Cooperations

Content SA

Duration: 11/2017 - 01/2018

Which texts are read more frequently on the web? Does a negative headline lead to more clicks? These findings are incorporated into the automatic sentiment analysis of editorially created texts at Content Garden Technologies in order to make the provision of information as attractive as possible for readers.

Categorizing job advertisements and determining fitting job advertisements

Duration: 02/2017 - 12/2017

Together with Jobware Online-Service GmbH, approaches for the automatic classification of job advertisements and for semantic searches in the job portal are being developed. More than 200,000 job advertisements that have already been categorized manually are available as training data. This offers the opportunity to implement rule-based approaches for information extraction using bootstrapping on the one hand and machine learning methods on the other, in order to create a hybrid process that is both knowledge-based and data-driven.

Conception of a solution assistant for customer questions

Duration: 02/2016 - 04/2016

LYNX B.V. is faced with an increasing number of customer requests every day. Only a small proportion of these can be answered reliably by the automatic question and answer system. One of the reasons for this is the large number of variants in the formulation of requests. The focus of the collaboration is the further development of an information retrieval system with regard to greater flexibility in the interpretation of customer requests (e.g. through extensive synonym dictionaries, suggestion system, error correction) in order to offer needs-oriented solution suggestions and further tips.

WöhlerIE

Duration: 11/2017 - 01/2018

As part of the cooperation, an application for extracting company-specific data from online resources is being developed to support the future evaluation of potential clients from the HVAC trade in terms of their company size and tradition. For this purpose, a supervised machine learning process is being developed to recognize relevant text modules in the self-presentations and service descriptions on the respective company websites.

Search Machine Optimization of the German Website

Duration: 03/2015 - 12/2016

Optimizing the existing website www.woehler.de with regard to the mostly unknown but sensitive criteria of the major search engine providers is particularly difficult due to the existing page structure and content, as changes must be prevented from having a negative impact on the current ranking.

Intelligent Tagging of Measuring, Inspection and Cleaning Technology for the Chimney Sweep and HVAC Trade

Duration: 07/2014 - 09/2014, 01/2015 - 03/2015, 07/2015, 10/2015

Intelligently indexing the product range in the Wöhler Messgeräte Kehrgeräte GmbH webshop is no easy task given the numerous variants. For example, while dust measuring device and dust knife mean the same technical product, measuring device and knife are only synonymous depending on the context. As part of the cooperation, new ways of semantically indexing technical terms in German, English, French, Italian and Dutch are being explored.

Conception of a Functional Search Prototype for the Helpdesk

Duration: 01/2014 - 03/2014

The aim of the cooperation was to redesign the existing help system of dSPACE digital signal processing and control engineering GmbH. Making technical documentation searchable is no easy challenge. Through intensive collaboration, the participants were able to identify and systematize user requirements and successfully implement the prototype of the help search.

Feasibility Study on Automated Text Analysis of Insurance-Relevant Documents

Duration: 12/2016 - 03/2017

Every day, insurance companies receive a large number of documents for processing. But how can the type of textual information (e.g. cover letter, invoice, expert opinion, contract, etc.) be recognized automatically? The aim of this project was to explore which methods, from OCR and classification to the structured representation of these documents, could be promising in the insurance context.

Analysis of Keywords and Emotions in Review Comments

Duration: 10/2015 - 12/2015, 03/2016 - 04/2016

Review BRIDGE Research GmbH deals with the acquisition and analysis of product reviews on the Internet. The aim is to obtain a simple and quick overview of what the world thinks about a product. As part of our cooperation, modern methods of sentiment analysis and customer satisfaction monitoring are expanded and applied. Among other things, extensive polarity lexicons are created with computer support in order to increase the accuracy and coverage of the analyses.