Exploratory research project for industrial R&D in the field of Business Intelligence (BI) for small to medium enterprises.
UNDERSTANDER is an exploratory research project for industrial R&D in the field of Business Intelligence (BI) for small to medium enterprises. We develop a knowledge-based model of BI, which can be parameterised by the end-user to cover their needs. The model can then be used to “prime” a software agent and the agent can then search autonomously for relevant information in the WWW or any intranet. The gathered information gets stored in a knowledge repository and can be retrieved via interactive reporting functions.
Initial situation and problem to solve
For years, there has been a high-end market for enterprise-level Business Intelligence (BI), which has been traditionally served by the large software houses such as Oracle, Microsoft, SAP or IBM. These solutions are much too complex and expensive to address the needs of small to medium enterprises (SMEs). These SMEs are at risk of being left out when it comes to utilizing the opportunities afforded by the WWW. UNDERSTANDER develops a basic technology that will enable small technology providers and consultant firms, to offer affordable BI solutions to SMEs whose core business is elsewhere, but who would like to acquire pertinent BI in their line of business and who can only afford a „self-serve“ approach. In this project, we focus on the BI aspect of competitive intelligence found on the Web, i.e. finding information that lets a firm assess its own position with respect to its competitors.
Planned goals and methodology
The aim of UNDERSTANDER is to give the target SME a highly automated BI solution for their line of business. This will be implemented as a software agent that searches for relevant web-resources (or intranet). At the heart of the search is an explicit domain model of BI that can be parameterised to suit the specific needs of the end-user SME. The domain model is injected into the evaluation function of the agent (this is the function which regulates the agent’s search behaviour). The BI model will be encoded in a variant of R.C. Schank’s Conceptual Dependency Theory from the early 1970s. The software agent translates documents that fit the customised schemata, into a series of statements that are valid according to the model. It then saves these statements in a knowledge base, together with references to the sources. Our chosen method is to tackle the four main issues in “R&D sprints” of 2 months duration, followed by two validation rounds. The sprints include the building of feasibility prototypes in SWI-Prolog. We will use RESTful services for communication.
Expected Results and Findings
We expect four major results: (1) a generally applicable, controlled vocabulary for BI; (2) an ontologised set of query scripts in the tradition of Conceptual Dependency Theory that the agent uses for finding relevant Web-resources for competitive intelligence and storing information about them; (3) the actual implementation of such a user agent and; (4) the implementation of a knowledge repository to which the agent can commit formal knowledge statements and which the end-user can query interactively.