A topic independent lexical and language modeling for robust key-phrase detection and verification is presented. Instead of assuming a domain specific lexicon and language model, our model is designed to characterize filler phrases depending on the speaking-style, thus can be trained with large corpora of different topics but the same style. Mutual information criterion is used to select topic independent filler words and their N-gram model is used for verification of key-phrase hypotheses. A dialogue-style dependent filler model improves the key-phrase detection in different dialogue applications. A lecture-style dependent model is trained with transcriptions of various oral presentations by filtering out topic specific words. It performs much better verification of key-phrases uttered during lectures of different topics compared with the conventional syllable-based model and large vocabulary model.

We propose an e ective application of speech recognition to foreign language pronunciation learning. The objective of our system is to detect pronunciation errors and provide diagnostic feedback through speech processing and recognition methods. Automatic pronunciation error detection is used for two kinds of mispronunciation, that is mistake and linguistical inheritance. The correlation between automatic detection and human judgement shows its reliability. For the feedback guidance to an erroneous phone, we set up classi ers for the well-recognized articulatory features, the place of articulation and the manner of articulation, in order to identify the cause of incorrect articulation. It provides feedback guidance on how to correct mispronunciation.

We propose an e cient two-pass search algorithm for LVCSR. Instead of conventional word graph, the rst preliminary pass generates \word trellis index, keeping track of all survived word hypotheses within the beam every time-frame. As it represents all found word boundaries non-deterministically, we can (1) obtain accurate sentence-dependent hypotheses on the second search, and (2) avoid expensive word-pair approximation on the rst pass. The second pass performs an e cient stack decoding search, where the index is referred to as predicted word list and heuristics. Experimental results on 5,000-word Japanese dictation task show that, compared with the word-graph method, this trellis-based method runs with less than 1/10 memory cost while keeping high accuracy. Finally, by handling inter-word context dependency, we achieved the word error rate of 5.6%.

In order to realize robust speech understanding, we present a phrase spotting approach. The use of concept-based phrases as the core unit for understanding is advantageous because the phrase-level constraint realizes wider coverage and stable matching and they are directly mapped to semantic cases. The phrase spotting and the sentence-level parsing are formulated as a progressive search. It realizes an optimal search to spot phrase candidates with Viterbi scoring and A* search to combine the phrase candidates into optimal sentence hypotheses. The approach achieved higher detection rates and robust interpretation of ill-formed utterances. We also examined the effect of the background language model. It is shown that lexical knowledge in the background is vital for spotting and the use of the acoustic score of the filler model is significant for parsing.

We present a new model of speech understanding, based on the cooperation of the speech recognizer and language analyzer, which interacts with the knowledge sources while keeping its modularity. The semantic analyzer is realized with a semantic network that represents the possible concepts in a task. The speech recognizer based on an LR parser interacts with the semantic analyzer to eliminate invalid hypotheses at an early stage. The coupling of a loose grammar and interactive semantic analysis accepts ill-formed sentences while filtering out non-sense ones, thus realizes robust understanding. Dialog-level knowledge is also incorporated to constrain both the syntactic and the semantic knowledge sources. The key to guide the search efficiently is powerful heuristics. The relationship between the heuristic power and search efficiency is examined experimentally. The stochastic word bigram is derived from the probabilistic LR grammar as A*-admissible heuristics.>

A novel hidden Markov model (HMM) architecture which realizes both high discriminating ability and stochastic scoring is presented. In modifying continuous HMM so that the states of the models are best separated, distinctive features vary for different states or different models, and different discriminant functions should be made for different competing states. For every pair of the states, a Bayes classifier which performs a vector transformation based on discriminant analysis is constructed. Each classifier ranks the two states and computes a relative value of the probabilities. Output probabilities of the HMM states are obtained by combining and normalizing the results of the pair-wise classifications. Training of the classifiers and HMMs is done interactively and iteratively so that they are optimized totally. Experimental results show that the method, called the pair-wise Bayes classifier-HMM (PWBC-HMM) is more effective than the conventional HMM. It realizes robust recognition by modifying pattern space to fully separate confusing classes, while retaining analog outputs by statistical Bayes classifiers.>

