Session Details

Day 1 (18 March, 2006)   Embedded Databases Prof. Krithi Ramamritham (KReSIT, IIT Bombay) Abstract : In this era of pervasive computing, computation platforms have extended to small intelligent devices like cell phones, sensors, smartcards, PDAs, etc. As new functionalities and features are being added to these devices, increasing number of applications dealing with significant amounts of data are being developed leading to the need of embedded database support on these devices]. The queries go beyond simple SPJ queries but still have to be locally computed on the device. Thus, there is an increasing need to facilitate execution of complex queries locally on a variety of lightweight computing devices. However, scaling down the database techniques poses challenges since these lightweight devices come with very limited computing resources. Also, the devices are not uniformly endowed with resources, the computing capability and main memory of a cell phone differs from that of a PDA. It is essential that the available resources be utilized optimally for a database system that is developed for such devices. Approaches to meeting these challenges will be the focus of this talk. About the Speaker : Prof. Krithi Ramamritham is currently at IIT Bombay as Head of Department, KReSIT and as the Vijay & Sita Vashee Chair professor in the Deptartment Of Computer Science & Engineering. He is a fellow of the IEEE, ACM. His areas of interest include real time systems, transaction processing in Database Systems and real-time database systems.   Content Dissemination Networks for Dynamic Data Shetal Shah (CSE, IIT Bombay) Abstract : Data collected from sensors, traffic information, sport scores, and stock prices is streaming and dynamic, i.e, it changes continuously, rapidly and unpredictably. Timely dissemination of such data to interested users is essential since it is used on-line monitoring and on-line decision making. Traditional dissemination methods involve a pull or a push of data between the sources and the clients. For dynamic data, it is very expensive and in certain cases infeasible to to send all changes to all users. Users of such data are often willing to tolerate a certain amount of imprecision in the value which makes this problem tractable. The focus of our work has been to design and build a content dissemination network for dynamic data that (a) preserves the user specified bound on tolerable imprecision, (b) is resilient to failures and (c) is scalable (it should meet the needs of a large number of clients). In this talk, we will first introduce the problems involved in the dissemination of dynamic data and then discuss some of the interesting issues which needed to be considered for building content dissemination networks for dynamic data. About the Speaker : Ms. Shetal Shah is a research scholar at CSE, IIT Bombay, working under Prof. Krithi Ramamritham. Her current work involves building content dissemination networks for dynamic data. Her areas of interest include Internet research, databases, algorithms, and compilers.   Computational Linguistics: A Data Management Perspective Dr. A. Kumaran (Microsoft) Abstract : With the explosive growth of data in the Internet and other media, it is becoming progressively more difficult for users to search for specific piece of information or to analyze the data for their general information needs. Given that most information is in natural language (whether in the form of text or other rich media), it is imperative that accurate and efficient technologies developed for processing the data for user consumption. Computational Linguistics research holds the key to develop tools that bridge the gap between the data and the user's information needs. Over the recent decade, there had been a paradigm shift in Computational Linguistics research: The research techniques have transformed from being an oriented towards development of tools for linguistics research, to active learning-based approaches that infers linguistic knowledge from data. The research has an underlying need for storing and processing large amounts of rich linguistic data. This talk presents the types of research in Computational Linguistics, the special data and processing requirements that go with them. Also we explore the sufficiency of conventional database systems to host the data to support such research. About the Speaker : Dr. A. Kumaran heads the Multilingual Systems Research group in Microsoft Research India, located in Bangalore. He completed his doctoral degree from Indian Institute of Science in the area of Multilingual Databases. Before Microsoft Research, he has worked for nearly a dozen years at the Oracle Corporation and at Bell Communications Research, where he was involved in developing document processing components for database systems.   