Area of research: Language and Knowledge Processing

Our research in the field of language and knowledge processing is divided into three streams:

Knowledge Discovery and Big Text-Data Analytics deals with methods for knowledge discovery that are scalable to large quantities of data, as well as with the methods for text mining. In these methods, data that is typically unstructured is automatically processed in such a way  that its inherent structures become visible and usable. For example, grammatical patterns in long texts can reveal information about the text type. Text-Data Analytics frequently uses methods from supervised text classification, in which relevant information (for example, the distribution of parts of speech in a sentence) is extracted to classify documents, or parts of documents, according to content-related characteristics (for example, the text type). The extracted information is used to train a model on the basis of pre-classified data. This model can then classify further documents fully automatically. Besides supervised methods, semi-supervised and unsupervised methods (i.e. those without a predetermined classification space) are used as well. Application areas include social networks and digital libraries.

Adaptive Language Processing researches methods for information preparation which adapt to different genres, domains and user groups. Flexible preparation, processing and presentation of linguistic data plays an important role in many areas. This is because, on the one hand, linguistic input can be extremely heterogeneous (for example, various language registers on the Web), and on the other hand, the users of linguistic output have different requests and requirements (for example, a language learner requires different automated feedback than a journalist). Cross-lingual or language-independent processing of language signals is becoming increasingly important in a globalized world.

Digital Humanities Algorithmics includes novel algorithmic approaches for the analysis of textual sources, which are implemented in end-to-end Web-based systems for users in the humanities and social science in order to answer new research questions of theirs. Because data from the humanities and social science domains, as opposed to data from the Web, is relatively small in size, its automated processing requires innovative and robust methods. In addition, experts must be given the opportunity to intervene in automated processes (interactively) in order to exploratively discover new patterns by means of appropriate visualization. Furthermore, intelligent methods can learn through interaction with the expert, just as the expert can gain a better understanding of the automated system.

Dean´s Office

TU Darmstadt
S2|02 Piloty-Building, Room D103
Hochschulstraße 10
64289 Darmstadt


A A A | Drucken Print | Impressum Impressum | Sitemap Sitemap | Suche Search | Kontakt Contact | Website Analysis: More Information
zum Seitenanfangzum Seitenanfang