Bach-Hiep Tran

Improvements in beam search for 10000-word continuous speech recognition

The author describes the improvements in a time synchronous beam search strategy for a 10000-word continuous speech recognition task. The improvements are based on two measures: a tree-organization of the pronunciation lexicon and a novel look-ahead technique at the phoneme level, both of which interact directly with the detailed search at the state levels of the phoneme models. Experimental tests were performed for four speakers on a 12306-word task. As a result of the above measures, the overall search effort was reduced by a factor of 17 without a loss in recognition accuracy.<<ETX>>

The Philips Research system for continuous-speech recognition

This paper gives an overview of the Philips Research system for continuous-speech recognition. The recognition architecture is based on an integrated statistical approach. The system has been successfully applied to various tasks in American English and German, ranging from small vocabulary tasks to very large vocabulary tasks and from recognition only to speech understanding. Here, we concentrate on phoneme-based continuous-speech recognition for large vocabulary recognition as used for dictation, which covers a significant part of our research work on speech recognition. We describe this task and report on experimental results. In order to allow a comparison with the performance of other systems, a section with an evaluation on the standard North American Business news (NAB2) task (dictation of American English newspaper text) is supplied.

AN OVERVIEW OF THE PHILIPS RESEARCH SYSTEM FOR LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION

This paper gives an overview of a research system for phoneme based, large vocabulary continuous speech recognition. The system to be described has been applied to the SPICOS task, the DARPA RM task and a 12000 word dictation task. Experimental results for these three tasks will be presented. Like many other systems, the recognition architecture is based on an integrated statistical approach. In this paper, we describe the characteristic features of the system as opposed to other systems: (1) The Viterbi criterion is consistently applied both in training and testing. (2) Continuous mixture densities are used without any tying or smoothing; this approach can be viewed as a sort of ‘statistical template matching’. (3) Time-synchronous beam search is used consistently throughout all tasks; extensions using a tree organization of the vocabulary and phoneme lookahead are presented so that a 12000 word task can be handled.

Method of and apparatus for deriving a plurality of sequences of words from a speech signal

The determination of a plurality of sequences of words from a speech signal with a decreasing probability of correspondence utilizes the best word sequence as a basis and as further word sequences there are determined only those which enclose a part of the best word sequence, that is to say the remainder of these word sequences. To this end, the recognition involves first the formation of a word graph and the best word sequence is separately stored as a tree which initially has one branch only. The word boundaries of this word sequence form nodes in this tree. Because only nodes of this tree have to be taken into account for the next-best word sequences, the calculation is substantially simpler than if the complete word graph were first completely expanded in the form of a tree and completely searched again for each new word sequence.

Method of recognizing a sequence of words and device for carrying out the method

When a language model is to be used for the recognition of a speech signal and the vocabulary is composed as a tree, the language model value cannot be taken into account before the word end. Customarily, after each word end the comparison with a tree root is started anew, be it with a score which has been increased by the language model value so that the threshold value for the scores at which hypotheses are terminated must be high and hence many, even unattractive hypotheses remain active for a prolonged period of time. In order to avoid this, in accordance with the invention a correction value is added to the score for at least a part of the nodes of the vocabulary tree; the sum of the correction values on the path to a word then may not be greater than the language model value for the relevant word. As a result, for each test signal the scores of all hypotheses are of a comparable order of magnitude. When a word end is reached, the sum of all added correction values is subtracted and the correct language model value is added.

Continuous speech dictation: from theory to practice

This paper gives an overview of the Philips research system for phoneme-based, large-vocabulary, continuousspeech recognition. The system has been successfully applied to various tasks in the German and (American) English languages, ranging from small vocabulary tasks to very large vocabulary tasks. Here, we concentrate on continuousspeech recognition for dictation in real applications, the dictation of legal reports and radiology reports in German. We describe this task and report on experimental results. We also describe a commercial PC-based dictation system which includes a PC implementation of our scientific recognition prototype. In order to allow for a comparison with the performance of other systems, a section with an evaluation on the standard Wall Street Journal task (dictation of American English newspaper text) is supplied. The recognition architecture is based on an integrated statistical approach. We describe the characteristic features of the system as opposed to other systems: 1. the Viterbi criterion is consistently applied both in training and testing; 2. continuous mixture densities are used without tying or smoothing; 3. time-synchronous beam search in connection with a phoneme look-ahead is applied to a tree-organized lexicon.