Ingels, P. (1996). Layered HMMs in Robust Natural Language Text Processing. Technical Report LiTH-IDA-R-96-11, Department of Computer and Information Science, Linköping University, Sweden. (bibtex),
Abstract: This paper describes and experimentally evaluates the application of speech recognition methods to the problem of processing distorted natural language text in a robust manner. The method implements an advanced tokenizer that can detect and correct spelling mistakes and segmentation errors in the input stream and it does so in the context of higher level linguistic knowledge. The idea is to arrange Hidden Markov Models (HMM) in multiple layers where the HMMs in each layer are responsible for different aspects of the processing of the input.
CS Dept TR Overview