The Lincoln Continuous Speech Recognition System: Recent Developments and Results

THE LINCOLN CONTINUOUS SPEECH RECOGNITION SYSTEM: RECENT DEVELOPMENTS AND RESULTS 1 Douglas B. Paul Lincoln Laboratory, MIT Lexington, MA 02173 ABSTRACT The Lincoln stress-resistant HMM CSR has been extended to large vocabulary continuous speech for both speaker-dependent (SD) and speaker-independent (SI) tasks. Performance on the DARPA Resource Manage- ment task (991 word vocabulary, perplexity 60 word-pair grammar) [1] is 3.4% word error rate for SD training of word-context-dependent triphone models and 12.6% word error rate for SI training of (word-context-free) tied mixture triphone models. INTRODUCTION Our earlier development efforts [2,3,4,5,6,7,8,9] centered on improving the SD speaker-stress robustness for both IWR and CSR tasks. Since our IWR database included a normal speech test section, we were able to determine that our enhancements for robustness also improved performance for normally spoken speech (0 errors/1680 test tokens, 105 word vocabulary, multi-style training). An independent test on the TI-20 word database [10] confirmed this normal speech performance with 3 errors out of 5120 test tokens on our first run on this database and no errors after a small amount of development [3]. Our robust CSR database was not useful for determining the large vocabulary normal speech performance. In order to work on a large vocabulary normal speech CSR task, we switched to the DARPA Resource Management database [1]. The SD portion of this database has 12 speakers with 600 training sentences and 100 development test sentences per speaker. This provided a total of 1,200 test sentences containing 10,242 words. For SI work we used the same development test sentences, but trained on 2,880 sentences from 72 speakers from the SI training portion of the database. (There was an overlap of 8 speakers between the SI and SD training sets making the total of 80 speakers reported in [1].) When additional SI training data was needed, we added the designated "SI development test" data, again avoiding test speaker overlaps, to the designated SI training data. This provided a total 3,990 training sentences from 109 speakers. The vocabulary of the Resource Management database is 991 words. There is also an "official" word-pair recognition grammar [11]. This grammar is just a list of allowable word pairs without probabilities for the purpose of reducing the recognition perplexity to about 60. (Including the probabilities slightly more than halves the perplexity.) Working with a single development test set carries a risk of tuning one's system to idiosyncrasies in the test set due to the multiple tests and decisions performed during algorithm development. Methodologies which focus on correcting the individual test set errors are particularly subject to this problem and become, in effect, corrective training [12] on the test set. In contrast, since different test sets have different inherent difficulties, comparisons of two systems using different test sets have a significantly reduced resolution. Therefore, the DARPA program has had several "official evaluation tests", the most recent of which was held in June 88. This test used 25 sentences from each of the 12 speakers (2,546 total words), all of which were …

Content truncated to 500 words