Deep Learning is a branch of Machine Learning that uses Deep Artificial Neural Networks for modeling problems. While the field is as old as the field of Machine Learning itself, it has experienced a tremendous revival in recent years, with a large portion of top publications devoted to the different facets of Neural Networks and their applications to NLP tasks. At UKP, we use Deep Learning for NLP problems ranging from sequence labeling tasks such as Named Entity Recognition, Event Detection, Metaphor Detection, to Text Classification and Information Retrieval problems. Recently, we have been applying Deep Learning in the form of word embedding features to the very active field of Argumentation Mining.
Argumentation is omnipresent in our daily communication and an important part of each decision making process. The recent research field of Argumentation Mining aims at automatically recognizing argumentation structures in written discourse in order to establish new intelligent systems for facilitating information access, writing skills acquisition and text summarization. This research area includes the following objectives:
- Identifying argument components in different text types
- Recognizing relations between argument components
- Automatic assessment of argumentation quality
Under the heading of Language Technology for Digital Humanities, UKP Lab conducts projects at the boundary between Natural Language Processing, Computer Science on the one hand, and Humanities, Social Sciences, and Educational Research on the other hand. In particular, we work on making digital analysis methods more accessible to text-based humanities, implement tools to explore and annotate text corpora, and contribute to the infrastructures supporting Digital Humanities. Our research interests in this area include:
- Creating user-friendly tools to explore and annotate text corpora
- Analyzing such corpora at the semantic level, e.g. performing opinion mining or identifying metaphoric language
- Processing and analyzing historical texts
- Interoperability with Digital Humanities infrastructures such as DARIAH and CLARIN
Lexical-Semantic Resources and Algorithms is concerned with the analysis, design, and application of lexical-semantic resources (LSRs) for natural language processing. At the core of our work is a multi-year effort of developing the large-scale sense-linked unified resource UBY. UBY contains multiple expert-built and collaboratively constructed LSRs for English and German. Moreover, we are also interested in UBY’s applications to semantic processing tasks, such as Word Sense Disambiguation or Semantic Role Labeling, and in end user applications like Question Answering. Another important topic is utilizing LSRs in the domain of Digital Humanities.
In the Text Mining and Text Analytics research area, we design algorithms to extract information from unstructured text. These algorithms are used in many contexts, e.g. Digital Humanities, educational research, research about the web 2.0, or information retrieval. We particularly focus on innovative automatized approaches to discover structure in textual documents by means of text classification.
The growing text analytics field heavily relies on supervised text classification to offer services such as sentiment analysis, document categorization, or scientific discovery. In a nutshell, supervised text classification extracts relevant information from manually classified documents and learns a model from the extracted information. Machine learning classifiers learn to take decisions autonomously, so that there is no need to programmatically implement rules that are later used to automatically take decisions.
We apply supervised text classification algorithms to complex language processing problems and novel datasets. In such settings, a textual document is typically enhanced with automatic annotations about grammatical and discourse structure, before the information relevant to the given problem is extracted. To reduce the effort of manually creating training data, we are currently also exploring the use of semi-supervised and unsupervised text mining algorithms.
Mastering a language is the result of a long process of learning and practice. Intelligent systems can effectively assist learners in this process by providing learning material, automatic tutoring, and feedback on the learning progress. Our research interests at UKP include:
- Automatic generation and analysis of exercises,
- Contextual linking to learning materials and language resources,
- Modelling text quality and providing feedback on quality issues,
- Automatic assessment of free-text answers and learner proficiency.
We put a special focus on studying written discourse and assisting writing tasks. Recent works deal with the difficulty prediction of exercises, the analysis of argumentation structures, the modeling of writing quality in collaborative texts, and the generation of instant quality feedback. To evaluate the effectiveness of our assistance systems, we conduct empirical user studies with cooperation partners from psychology, didactics, and language learning research.