for knowledge discovery. In this article, we shed light on some interesting phenomena of the Web: the deep Web, which surfaces database records as Web pages; the Semantic Web, which defines meaningful data exchange formats; XML, which has established itself as a lingua franca for Web data exchange; and domain-specific markup languages, which are designed based on XML syntax with the goal of preserving semantics in targeted domains. We detail these four developments in Web technology, and explain how they can be used for data mining. Our goal is to show that all these areas can be as useful for knowledge discovery as the HTML-based part of the Web.

The purpose of this paper is to analyze the carbon footprint and utilization rates in a data center. The long-term goal of this work is to give data center administrators an enhanced perspective of data center operations to allow for more energy efficient operation, to lower the carbon footprint, and to promote green data centers. Previous literature shows that low utilization rates in data centers are due to the forecasting of demand to meet spikes in data center use. This management policy has led to many servers running idle the majority of the time which is a waste of resources. We argue that a majority of the data centers should be down sized through decommissioning of phantom servers, virtualization, and shifting spikes in demand to a cloud provider. We use data from the operations of a mid-to-large scale data center in a university. We deploy data mining techniques of decision trees and case-based reasoning to conduct analysis for decision support in cloud computing at data centers. We provide recommendations based on a literature search and our own work. This paper describes our work in progress in the area of developing green data centers.

In this paper, we explore academic data center utilization rates from an energy management perspective with the broader goal of providing decision support for green computing. The utilization rate is defined as the overall extent to which data center servers are being used and is usually recorded as a percentage. Recent literature states that utilization rates at many data centers are quite low resulting in poor usage of resources such as energy and labor. Based on our study we attribute these lower utilization rates to not fully taking advantage of virtualization and cloud technology. This paper describes our research including energy data analysis with our proposed equations for performance measurement and forecasting, corroborated by evaluation with real data in a university setting. We suggest that future data centers will need to increase their utilization rates and thus shift more towards the cloud in order to lower costs and increase services despite current concerns for security of cloud technology.

Adoption of electronic records streamlines how critical information is stored and analyzed. A cloud computing environment provides vast storage capacity and CPU processing power for archival and analytics in a cost-effective manner. In this paper we present a study of deploying the medical markup language MML for storing and analyzing data using cloud services. In addition to emphasizing the advantages of the cloud, we provide a brief risk analysis of using cloud services in such a crucial domain. While we focus our discussion on the medical domain, similar arguments are applicable to other domains with sensitive information.

International audience; The database community is exploring more and more multidisciplinary avenues: Data semantics overlaps with ontology management; reasoning tasks venture into the domain of artificial intelligence; and data stream management and information retrieval shake hands, e.g., when processing Web click-streams. These new research avenues become evident, for example, in the topics that doctoral students choose for their dissertations. This paper surveys the emerging multidisciplinary research by doctoral students in database systems and related areas. It is based on the PIKM 2010, which is the 3rd Ph.D. workshop at the International Conference on Information and Knowledge Management (CIKM). The topics addressed include ontology development, data streams, natural language processing, medical databases, green energy, cloud computing, and exploratory search. In addition to core ideas from the workshop, we list some open research questions in these multidisciplinary areas.

The World Wide Web no longer consists just of HTML pages. Our work sheds light on a number of trends on the Internet that go beyond simple Web pages. The hidden Web provides a wealth of data in semi-structured form, accessible through Web forms and Web services. These services, as well as numerous other applications on the Web, commonly use XML, the eXtensible Markup Language. XML has become the lingua franca of the Internet that allows customized markups to be defined for specific domains. On top of XML, the Semantic Web grows as a common structured data source. In this work, we first explain each of these developments in detail. Using real-world examples from scientific domains of great interest today, we then demonstrate how these new developments can assist the managing, harvesting, and organization of data on the Web. On the way, we also illustrate the current research avenues in these domains. We believe that this effort would help bridge multiple database tracks, thereby attracting researchers with a view to extend database technology.

This paper demonstrates a system called SITAC based on our proposed approach to automate the discovery of concepts (called SITACs) in text sources that are identical semantically but alter their names over time. This system is developed to perform time-aware translation of queries over text corpora by incorporating terminology evolution, thus providing more accurate responses to users, e.g., query processing on Mumbai should automatically take into account its former name Bombay. The SITAC system constitutes a novel collaborative framework of natural language processing, association rule mining and contextual similarity.

In this paper, we propose a decision support system for green computing in data centers on campuses. Green computing aims at the development of technologies for a greener and more sustainable planet. In this work we focus on the greening of data centers that house servers on campuses. It is known that servers consume a huge amount of power and incur other costs in cooling and related operations. It is challenging to meet the demands of efficiency and accuracy in data centers while yet promoting a greener environment along with cost cutting. Our work proposes a decision support system based on decision trees and case-based reasoning that mines existing data in order to assist decision-making for better management in data centers. We consider various aspects of the data from an environmental management perspective such as carbon footprint, thermal profiles and virtualization. Issues such as efficiency and accuracy are taken into account here, In this proposal paper, we outline the need for our work, the challenges involved, the proposed approach for building the system, preliminary evaluation and future plans. This work would be of interest to the data and knowledge management community as well as environmental scientists and researchers in related areas.

