Devasis Bassu

Telcordia LSI Engine: implementation and scalability issues

Latent Semantic Indexing (LSI), a vector space-based approach to information retrieval, has been proven to be an effective tool in correlating and retrieving relevant documents. While much work has been published on LSI, most of it addresses the algorithmic or theoretical basis of the model. Little, if any, presents implementation issues in practice. We describe a production-level implementation of LSI. The system integrates components including document collection and preprocessing, singular value decomposition (SVD), multilingual processing, and a tree-based access method for similarity querying. We discuss implementation issues encountered during the development of the system. In particular, we address scalability issues in the query engine and various components of the system, and present lessons learned.

Generating Domain Specific Graphical Modeling Editors from Meta Models

We describe an approach for automatically generating application aware graphical modeling environments from the meta-model specification of an application domain. A generated graphical modeling environment: a) provides domain-specific graphical metaphors in the modeling palette, b) imposes domain-specific modeling constraints to prevent semantically incorrect models, and c) provides domain-specific operators and languages to capture application domain constraints. The domain meta-model is specified using a meta-model which is an extension of the UML meta-model. One of the major advantage of using application domain-specific modeling environment is to make it easy for business analyst to create semantically correct models. We use UML2.0 to specify the domain metamodel. The advantage of using UML2.0 representation is the reuse of vendor-supported technologies including MDA tools. An implementation using the Eclipse framework is also discussed

Mapping domain expertise within teams: visual stimulation of knowledge-building through collaboration

Recent R&D efforts for the U.S. Intelligence Community have been directed towards providing new information technology that better supports collaborative war-gaming and decision-modeling among expert teams so as to produce valid and reliable intelligence in a more-timely, less-costly manner. This paper describes Schemer, a Web service that provides knowledge validation and collaboration support to virtual panels of experts. Of particular interest is the way that Schemer uses visualization of knowledge distribution among panelists to motivate collaboration, with the goal of fostering greater consensus and knowledge-building, hence, better intelligence. Since an additional requirement for supporting virtual panels is incremental data acquisition from panel members, models must often be computed from incomplete data. The same visualization techniques for evaluating consensus formation and knowledge-building among panelist also prove useful for evaluating the impact of data imputation on model validity and reliability.

Collaborative Panel Administrator: A framework and implementation for expert panel lifecycle management

In this paper, we describe the Collaborative Panel Administrator, a collaboration framework designed to manage the lifecycle of virtual teams of domain experts working in collaborative gaming and modeling. A unique feature of the CPA is the way it is guided by sound experimental research design so that information collected from panelists can be vetted, and knowledge built from this information can be validated quantitatively. The framework is deployed as a J2EE application and provides extension points to integrate critical services needed to foster collaboration among panelists and analyze the information acquired from them. We examine in detail the CPApsilas integration with a commercial collaboration platform.