Many natural language processing (NLP) applications make use of lexical semantic knowledge, i.e. meanings of words decoupled from the more complex tasks of compositional sentence analysis and language understanding. The course will represent a mixture of theoretical studies and hands-on research work.
At first, broad-coverage knowledge sources, such as e.g. WordNet for English and GermaNet for German, used in lexical semantic processing will be introduced. We will discuss in detail a set of algorithms for computing semantic relatedness of words. The second half of the course will look at some NLP applications using lexical semantic knowledge, such as spelling correction, information
retrieval, text summarization, essay grading. In particular, we will look at the ways of integrating semantic relatedness of words in NLP applications.

During the course, we will work at a shared task aimed at computing semantic coherence of texts. We will annotate the data with the help of an annotation tool MMAX, investigate the inter-annotator agreement, define the approaches to quantify semantic coherence and work on those in several teams. In the end, we will evaluate the results against the annotation and discuss them.

The plan of the seminar including the slides is published here.