Search

This paper describes the architecture of a web farming system, that has been created during AMI-SME project as a tool for supporting small and medium sized enterprises employees in performing marketing research. Due to business objectives that have been defined by the project managers, the system must be extremely flexible and configurable allowing easy adaptation to a variety of text mining and information retrieval related tasks, from simple web based document searching to complex concept visualization using ontologies and information clustering algorithms. As such it constitutes an interesting component framework, upon which different research and business oriented systems using text mining methodology can be built. [ABSTRACT FROM AUTHOR]

This paper is an overview of a Text Mining project carried out by the Warsaw University of Technology in the business analysis department of a Polish mass services company. This project has provided an excellent opportunity to test the usefulness of existing TM tools and algorithms for processing a large CRM database constituting a text corpus in Polish language. As these tools proved to be highly inadequate for the task, some new algorithms and methods have been devised, that would be specifically tailored to peculiarities of CRM data in general and Polish language in particular. Noteworthy, some of the solutions implemented were based on publicly available open source software, which was modified and extended by the project team, thus allowing one to evaluate the robustness of open source code for large scientific and business deployments. The paper describes briefly the Text Mining problem that had to be tackled and contains overview of the implemented solutions, together with information about obtained experimental results. [ABSTRACT FROM AUTHOR]