ASR Technology Research Articles

The impact of automatic speech recognition technology on second language pronunciation and speaking skills of EFL learners: a mixed methods investigation.

This study employed an explanatory sequential design to examine the impact of utilizing automatic speech recognition technology (ASR) with peer correction on the improvement of second language (L2) pronunciation and speaking skills among English as a Foreign Language (EFL) learners. The aim was to assess whether this approach could be an effective tool for enhancing L2 pronunciation and speaking abilities in comparison to traditional teacher-led feedback and instruction. A total of 61 intermediate-level Chinese EFL learners were randomly assigned to either a control group (CG) or an experimental group (EG). The CG received conventional teacher-led feedback and instruction, while the EG used ASR technology with peer correction. Data collection involved read-aloud tasks, spontaneous conversations, and IELTS speaking tests to evaluate L2 pronunciation and speaking skills. Additionally, semi-structured interviews were conducted with a subset of the participants to explore their perceptions of the ASR technology and its impact on their language learning experience. The quantitative analysis of the collected data demonstrated that the EG outperformed the CG in all measures of L2 pronunciation, including accentedness and comprehensibility. Furthermore, the EG exhibited significant improvements in global speaking skill compared to the CG. The qualitative analysis of the interviews revealed that the majority of the participants in the EG found the ASR technology to be beneficial in enhancing their L2 pronunciation and speaking abilities. The results of this study suggest that the utilization of ASR technology with peer correction can be a potent approach in enhancing L2 pronunciation and speaking skills among EFL learners. The improved performance of the EG compared to the CG in pronunciation and speaking tasks demonstrates the potential of incorporating ASR technology into language learning environments. Additionally, the positive feedback from the participants in the EG underscores the value of using ASR technology as a supportive tool in language learning classrooms.

Exploring the Effects of ASR Technology on Improving English Education Majors’ Oral Accuracy

Exploring the Effects of ASR Technology on Improving English Education Majors’ Oral Accuracy

Application of Wearable Computer and ASR Technology in an Underground Mine to Support Mine Supervision of the Heavy Machinery Chamber.

Systems that use automatic speech recognition in industry are becoming more and more popular. They bring benefits especially in cases when the user’s hands are often busy or the environment does not allow the use of a keyboard. However, the accuracy of algorithms is still a big challenge. The article describes the attempt to use ASR in the underground mining industry as an improvement in the records of work in the heavy machinery chamber by a foreman. Particular attention was paid to the factors that in this case will have a negative impact on speech recognition: the influence of the environment, specialized mining vocabulary, and the learning curve. First, the foreman’s workflow and documentation were recognized. This allowed for the selection of functionalities that should be included in the application. A dictionary of specialized mining vocabulary and a source database were developed which, in combination with the string matching algorithms, aim to improve correct speech recognition. Text mining analysis, machine learning methods were used to create functionalities that provide assistance in registering information. Finally, the prototype of the application was tested in the mining environment and the accuracy of the results were presented.

New approach to heat treatment of spring steel

In the 54SiCr6 spring steel, a spheroidized structure was achieved by two processing methods: conventional soft annealing above the temperature A1 in a furnace and accelerated ASR annealing by an induction heating equipment. The ASR technology is much faster, even for materials with a worse tendency to the spheroidisation, such as steels with higher silicon content. Thanks to this, a finer structure is obtained. This is the basis of other advantages for the subsequent hardening. Quenching can be done at lower temperatures, higher yield and tensile strength is obtained while maintaining same or improved ductility and contraction. This also results in influencing of fatigue behavior.

Application of Automatic Speech Recognition (ASR) Algorithm in Smart Home

Study on practical ASR system for smart home had significant meanings for the development of the smart home. Through analysis of embedded ASR technology and the control technology of smart home, people could utilize VS1003 to record the speech for audio decoder chip by taking the NL6621 template as the platform. And, the ASR system on smart home could be realized by utilizing the Hidden Markov Model (HMM) algorithm to conduct speech model training and speech matching. Test showed that, the ASR system has relatively high recognition rate and real-time property.

