Invited talk by Advaith Siddharthan:
"Is Automatic Text Simplification Useful?"
Abstract: In this talk I'll provide a personal perspective on evaluating Automatic Text Simplification (ATS) systems, particularly when I have made the leap from asking "Does my ATS system produce better output than something else?" to asking "Is my ATS system useful to anyone?" When asking the first question, I have found evaluations relatively straightforward to perform, with clear results that are easy to interpret and therefore publish. When asking the second question, I have encountered often fiddly methodological issues, and my experiments to date have thrown up a mixed bag of results that I haven't always anticipated. There is now enough data from such experiments that some patterns have begun to emerge. I will discuss some of these and ask what they mean for Quality Assessment for Text Simplification.
Bio: Dr Advaith Siddharthan is a Senior Lecturer at the University of Aberdeen, with interests in Text Simplification and Summarisation and Natural Language Generation (NLG). He completed a PhD on text simplification at Cambridge in 2003, and worked on projects at Cambridge and Columbia before joining the NLG group at Aberdeen in 2009, He is one of the founding organisers of the workshop series on "Predicting and Improving Text Readability (PITR)" and has also held numerous grants related to Text Simplification, NLG and Text Analytics.