Dragon Systems Resource Management Benchmark Results -February 1991

In this paper, we present preliminary results obtained at Dragon Systems on the Resource Management benchmark task. The basic conceptual units of our system are Phonemes-in-context PICs, which are represented as Hidden Markov Models, each of which is expressed as a sequence of Phonetic Elements PEIs. The PEIs corresponding to a given phoneme constitute a kind of alphabet for the representation of PICs. For the speaker-dependent tests, two basic methods of training the acoustic models were investigated. The first method of training the Resource Management models is to re-estimate the models for each test speaker from that speakers training data, keeping the PEL spellings of the PICs fixed. The second approach is to use the re-estimated models from the first method to derive a segmentation of the training data, then to respell the PICs in a largely speaker-dependent manner in order to improve the representation of speaker differences. A full explanation of these methods is given, as are results using each method. In addition to reporting on two different training strategies, we discuss N-Best results. The N-Best algorithm is a modification of the algorithm proposed by Soong and Huang at the June 1990 workshop. This algorithm runs as a post-processing step and uses an A-search an algorithm also known as a stack decoder.