The authors present a novel phoneme recognition method which combines two stochastic methods; discriminant analysis and the hidden Markov model (HMM) method. The HMM is powerful in time-warping and in capturing the global dynamic features, but its discriminating ability is not sufficient. The approach used is to construct the HMM with a phonetic element classifier front-end. Each phonetic element belongs to one phoneme and represents a local pattern of the phoneme. The classifier is a modified version of discriminant analysis, that is, a combination of Bayes classifiers. It extracts optimal features to separate the phonetic elements and consequently contributes to separate HMMs of the phonemes. Furthermore, the score of the classifier is combined with that of the HMM. Since the classifier is based on a statistical method, combination of the scores is straightforward both in theory and in practice. Experimental results showed that the combined method is more effective than the conventional VQ (vector quantization) HMM and that utilizing the score of the classifier on the local features is significant.>

Place and/or manner of articulation of the 7 consonants (/?,p,t,k,b,d,g/) are discriminated in an enviroment of /a,i,u, e,o/. We considered that isolated vowels are approximately characteriszed as the glottal stop /?/ and included them as stop consonants. The feature vector used was spectrum of the initial 50ms after vowel voicing or stop burst point. The burst point of time was manually specified referring to waveform display. We used a stepwise discriminant analysis program with variable selection to classify the samples of speech. The sample set comprised 3402 samples from 89 male speakers. Some invariance was found in the discrimination between /?,p,t,k/ achieving 89% correct score across vowel context and speakers. This is a 3% degradation from that between /p,t,k/. A few samples of voiced stops lacking murmur preceding the burst point was well discriminated from voiceless stops of same point of articulation.

This paper proposes a method for organizing linguistic knowledge in both systematic and flexible fashion. We introduce a purely applicative language (PAL) as an intermediate representation and an object-oriented computation mechanism for its interpretation. PAL enables the establishment of a principled and well-constrained method of interaction among lexicon-oriented linguistic modules. The object-oriented computation mechanism provides a flexible means of abstracting modules and sharing common knowledge.

This paper describes an analysis method which uses heuristic knowledge to find local syntactic structures of Chinese sentences. We call it a preprocessing, because we use it before we do global syntactic structure analysis (1) of the input sentence. Our purpose is to guide the global analysis through the search space, to avoid unnecessary computation.To realize this, we use a set of special words that appear in commonly used patterns in Chinese. We call them "characteristic words". They enable us to pick out fragments that might figure in the syntactic structure of the sentence. Knowledge concerning the use of characteristic words enables us to rate alternative fragments, according to pattern statistics, fragment length, distance between characteristic words, and so on. The preprocessing system proposes to the global analysis level a most "likely" partial structure. In case this choice is rejected, backtracking looks for a second choice, and so on.For our system, we use 200 characteristic words. Their rules are written by 101 automata. We tested them against 120 sentences taken from a Chinese physics text book. For this limited set, correct partial structures were proposed as first choice for 94% of sentences. Allowing a 2nd choice, the score is 98%, with a 3rd choice, the score is 100%.

This paper describes the method and the system investigated to solve the problem encountered in the automatic recognition of speech sound. From research in the automatic analyzer of speech sound, a monosyllable recognition system was constructed in which the phoneme is used as the basic recognition unit. Recently this system has been developed to accept the conversational speech sound with unlimited vocabulary. The mechanical recognition of conversational speech sound requires two basic operations. One is the segmentation of the continuous speech sound into several discrete intervals (or segments), each of which may be thought to correspond to a phoneme, and the other is the pattern recognition of such segments. For segmentation, by defining two criteria, ``stability'' and ``distance,'' the properties of the time pattern obtained by the analysis of input speech sound may be examined. The principle of the recognition is based on the mechanism of the articulation in our speech organ. Corresponding to this, the machine has the functions called phoneme classification, vowel analysis and consonant analysis. A conversational speech recognition system with the phonetic contextual approach is also applied to the vowel recognition where the time pattern of input speech is matched with the stored standard patterns in which the phonetic contextual effects are taken into consideration. The time pattern which has great variety may be effectively expressed by the new representation of ``sequential pattern'' and ``weighting pattern.''