Time-stamped documents such as newswire articles, blog posts and other web-pages are often archived online. When these archives cover long spans of time, the terminology within them could undergo significant changes. Hence, when users pose queries pertaining to historical information, over such documents, the queries need to be translated, taking into account these temporal changes, to provide accurate responses to users. For example, a query on Sri Lanka should automatically retrieve documents with its former name Ceylon. We call such concepts SITACs, i.e., Semantically Identical Temporally Altering Concepts. In order to discover SITACs, we propose an approach based on a novel framework constituting an integration of natural language processing, association rule mining, and contextual similarity as a learning technique. The proposed approach has been experimented with real data and has been found to yield good results with respect to efficiency and accuracy.

A challenging area in web based support systems is the study of human activities in connection with the web, especially with reference to certain domains. This includes capturing human reasoning in information retrieval, facilitating the exchange of domain-specific knowledge through a common platform and developing tools for the analysis of data on the web from a domain expert’s angle. Among the techniques and standards related to such work, we have XML, the eXtensible Markup Language. This serves as a medium of communication for storing and publishing textual, numeric and other forms of data seamlessly. XML tag sets are such that they preserve semantics and simplify the understanding of stored information by users. Often domain-specific markup languages are designed using XML, with a user-centric perspective. Standardization bodies and research communities may extend these to include additional semantics of areas within and related to the domain. This chapter outlines the issues to be considered in developing domain-specific markup languages: the motivation for development, the semantic considerations, the syntactic constraints and other relevant aspects, especially taking into account human factors. Illustrating examples are provided from domains such as Medicine, Finance and Materials Science. Particular emphasis in these examples is on the Materials Markup Language MatML and the semantics of one of its areas, namely, the Heat Treating of Materials. The focus of this chapter, however, is not the design of one particular language but rather the generic issues concerning the development of domain-specific markup languages.

Analyzing complex scientific data, e.g., graphs and images, often requires comparison of features: regions on graphs, visual aspects of images and related metadata, some features being relatively more important. The notion of similarity for comparison is typically distance between data objects which could be expressed as distance between features. We refer to distance based on each feature as a component. Weights of components representing relative importance of features could be learned using distance function learning algorithms. However, it is seldom known which components optimize learning, given criteria such as accuracy, efficiency and simplicity. This is the problem we address. We propose and theoretically compare four component selection approaches: Maximal Path Traversal, Minimal Path Traversal, Maximal Path Traversal with Pruning and Minimal Path Traversal with Pruning. Experimental evaluation is conducted using real data from Materials Science, Nanotechnology and Bioinformatics. A trademarked software tool is developed as a highlight of this work.

This paper describes a graphical data mining system called AutoDomainMine. It is based on our proposed approach of integrating clustering and classification to mine scientific data stored in a database. The data consists of input conditions of scientific experiments and graphs plotted as their results. This system mines the stored data in order to submit exact or approximate ranked responses to user queries intended to optimize the scientific processes.

In computational analysis in scientific domains, images are often compared based on their features, e.g., size, depth and other domain-specific aspects. Certain features may be more significant than others while comparing the images and drawing corresponding inferences for specific applications. Though domain experts may have subjective notions of similarity for comparison, they seldom have a distance function that ranks the image features based on their relative importance. We propose a method called features rank for learning such a distance function in order to capture the semantics of the images. We are given training samples with pairs of images and the extent of similarity identified for each pair. Using a guessed initial distance function. Features rank clusters the given images in levels. It then adjusts the distance junction based on the error between the clusters and training samples using heuristics proposed in this paper. The distance junction that gives the lowest error is the output. This contains the features ranked in the order most appropriate the domain. Features rank is evaluated with real image data from nanotechnology and bioinformatics. The results of our evaluation are presented in the paper.

This paper describes the development and use of MatML, the Materials Markup Language. MatML is an emerging XML standard intended primarily for the exchange of materials property information. It provides a medium of communication for users in materials science and related fields such as manufacturing and aerospace. It sets the stage for the development of semantic web standards to enhance knowledge discovery in materials science and related areas. MatML has been used in applications such as the development of materials digital libraries and analysis of contaminant emissions data. Data mining applications of MatML include statistical process control and failure analysis. Challenges in promoting MatML involve satisfying a broad range of constituencies in the international engineering and materials science community and also adhering to other related standards in web data exchange. These issues are being addressed through the development of a good ontology, the automation of format conversions and possible schema extensions. MatML aims to be a lingua franca for data exchange in materials science and its broader horizons.

In scientific domains, knowledge is often discovered from experiments by grouping or clustering them based on the similarity of their output. The causes of similarity are analyzed based on the input conditions characterizing a given type of output, i.e., a given cluster. This analysis helps in applications such as decision support in industry. Cluster representatives form at-a-glance depictions for such applications. Randomly selecting a set of conditions in a cluster as its representative is not sufficient since distinct combinations of inputs could lead to the same cluster. In this paper, an approach called DesCond is proposed to design semantics-preserving cluster representatives for scientific input conditions. We define a notion of distance for conditions to capture semantics based on the types of their attributes and their relative importance. Using this distance, methods of building candidate cluster representatives with different levels of detail are proposed. Candidates are compared using the DesCond Encoding proposed in this paper that assesses their complexity and information loss, given user interests. The candidate with the lowest encoding for each cluster is returned as its designed representative. DesCond is evaluated with real data from Materials Science. Evaluation with domain expert interviews and formal user surveys shows that designed representatives consistently outperform randomly selected ones and different candidates suit different users.