Business Information Service On the Network (BISON)

Information analytics technologies, especially text mining, face several key challenges that inhibit such technologies to reach their full potential.

• Complexity: Many information analytics technologies are extremely complex to use, making it difficult for business users to identify the right data sets and enhance the business decision-making process. Is annotation on names and organizations alone sufficient, or is clustering and taxonomy creation required? Our experience with customers in many different domains indicate that different combinations of analytics technologies are needed for different customers. There is a significant need to develop technologies that can quickly compose the right analytics components into solutions to solve business problems in different domains.
• Performance and Scalability: Technology advances have enabled users to generate and store vast amounts of unstructured information. Text mining must scale in order to discover insights hidden in large amounts of unstructured data. This requires fast crawling, indexing and search capabilities as well as efficient algorithms for clustering, classification and annotation. Today, significant gaps exist in this area.
• Data Source Identification, Access, Transformation, and Loading: Enterprise information comes from various data sources including transactional systems, business support systems, partner systems, etc. The correct data sources must be identified and proper data must be extracted and cleansed before performing analytics on the data. Data that is not sufficiently normalized and carefully preprocessed may lead to an inconclusive or incorrect analysis result.

To address these challenges, we are developing an approach that treats business information analytics as services. Recent advances in networking and internet technologies have fueled the emergence of new computing models for industries. One of the most prominent new computing paradigms is the Service-Oriented Architecture (SOA). SOA allows organizations to quickly leverage various IT technologies offered by service providers to solve their business problems without purchasing expensive hardware and software each time. We envision that by combining business information analytics and SOA, we can create a new business information analytics service framework, offering on-demand information analytics services for different business needs with a number of key advantages. These advantages include better integration with intra- and inter-organization computing capabilities; enhanced and custom computing environments (by integrating available analytics services from different providers); larger customer base coverage; and a high level of solution scalability, reliability, extensibility and dynamic provisioning.

Business Information Service On the Network (BISON) is the next generation of BIW, based on SOA. BISON divides information analytics into granular analytics service units (ASUs), e.g., clustering/taxonomy services, annotation services, statistical analysis services, etc. With the emerging adoption of Web Services technologies, ASUs can be defined by standard WSDLs (Web Services Definition Languages), and the service discovery and registration mechanisms can be driven by UDDI (Universal Description, Discovery and Integration) mechanisms. An overall SOA-based analytics framework allows the independent development of different ASUs. As long as the web services are defined by standards such as WSDL, they can be registered, discovered, aggregated into a ubiquitous framework and used to compose different analytics solutions based on customer needs.

In addition, BISON combines a Solution Composer runtime service to allow real-time analytics solution composition without lengthy and tedious programming or customization. With such capabilities, we envision that users will be able to develop analytics solutions in minutes or hours, rather than months, and without a single line of code. The figure below shows the overall BISON architecture:

BISON Architecture

BISON uses J2EE standards as the implementation framework. The overall BISON architecture contains several key Web Services components, i.e., Text Analytics Services (TAS), Data Access Services (DAS), Data Source Services (DSS), and ETL Services as shown in the figure below. TAS encompasses various ASUs from which users can choose. DAS is mainly responsible for querying data sources to retrieve data of interests and performing additional functions such as data filtering data joins. DSS may provide structured and unstructured data sources and metadata information. ETL provides services for information warehouse building.

Both DAS and DSS are designed to be interoperable with other SOA-based information integrators such that additional data sources can be easily incorporated. In addition to the core services components, BISON provides a workflow to allow easy integration of analytics solutions developed for different domains into business processes, as shown at the top layer of the system stack. The overall BISON infrastructure also includes administration, security, monitoring and service registration, as well as pricing and metering support. Such capabilities are necessary to ensure that the overall SOA-based information analytics framework can be realistically and successfully deployed for different customer cases.

BISON Layers