Fusing ASR Outputs in Joint Training for Speech Emotion Recognition

Until very recently, the generation of punctuation marks for automatic speech recognition (ASR) output has been mostly done by looking at the syntactic structure of the recognized utterances. Prosodic cues such as breaks, speech rate, pitch intonation that influence placing of punctuation marks on speech transcripts have been seldom used. We propose a method that uses recurrent neural networks, taking prosodic and lexical information into account in order to predict punctuation marks for raw ASR output. Our experiments show that an attention mechanism over parallel sequences of prosodic cues aligned with transcribed speech improves accuracy of punctuation generation.

We present a Conformer-based end-to-end neural diarization (EEND) model that uses both acoustic input and features derived from an automatic speech recognition (ASR) model. Two categories of features are explored: features derived directly from ASR output (phones, position-in-word and word boundaries) and features derived from a lexical speaker change detection model, trained by fine-tuning a pretrained BERT model on the ASR output. Three modifications to the Conformer-based EEND architecture are proposed to incorporate the features. First, ASR features are concatenated with acoustic features. Second, we propose a new attention mechanism called contextualized self-attention that utilizes ASR features to build robust speaker representations. Finally, multi-task learning is used to train the model to minimize classification loss for the ASR features along with diarization loss. Experiments on the two-speaker English conversations of Switchboard+SRE data sets show that multi-task learning with position-in-word information is the most effective way of utilizing ASR features, reducing the diarization error rate (DER) by 20% relative to the baseline.

This article concerns comparative studies on the Automatic Speech Recognition (ASR) model incorporated with the Large Language Model (LLM) used for medical interviews. The proposed solution is tested on polish language benchmarks and dataset with medical interviews. The latest ASR technologies are based on convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers. Most of them work as end-to-end solutions. The presented approach in the case of the Whisper model shows a two-stage solution with End-To-End ASR and LLM working together in a pipeline. The ASR output is an input for LLM. The LLM is a component by which the output from ASR is corrected and improved. Comparative studies for automatic recognition of the Polish language between modern End-To-End deep learning architectures and the ASR hybrid model were performed. The medical interview tests were performed with two state-of-the-art ASR models: OpenAI Whisper incorporated with LLM and Scribe ElevenLabs. Additionally, the results were compared with five more end-to-end models (QuartzNet, FastConformer, Wav2Vec 2.0 XLSR and ESPnet Model Zoo) on Mozilla Common Voice and VoxPopuli databases. Tests were conducted for clean audio signal, signal with bandwidth limitation, and degraded. The tested models were evaluated on the basis of Word Error Rate (WER) and Character Error Rate (CER). The results show that the Whisper model performs by far the best among the open-source models. ElevenLabs Scribe model, on the other hand, performs best for Polish on both general benchmark and medical data.

We report insights from translating Spanish conversational telephone speech into English text by cascading an automatic speech recognition (ASR) system with a statistical machine translation (SMT) system. The key new insight is that the informal register of conversational speech is a greater challenge for ASR than for SMT: the BLEU score for translating the reference transcript is 64%, but drops to 32% for translating automatic transcripts, whose word error rate (WER) is 40%. Several strategies are examined to mitigate the impact of ASR errors on the SMT output: (i) providing the ASR lattice, instead of the 1-best output, as input to the SMT system, (ii) training the SMT system on Spanish ASR output paired with English text, instead of Spanish reference transcripts, and (iii) improving the core ASR system. Each leads to consistent and complementary improvements in the SMT output. Compared to translating the 1-best output of an ASR system with 40% WER using an SMT system trained on Spanish reference transcripts, translating the output lattice of a better ASR system with 35% WER using an SMT system trained on ASR output improves BLEU from 32% to 38%.

Reverberation effects dramatically degrade the performance of automatic speech recognition (ASR). In conventional approaches, the microphone array-based beamformers try to filter rout the reverberation and re-build the enhanced speech signals for ASR based on the SSR (signal-to-reverberation) criterion. In such approach, the inconsistent criterions exists between the operation of array-based beamforming and model training of ASR. In this study, we propose a HCRF-driven approach for beamformer in which the acoustic models employ the framework of HCRFs and the adaptation of beamformer is based on the ASR output Both the models and the ASR result are incorporated into the adaptation process to iteratively estimate the parameters of an microphone array. A significant error rate reduction is obtained by the proposed method through the experimental results of RWCP room impulse database.

