1. Title of Database Automatic transcription of read out document subset of Reuters-21578 2. Data Type Text 3. Abstract This dataset is created by reading out 200 files from the 10 largest Reuters classes and using an Automatic Speech Recognition system to create corresponding transcriptions. 4. Sources (a) Original owners of database (name/snail address/phone/email/homepage) Shourya Roy shourya.roy 'at' gmail.com and Shantanu Godbole shantanu 'at' godbole.net 5. Data Characteristics This data was created by selecting 20 files each from the 10 largest classes in the Reuters-21578 collection (http://www.daviddlewis.com/resources/testcollections/reuters21578/). The files were read out by 3 Indian speakers and an Automatic Speech Recognition (ASR) system was used to generate the transcripts. More about the ASR system can be found in [1]. Such a dataset will be really helpful to study the effect of speech recognition noise on text mining algorithms. The first work which refered to this dataset was on noisy text classification[2]. 7. Data Format There are 10 directories labeled by the topic name. Each contains 20 files of transcriptions. References [1] L. R. Bahl, S. Balakrishnan-Aiyer, J. Bellegarda, M. Franz, P. Gopalakrishnan, D. Nahamoo, M. Novak, M. Padmanabhan, M. Picheny, and S. Roukos, Performance of the IBM large vocabulary continuous speech recognition system on the ARPA wall street journal task. In Proc. of ICASSP ’95, pages 41–44, Detroit, MI, 1995. [2] S. Agarwal, S. Godbole, D. Punjani and S. Roy, How Much Noise is too Much: A Study in Automatic Text Classification", In Proc. of ICDM 2007