Technology challenges in a database system for implementing a data warehouse Mr. Ramesh Bhashyam(Teradata) Abstract : A data warehouse integrates large volumes of detailed and current data across entire organizations. It provides a unified view of operational and historical data and enables diverse forms of decision making from the same data model. It often uses as a foundation a detailed data model and enables different summaries and views, as appropriate to the business, to be built on top of the detailed model. This talk will discuss some of the technology challenges that need to be addressed when designing a database system for a data warehouse. It will touch upon the challenges posed by high capacity drives and the mechanisms that are appropriate for addressing that. It will consider the features and capabilities required of a database in a mixed application environment of a warehouse. The usage of a data warehouse has evolved from reporting and decision support system to mission critical decision-making operational systems. Data warehouses are often used for mining types of applications. These applications read massive volumes of data from within the data warehouse and are demanding of both CPU and IO resources. Data warehouses are also used for operational decision-making applications, applications which read a small but well-focused set of data from the warehouse, and use few resources. These operational applications are differentiated by short response time requirements, making scalability a challenge. Data warehouses that combine both these types of applications require that the operational data be integrated, current and up to date. The common belief that warehouse data is static is no longer valid. This talk will examine some of the technology challenges in a database system for implementing a data warehouse About the Speaker : Mr. Ramesh Bhashyam has been with Teradata Corp for the last 15 years. He has experience is various aspects of building a data base system. His primary interests are in query optimization, scalable algorithms, and performance analysis. Ramesh was the CTO for Teradata Database Engineering organization from 1999 to 2005. Prior to joining Teradata Ramesh was with Inference Corporation. Ramesh has a Masters in Computer Science from USC.   Context-Oriented Information Integration Dr. Mukesh Mohania (IBM IRL, New Delhi) Abstract : With critical business information distributed across both structured and unstructured data sources, enterprises are increasingly realizing the importance of seamlessly integrating relevant structured and unstructured data. Existing information integration solutions typically address this issue by providing a single point of access for both structured and unstructured data sources. This is not enough, since the application still needs to formulate the SQL logic to retrieve the needed structured data on one hand, and identify a set of keywords to retrieve the related unstructured data on the other. This is a limitation since (a) the same information need needs to be formulated using two disparate paradigms, which is redundant effort, and (b) in many cases, it is hard (even impossible) for the application to identify appropriate keywords needed as above to retrieve related unstructured data. The SCORE project addresses this limitation by following a novel approach to information integration. In this approach, the application specifies its information needs using only a SQL query on the structured data, and the system automatically "translates" this query into a set of keywords that can be used to retrieve relevant unstructured data. In this talk, we describe the techniques used in SCORE for this query translation, and also present an experimental study that illustrates the effectiveness of these techniques. About the Speaker : Dr. Mukesh Mohania is a Research Staff Member at IBM India Research Lab, Delhi, and manages a database group. He has worked in the areas of distributed databases, data warehousing, semistructured databases, and autonomic computing. He was awarded Technical Achievement Award in the area of Web Database Management and Data Warehousing by Association of Database and Expert Systems Applications in Greenwich, U.K., 2000.     Day 2 (19 March, 2006)   Integrating unstructured data into structured relational databases Prof. Sunita Sarawagi (KReSIT, IIT Bombay) Abstract : Database systems are islands of structure in a sea of unstructured data sources. Several real-world applications now need to create bridges for smooth integration of semi-structured sources with existing structured databases for seamless querying. This integration requires extracting structured column values from the unstructured source and mapping them to known database entities. Existing methods of data integration do not effectively exploit the wealth of information available in multi-relational entities. We present statistical models for co-reference resolution and information extraction in a database setting. We then go over the performance challenges of training and applying these models efficiently over very large databases. About the Speaker : Prof. Sunita Sarawagi is currently working as an associate professor at KReSIT. She is one of the top researchers in field of datamining and data warehousing. Her fields of interest also include machine learning and databases. She holds many patents in data mining and related fields.   Type-enabled Proximity Search Using Entities and Relations Prof. Soumen Chakrabarti (CSE, IIT Bombay) Abstract : We introduce a new, powerful class of text proximity queries: find an instance of a given answer type (person, place, distance) near selector tokens matching given literals or satisfying given ground predicates. An example query is type=distance NEAR Hamburg Munich, which might be a rough translation of the information need "find the distance between Hamburg and Munich". Nearness is defined as a flexible, trainable parameterized aggregation function of the selectors, their frequency in the corpus, and their distance from the candidate answer. Such queries provide a key data reduction step for information extraction, data integration, question answering, and other text-processing applications. We describe the architecture of a next-generation information retrieval engine for