PRIME Research is a company based in Mainz performing market research for companies around the world. PRIME Research analyzes news, social media, and other data sources to extract relevant information. Research areas include: analyzing news coverage of different topics relevant to the client, detecting upcoming new trends, discovering the sentiment of texts etc.
Currently, two PhD scholarships are funded by PRIME Research. The recipients of these scholarships are Zahra Ahmadi and Sophie Burkhardt. Below you find a short summary of the two PhD projects.
The PhD project of Sophie Burkhardt focuses on online multi-label classification and active learning. This means that data arriving in a stream are automatically tagged or labeled with multiple tags. The learning process is sped up to increase time efficiency and annotation cost by actively selecting informative instances to be tagged by a human annotator. The main goal is to combine techniques from online learning, multi-label classification, and active learning to develop scalable methods for analyzing and classifying text data streams.
Zahra Ahmadi works on text stream classification in the presence of recurring concepts. One of biggest challenges of data streams is the occurrence of concept drift and the reappearance of previous concepts. The goal of the project is to propose a framework which handles recurring concepts and keeps the relationship among concepts. To specify the framework for text data, hybrid text mining methods which integrate linguistics knowledge and statistical methods are being used.