Roland Fleischhacker (Dr. Sonja Kabicher-Fuchs)


Using semantic technologies to identify the concerns of a caller in a big contact center of Stadt Wien – Wiener Wohnen

Co-presenter: Dr. Sonja Kabicher-Fuchs, Product Owner, Stadt Wien – Wiener Wohnen Kundenservice GmbH


Description Stadt Wien – Wiener Wohnen

As one of the largest property management companies in Europe Stadt Wien - Wiener Wohnen (WW) manages approximately 220,000 community-owned apartments, 47,000 parking spaces and 5,500 shops. More than half a million tenants, and thus about a quarter of Vienna's city population, cause about 1.5 million inquiries to the contact center per year.

Description of the problem

The topics of the callers, for example tenants, are manifold and can range from simple information to the description of technical problems. Caller’s concerns need to be identified as well as recorded in an accurate, documented and fast way.

Former solution

WW works with a range of standard business processes, which continuously increase . The goal of the call center agent is to identify the correct business process by listening to the caller. The business process covers the wording of the answer, the questions which have to be answered for the detailed problem description and which technical business process has to be started within the workflow system.
In the past the call center agents used a hierarchical topic tree and a full text search engine.

Disadvantages of the former solution

Whereas experienced agents knew the position of frequent topics in the topic tree out of their mind, unexperienced colleagues or rarely occurring problems caused longer search times. The agents selected business cases without any further choice support by the system.

Consequences of the former solution

The consequences of the former solutions were long search times, high efforts in changing the business case during the call if the appropriate business case was not selected, and highly specific trainings. 

The goal

The intended target was to combine two steps, which were in the past separated: the recording of the concern and the identification of the business process. The call center agent should concentrate on the conversation and the expert system should ensure that every agent, whether experienced or not, has the possibility of an optimal information provisioning.

Solution roadmap

The first step was the analysis of earlier call notes to identify the domain-specific language used by the call center agents. DEEP.knowledge, the knowledge base of the DEEP.delphi Product Suite, which covers about 90.000 concepts and their relations was supplemented by approximately 5,000 domain-specific concepts and abbreviations.
The second step was the implementation of a prototype, which covered all concerns dealing with defects and water, like tube defects.
A huge amount of test cases brought important insights for the implementation of the final version, which was implemented parallel to the training of standardized business processes. 

Technical solution

DEEP.assist is an application based on the semantic platform DEEP.delphi Product Suite. It covers the whole process of knowledge management, from importing the standardized business processes, the semantic training and the management of the knowledgebase, to analyzing and decision finding and finally to a bunch of monitoring tools as a basic component for the self-learning mechanisms.

Organizational solution

Besides the technical implementation the team developed a knowledge maintenance process, led by a Knowledge Officer, which covers the whole life-cycle of knowledge management within WW.


DEEP.assist resulted in a quantum leap in terms of identifying speed and accuracy. 
The original business plan considered the main advantage in the acceleration of the telephone conversation and the reduction of the call costs. However, it turned out that the reduction of the on boarding costs for new call center staff open the real potential for cost reduction. 

Lessons learned

The most important lessons learned were: 
Misspelling tolerance is even more important than expected. 
Updates of new business processes must be able within a very short period of time.
Software represents only one part of the solution. Organization is as important as technology.
Changing typical search behaviors (such as the keyword based search) requires time.

Next steps

Next steps are to develop an enhanced business process training tool and an enhanced user monitoring tool, to detect vulnerabilities of business process training.
Furthermore, there are considerations to integrate DEEP.assist into a self-service helpdesk for administrative processes, in order to be able to answer questions like “What do I have to do, when I need care leave, because my child has mumps?”


  • Technical University Munich and Vienna
  • Foundation of Austria’s first SAP Consulting Company
  • Tradesale and further management of one of Austria’s leading consulting companies
  • Foundation of software companies dealing with semantic and cognitive technologies