Research on Question Answering is focused mainly on classifying the question type and finding the answer, while presenting the answer in a way that suits the user’s needs has received little attention. This paper shows how existing question answering systems can be improved by exploiting Rhetorical Structure Theory-based summarization techniques in order to extract more than just the exact answer from the document in which the answer resides. The output is an extensive answer, which also provides additional information related to the question, and which may give the user an opportunity to assess the accuracy of the answer (is this what I am looking for?). A first experiment confirms that the proposed summarization method performs better than a baseline summarization method. 1

which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Background: The practice of evidence-based medicine (EBM) requires clinicians to integrate their expertise with the latest scientific research. But this is becoming increasingly difficult with the growing numbers of published articles. There is a clear need for better tools to improve clinician's ability to search the primary literature. Randomized clinical trials (RCTs) are the most reliable source of evidence documenting the efficacy of treatment options. This paper describes the retrieval of key sentences from abstracts of RCTs as a step towards helping users find relevant facts about the experimental design of clinical studies. Method: Using Conditional Random Fields (CRFs), a popular and successful method for natural language processing problems, sentences referring to Intervention, Participants and Outcome Measures are automatically categorized. This is done by extending a previous approach for labeling sentences in an abstract for general categories associated with scientific argumentation or rhetorical roles: Aim, Method, Results and Conclusion. Methods are tested on several corpora of RCT abstracts. First structured abstracts with headings specifically indicating Intervention, Participant

We present a system that automatically identifies Attribution, an intrasentential relation in the RST Treebank. The system uses uses syntactic information from Penn Treebank parse trees. It identifies Attributions as structures in which a verb takes an SBAR complement, and achieves a f-score of.92. This supports our claim that the Attribution relation should be eliminated from a discourse treebank, since it represents information that is already present in the Penn Treebank, in a different form. More generally, we suggest that intra-sentential relations in the RST Treebank might all be eliminable in this way. 1

The final copy of this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable presentation standards of scholarly work in the above mentioned discipline. Bethard, Steven John (Ph.D., Computer Science)

This thesis focuses on computational discourse models for collaboratively edited corpora. Due to the exponential growth rate and significant stylistic and content variations of collaboratively edited corpora, models based on professionally edited texts are incapable of processing the new data effectively. For these methods to succeed, one challenge is to preserve the local coherence as well as global consistence. We explore two corpus-based methods for processing collaboratively edited corpora, which effectively model and optimize the consistence of user generated text. The first method addresses the task of inserting new information into existing texts. In particular, we wish to determine the best location in a text for a given piece of new information. We present an online ranking model which exploits this hierarchical structure – representationally in its features and algorithmically in its learning procedure. When tested on a corpus of Wikipedia articles, our hierarchically informed model predicts the correct insertion paragraph more accurately than baseline methods. The second method concerns inducing a common structure across multiple articles in similar domains to

Using the RST annotated corpus [Carlson et al., 2003], we use simple statistics on the distribution of discourse markers or cue phrases as evidence of the three-way distinction of Contrast relations, Contrast, Antithesis and Concession, recognized in standard Rhetorical Structure Theory (RST, Mann and Thompson 1987). We also show that however, an intuitive marker of Contrast, is not actually used statistically signi cantly more often in Contrast relations than in Cause-E ect relations. These results highlight the need for empirically based discourse marker identi cation rather than the intuitive method that is the current norm. 1