Support vector machine-based stuttering dysfluency classification using GMM supervectors

It is generally acknowledged that recognition and classification of dysfluencies are an important criterion in the objective and accurate assessment of stuttered speech. For this reason, there is a growing interest in the application of Automatic Speech Recognition ASR technology to automate the dysfluency recognition. In this perspective, several studies have been carried out on the classification of dysfluencies by means of acoustic analysis, parametric and non-parametric feature extraction and statistical methods. This work is focused on introducing and evaluating Support Vector Machine SVM based dysfluency recognition system using a Gaussian Mixture Model GMM supervector. The experimental evaluation of the proposed system reveals that an SVM-based GMM supervector is effective for dysfluency classification. We have obtained substantial improvements in the performance by considering cepstral and their delta features.

Using ASR Technology in Language Training for Specific Purposes: A Perspective from Quebec, Canada

For many patients throughout the world, access to healthcare depends on the patients’ and healthcare providers’ ability to communicate efficiently in each other’s language. One way to reduce linguistic barriers to healthcare access is to increase the number of linguistically and culturally competent healthcare professionals. Conspicuously absent in the literature on second language (L2) training of healthcare professionals, however, is the use of technology that combines meaningful interaction, feedback, simulation, and asynchronous access. The goal of this paper is to fill this gap by describing and evaluating the “Virtual Language Patient,” a computer-based L2 training module for healthcare professionals. The module employs automatic speech recognition technology, pronunciation assessment, and video clips of a simulated medical history interview with a minority language patient. Five nurses-in-training at a French-language nursing college in Quebec reported that the module was easy to operate and that it addressed their anticipated language learning needs. More importantly, analysis of the data file automatically generated by the module revealed improvements in acceptability of the nurses’ pronunciation of the medical interview questions. These findings suggest that the module can be effective in language training for healthcare professionals. Implications for the improvement of virtual dialogue systems are discussed.

Oral proficiency training in Dutch L2: The contribution of ASR-based corrective feedback

In this paper, we introduce a system for providing automatically generated corrective feedback on pronunciation errors in Dutch, Dutch-CAPT. We describe the architecture of the system paying particular attention to the rationale behind it, to the performance of the error detection algorithm and its relationship to the effectiveness of the corrective feedback provided. It appears that although the system does not achieve 100% accuracy in error detection, learners enjoy using it and the feedback provided is still effective in improving pronunciation errors after only a few hours of use over a period of one month. We discuss which factors may have led to these positive results and argue that it is worthwhile studying how ASR technology could be applied to the training of other speaking skills.

Decoding speech in the presence of other sources

The statistical theory of speech recognition introduced several decades ago has brought about low word error rates for clean speech. However, it has been less successful in noisy conditions. Since extraneous acoustic sources are present in virtually all everyday speech communication conditions, the failure of the speech recognition model to take noise into account is perhaps the most serious obstacle to the application of ASR technology. Approaches to noise-robust speech recognition have traditionally taken one of two forms. One set of techniques attempts to estimate the noise and remove its effects from the target speech. While noise estimation can work in low-to-moderate levels of slowly varying noise, it fails completely in louder or more variable conditions. A second approach utilises noise models and attempts to decode speech taking into account their presence. Again, model-based techniques can work for simple noises, but they are computationally complex under realistic conditions and require models for all sources present in the signal. In this paper, we propose a statistical theory of speech recognition in the presence of other acoustic sources. Unlike earlier model-based approaches, our framework makes no assumptions about the noise background, although it can exploit such information if it is available. It does not require models for background sources, or an estimate of their number. The new approach extends statistical ASR by introducing a segregation model in addition to the conventional acoustic and language models. While the conventional statistical ASR problem is to find the most likely sequence of speech models which generated a given observation sequence, the new approach additionally determines the most likely set of signal fragments which make up the speech signal. Although the framework is completely general, we provide one interpretation of the segregation model based on missing-data theory. We derive an efficient HMM decoder, which searches both across subword state and across alternative segregations of the signal between target and interference. We call this modified system the speech fragment decoder. The value of the speech fragment decoder approach has been verified through experiments on small-vocabulary tasks in high-noise conditions. For instance, in a noise-corrupted connected digit task, the new approach decreases the word error rate in the condition of factory noise at 5 dB SNR from over 59% for a standard ASR system to less than 22%.