such applications, and investigate three key technical problems faced in building it. First, we discuss how to translate natural language questions into queries in our framework. Second, we propose a new algorithm that estimates a scoring function from past logs of queries and answer spans. Plugging the scoring function into the query processor gives high accuracy: typically, an answer is found at rank 2 or 3. Third, we exploit the skew in the distribution over types seen in query logs to optimize the space required by the new index structures required by our system. Extensive performance studies with a 10GB, 2-million document TREC corpus and several hundred TREC queries show both the accuracy and the e ciency of our system. From an initial 4.3GB index using 18,000 types from WordNet, we can discard 88% of the space, while inflating query times by a factor of only 1.9. Our final index overhead is only 20% of the total index space needed. We also describe the architecture of SPIN, a PIM system for searching personal information networks, which provides a wider class of proximity queries on graphs with text data. Joint work with Sujatha Das, Srivatsa Iyengar, Harsh Jain, Vijay Krishnan and Kriti Puniyani. About the Speaker : Prof. Soumen Chakrabarti is currently at the Department of Computer Science and Engineering at the Indian Institute of Technology, Bombay, where he has been an Associate professor since 2003. His current research interests include question answering, Web analysis, monitoring and search, mining irregular and relational data, and textual data integration.   DBMS Performance - A Multi-dimensional Challenge Mr. Vadiraja Bhatt (Sybase) Abstract : Over the past 2 decades traditional relational database servers have grown to become critical back-end for almost all applications that there is. With that they have endured several challenges that emerged out of user requirements, new generation applications, ever changing hardware architectures and numerous operating environments. Each of them have posed very unique problem to solve and DBMS have adapted very well to these challenges acceptable throughput and response times. This talk will go over some of these challenges and the ways traditional DBMS have answered them with various features, tunable and architectural changes. About the Speaker : Mr. Vadiraja Bhatt is a Senior Software Engineer with Sybase Inc., Pune and has actively contributed to database products of Sybase. He is an alumnus of IIT Bombay and had completed his postgraduation from CSE department in 1994.   Multilingual Databases Prof. Jayant Haritsa (IISc. Bangalore) Abstract : Efficient storage and query processing of data spanning multiple natural languages are of crucial importance in today's globalized world. A primary prerequisite to achieve this goal is that the defacto standard data repositories -- relational database systems -- should efficiently and seamlessly support multilingual data. In this talk, we will first investigate how good today's database systems are with regard to the storage, management and processing of multilingual data, and demonstrate that there are significant performance inefficiencies for languages based on scripts other than Latin (such as Devanagari). We will then outline techniques for alleviating these problems. With regard to functionality, current SQL does not support querying of data across different natural languages, that is, cross-lingual queries. To address this lacuna, we will present two new SQL operators that support phoneme-based matching of names, and ontology-based matching of concepts, in the multilingual world. In a nutshell, we will discuss in this talk the research challenges that have to be addressed in achieving the ultimate goal of "natural-language-neutral" database engines. About the Speaker : Prof. Jayant Haritsa is on the faculty of the Supercomputer Education & Research Centre and the Department of Computer Science & Automation at the Indian Institute of Science, Bangalore. His research interests are in Database Systems (Query Optimization, Data Mining, XML Databases, Biological Databases, Multi-lingual Databases)   Fine Grained Authorization in Databases Prof. Sudarshan (IIT Bombay) Abstract : Current day database applications, with large numbers of users, require fine-grained access control mechanisms, at the level of individual tuples, not just entire relations/views. Unfortunately, standard SQL does not have a way to specify such access control, and most database systems cannot handle thousands of users, let alone millions for Web scale applications. Fine-grained access control is therefore typically enforced in the application code. This approach has numerous drawbacks, such as difficulty of management, and lack of transparency, which can be avoided by devising mechanisms for specifying/enforcing fine-grained access control at the database level. In recent years, there has been a fair amount of research in this area, and several database products, such as Oracle and Sybase, have incorporated forms of fine-grained access control. In this talk we present an overview of fine-grained access control in databases. We concentrate on conceptual issues, which apply across different products. About the Speaker : Prof. Sudarshan is a Professor in the Department of Computer Science and Engineering at IIT Mumbai (Bombay). He is an author of over 60 papers in different areas of database systems, and holds 13 patents. His current research interests include query processing and optimization, authorization, and keyword querying of databases.