Interest in automatic summarization continues to grow, motivated by the explosion in access to on-line information, together with advances in natural language processing and information retrieval. In fact, some form of automatic summarization may be indispensable given the massive information universes that will be explored in the 21st century. I will begin with a brief introduction to the field, and then move on to two two recent trends. The first is located within the paradigm of summarization as sentence extraction. I discuss several experimental results in learning sentence extraction rules from a corpus, along with some remaining challenges. I then move on to the problem of synthesizing summaries from multiple documents, for example, a biographical sketch of a person assembled from a document collection. The focus here is on narrative summarization, and I describe an approach that allows multimedia narratives to be generated from high-level outlines. A brief demo will also be presented.
For the colloquium series schedule, see the UMD Computational Linguistics Colloquium Series web page at http://umiacs.umd.edu/~resnik/cl_colloquium/. If you are interested in meeting with the speaker, please contact Philip Resnik (resnik@umiacs.umd.edu).