A Comparison of Dialog Strategies for Call Routing

Advances in commercially-available ASR technology have enabled the deployment of “How-may-I-help-you?” interactions to automate call routing. While often preferred to menu-based or directed dialog strategies, there is little quantitative research into the relationships among prompt style, task completion, user preference/satisfaction, and domain. This work applies several dialog strategies to two domains, drawing on both real callers and usability subjects. We find that longer greetings produce higher levels of first-utterance routability. Further, we show that a menu-based dialog strategy produces a uniformly high level of routability at the first utterance in two domains, whereas an open-dialog approach varies significantly with domain. In a domain where users lack an expectation of task structure, users are most successful with a directed strategy, for which preference scores are highest, even though it does not result in the shortest dialogs. Callers rarely provide more than one piece of information in their responses to all types of dialog strategies. Finally, a structured dialog repair prompt is most helpful to callers who were greeted with an open prompt, and least helpful to callers who were greeted with a structured prompt.

Is ASR ready for wireless primetime: Measuring the core technology for selected applications

It is estimated that by the end of 2001 as many as 500 million people worldwide will use cellular services. The nature of hands-busy and eyes-busy situations inherent in the anywhere and anytime wireless communication paradigm presents exciting marketing opportunities and, at the same time, unique technical challenges to the current-generation ASR technology and their new applications. Current industry trends clearly show that incorporating ASR technology into existing or new wireless services as a replacement for touch-tone input is a natural progression in user interface. But is the current-generation ASR technology ready for prime time over wireless channels? Both qualitative and quantitative assessments for the core technology must be adopted by the industry before answering this question. In this paper, we will describe a set of benchmark tasks designed to evaluate the state-of-the-art ASR technologies from a wireless perspective and present the results of these benchmark tests on two commercially available software-based ASR systems that represent the best core ASR technology on the market.

Modelling of phone duration (using the TIMIT database) and its potential benefit for ASR

As indicated by Bourlard et al. (1996), the best and simplest solution so far in standard ASR technology to implement durational knowledge, seems to consist of imposing a (trained) minimum segment duration, simply by duplicating or adding states that cannot be skipped. We want to argue that recognition performance can be further improved by incorporating “specific knowledge” (such as duration and pitch) into the recognizer. This can be achieved by optimising the probabilistic acoustic and language models, and probably also by a postprocessing step that is fully based on this specific knowledge. The widely available, hand-segmented, TIMIT database was used by us to extract duration regularities, that persist despite the great speaker variability. Two main approaches were used. In the first approach, duration distributions are considered for single phones, as well as for various broader classes, such as those specified by long or short vowels, word stress, syllable position within the word and within an utterance, post-vocalic consonants, and utterance speaking rate. The other approach is to use a hierarchically structured analysis of variance to study the numerical contributions of 11 different factors to the variation in duration. Several systematic effects have been found, but several other effects appeared to be obscured by the inherent variability in this speech material. Whether this specific use of knowledge about duration in a post-processor will actually improve recognition performance still has to be shown. However, in line with the prophetic message of Bourlard et al.'s paper, we here consider the improvement of performance as of secondary importance.

The role of automated speech recognition in endoscopic data collection.

Speech recognition technology has developed substantially in the past half decade. Currently, large vocabulary, speaker independent, discrete recognizers are the state-of-the-art. This will change. Moderate sized, continuous recognition systems now exist in research settings. However, it is unlikely that such systems will be widely available until the mid to late 1990's. The accuracy rates of current speech recognition systems are high. Consequently, speech accuracy is not the current limiting aspect of using ASR. The limiting aspect of using ASR technology is the approach to integrating speech functionality into applications. One approach is to use ATNs as models of natural language to support both an input strategy and a text generation system. ATNs provide approaches to both syntactical correctness and semantic richness. This is an approach which plays to the strengths of the discrete nature of current speech technology and also provides a methodology for the capture and archiving of highly detailed information. The ATN approach avoids the natural language parsing problem created by a fully free form dictation interface. Evolving along with the underlying speech technology are standards in the definitions and criteria used in endoscopic practice. There are clear benefits from standards in this area. However, it is likely that this will also take several years and may never yield a universally accepted lexicon. Furthermore, there will be user interface barriers to surmount in any system attempting to use speech as an input modality. Because of the relatively large vocabularies used in medical discourse, the user interface will need to be carefully crafted.(ABSTRACT TRUNCATED AT 250 WORDS)