Maria Muslea

Integration of heterogeneous knowledge sources in the CALO query manager

We report on our effort to build a real system for integrating heterogeneous knowledge sources with different query answering and reasoning capabilities. We are conducting this work in the context of CALO (Cognitive Assistant that Learns and Organizes), a multidisciplinary project funded by DARPA to create cognitive software systems.

Semi-automatically mapping structured sources into the semantic web

Linked data continues to grow at a rapid rate, but a limitation of a lot of the data that is being published is the lack of a semantic description. There are tools, such as D2R, that allow a user to quickly convert a database into RDF, but these tools do not provide a way to easily map the data into an existing ontology. This paper presents a semi-automatic approach to map structured sources to ontologies in order to build semantic descriptions (source models). Since the precise mapping is sometimes ambiguous, we also provide a graphical user interface that allows a user to interactively refine the models. The resulting source models can then be used to convert data into RDF with respect to a given ontology or to define a SPARQL end point that can be queried with respect to an ontology. We evaluated the overall approach on a variety of sources and show that it can be used to quickly build source models with minimal user interaction.

Mixed-initiative, multi-source information assistants

While the information resources on the Web are vast, the sources are often hard to find, painful to use, and difficult to integrate. We have developed the Heracles framework for building Web-based information assistants. This framework provides the infrastructure to rapidly construct new applications that extract information from multiple Web sources and interactively integrate the data using a dynamic, hierarchical constraint network. This paper describes the core technologies that comprise the framework, including information extraction, hierarchical template representation, and constraint propagation. In addition, we present an application of this framework, the Travel Assistant, which is an interactive travel planning system. We also briefly describe our experience using the same framework to build a second application, the WorldInfo Assistant, which extracts and integrates geographic-related data about countries thorughout the world. We believe these types of information assistants provide a significant step forward in fully exploiting the information available on the Internet.

Neuroscience data integration through mediation: An (F)BIRN case study

We describe an application of the BIRN mediator to the integration of neuroscience experimental data sources. The BIRN mediator is a general purpose solution to the problem of providing integrated, semantically-consistent access to biomedical data from multiple, distributed, heterogeneous data sources. The system follows the mediation approach, where the data remains at the sources, providers maintain control of the data, and the integration system retrieves data from the sources in real-time in response to client queries. Our aim with this paper is to illustrate how domain-specific data integration applications can be developed quickly and in a principled way by using our general mediation technology. We describe in detail the integration of two leading, but radically different, experimental neuroscience sources, namely, the human imaging database, a relational database, and the eXtensible neuroimaging archive toolkit, an XML web services system. We discuss the steps, sources of complexity, effort, and time required to build such applications, as well as outline directions of ongoing and future research on biomedical data integration.

Karma: A System for Mapping Structured Sources into the Semantic Web

The Linked Data cloud contains large amounts of RDF data generated from databases. Much of this RDF data, generated using tools such as D2R, is expressed in terms of vocabularies automatically derived from the schema of the original database. The generated RDF would be significantly more useful if it were expressed in terms of commonly used vocabularies. Using today’s tools, it is labor-intensive to do this. For example, one can first use D2R to automatically generate RDF from a database and then use R2R to translate the automatically generated RDF into RDF expressed in a new vocabulary. The problem is that defining the R2R mappings is difficult and labor intensive because one needs to write the mapping rules in terms of SPARQL graph patterns. In this work, we present a semi-automatic approach for building mappings that translate data in structured sources to RDF expressed in terms of a vocabulary of the user’s choice. Our system, Karma, automatically derives these mappings, and provides an easy to use interface that enables users to control the automated process to guide the system to produce the desired mappings. In our evaluation, users need to interact with the system less than once per column (on average) in order to construct the desired mapping rules. The system then uses these mapping rules to generate semantically rich RDF for the data sources. We demonstrate Karma using a bioinformatics example and contrast it with other approaches used in that community. Bio2RDF [7] and Semantic MediaWiki Linked Data Extension (SMW-LDE) [2] are examples of efforts that integrate bioinformatics datasets by mapping them to a common vocabulary. We applied our approach to a scenario used in the SMW-LDE that integrate ABA, Uniprot, KEGG Pathway, PharmGKB and Linking Open Drug Data datasets using a

Mind your metadata: exploiting semantics for configuration, adaptation, and provenance in scientific workflows

Scientific metadata containing semantic descriptions of scientific data is expensive to capture and is typically not used across entire data analytic processes. We present an approach where semantic metadata is generated as scientific data is being prepared, and then subsequently used to configure models and to customize them to the data. The metadata captured includes sensor descriptions, data characteristics, data types, and process documentation. This metadata is then used in a workflow system to select analytic models dynamically and to set up model parameters automatically. In addition, all aspects of data processing are documented, and the system is able to generate extensive provenance records for new data products based on the metadata. As a result, the system can dynamically select analytic models based on the metadata properties of the data it is processing, generating more accurate results. We show results in analyzing stream metabolism for watershed ecosystem management.

Conditional constraint networks for interleaved planning and information gathering

We have developed Heracles II, a framework for mixed-initiative planning and information gathering. Heracles II maps the hierarchical task structure of the planning domain into a conditional constraint network. It also ensures correct constraint propagation in the presence of cycles, user interaction, and asynchronous sources. We have applied the Heracles II framework to several domains including travel planning and geospatial data integration.