Challenges always arise during information fusion, especially for large enterprise information systems (EIS) such as British Telecommunications (BT), which wishes to have its people, especially its sales staff, forge stronger links between one another in order to make connecting with customers easier.

This paper is specifically a live example of how to solve such challenges. Many solutions in this work are valuable reference sources for researchers focusing on designing a multi-source information integration system.

In this paper, the authors describe how they used ontology-based data management (OBDM) and linked data to integrate enterprise data with open and social media. They built a system using linked data principles to bring data from Web 2.0 sites and other external business sites and link them together with pertinent internal BT enterprise data into that enterprise data space. Furthermore, this work intensively covers the challenges they have encountered and their solutions and general principles.

One of the key challenges when meshing their internal LinkedIn data is in disambiguating a person’s identification, especially to decide whether similar names refer to the same person. Instead of using common natural language processing (NLP) techniques, they use employee identification numbers (EIN) and their company email addresses as the identifiers. The system also integrates data from their local Salesforce system. In order to input further understanding of this data, public information of these accounts is sourced from OpenCorporates.

Another major challenge they faced is the potential loss of context and semantics of the data when it is transported across the boundaries of departments and companies.

In general, the two principles in designing the proposed system are building an appropriate business case and using a unified ontology that describes the domains of interest. For future work, they will be extending the system and integrating additional resources both from the enterprise and externally.