With the surge of online meetings, it has become more critical than ever to provide high-quality speech audio and live captioning under various noise conditions.However, most monaural speech enhancement (SE) models introduce processing artifacts and thus degrade the performance of downstream tasks, including automatic speech recognition (ASR).This paper proposes a multi-task training framework to make the SE models unharmful to ASR.Because most ASR training samples do not have corresponding clean signal references, we alternately perform two model update steps called SE-step and ASR-step.The SEstep uses clean and noisy signal pairs and a signal-based loss function.The ASR-step applies a pre-trained ASR model to training signals enhanced with the SE model.A cross-entropy loss between the ASR output and reference transcriptions is calculated to update the SE model parameters.Experimental results with realistic large-scale settings using ASR models trained on 75,000-hour data show that the proposed framework improves the word error rate for the SE output by 11.82% with little compromise in the SE quality.Performance analysis is also carried out by changing the ASR model, the data used for the ASR-step, and the schedule of the two update steps.

We are interested in the problem of robust understanding from noisy spontaneous speech input. In goal driven human-machine dialog, utterance classification is a key component of the understanding process to determine the intent of the speaker. We propose a novel algorithm for exploiting ASR word confidence scores for better classification of spoken utterances. Word confidence scores for automatic speech recognition (ASR) provide estimates for word error rates. While previous work has focused on straightforward combination of word confidence scores into Bayesian classifiers, we extend the mathematical formulation for boosting classifiers. This extension of the algorithm allows confidence scores to be exploited from a 1-best ASR output or from word confusion networks (WCNs). We present methods for on-line and off-line score combinations. The results we show are for a large database of utterances collected using the AT&T VoiceTone/sup SM/ spoken dialog system. Our experiments show between 5% and 10% reduction in error (1-precision) for a given recall using WCNs compared to ASR output.

Sequence labeling to detect stuttering events in read speech

Graph-based temporal classification (GTC), a generalized form of the connectionist temporal classification loss, was recently proposed to improve automatic speech recognition (ASR) systems using graph-based supervision. For example, GTC was first used to encode an N-best list of pseudo-label sequences into a graph for semi-supervised learning. In this paper, we propose an extension of GTC to model the posteriors of both labels and label transitions by a neural network, which can be applied to a wider range of tasks. As an example application, we use the extended GTC (GTC-e) for the multi-speaker speech recognition task. The transcriptions and speaker information of multi-speaker speech are represented by a graph, where the speaker information is associated with the transitions and ASR outputs with the nodes. Using GTC-e, multi-speaker ASR modelling becomes very similar to single-speaker ASR modeling, in that tokens by multiple speakers are recognized as a single merged sequence in chronological order. For evaluation, we perform experiments on a simulated multi-speaker speech dataset derived from LibriSpeech, obtaining promising results with performance close to classical benchmarks for the task.

This research introduces a comprehensive strategy to enhance the performance of an existing automatic speech recognition (ASR) model, which has been previously documented in published articles. The study sets out to achieve several objectives. Firstly, it concentrates on updating the ASR model by retraining it with new datasets. This involves integrating samples from the latest Common Voice corpus release and data collected independently via the armspeech.com web application. Another key focus lies in optimizing the ASR model for near-real-time processing, intending to improve its speed and efficiency. The proposed adjustments to the model’s architecture aim to balance accuracy and processing speed, which is essential for applications requiring prompt speech recognition. Furthermore, the research explores the integration of Transformer models into the post-processing pipeline to introduce punctuation and capitalization into the ASR output. This step not only enhances the linguistic quality of transcriptions but also improves their readability and usability. In tandem with these advancements, the research presents a systematic approach to gathering, annotating, and storing datasets specifically tailored for punctuation and capitalization tasks. The methodology outlines the acquisition and organization of a dataset conducive to training Transformer models for these linguistic tasks. This comprehensive approach, which encompasses dataset enrichment, architectural modifications, and post-processing enhancements, aims to elevate the ASR model’s accuracy, speed, and linguistic refinement, with a particular focus on addressing the intricacies of the Armenian language. The research contributes valuable insights into the optimization of ASR systems, tackling both language-specific challenges and broader issues related to linguistic post-processing.

The purpose of this study was to determine the extent to which automatic speech recognition (ASR) systems and human speech output to ASR systems are affected when users perform complex tasks in a noisy, stressful simulated tank environment. A second objective was to determine whether the type of auditory information input (monaural or spatial) affected user task performance, including the output of speech to ASR systems. Subjects performed simultaneous radio communication and tank driving tasks. Independent variables included auditory display configuration, driving speed, and subject type (soldier or civilian). Dependent variables included 1) the proportion of radio channel changes correctly recognized by the ASR; 2) the proportion and types of speech errors in radio channel changes; 3) the number of accurate responses to the radio messages; and 4) subject ratings of stress and workload. Subjects were six male U.S. Army personnel and six male civilians. Results indicated that ASR and human speech output were not significantly affected by any of the independent variables. A mean of 95.5% of all subject verbal commands was recognized by the ASR system. Spatialized radio messages significantly enhanced the ability of the user to understand speech communications from multiple channels. However, spatial audio information input had no significant effect on user speech output or on ASR performance. In general, ASR and spatial audio displays have potential as technologies of interest in high noise, stressful tank environments in which complex, concurrent tasks are performed.

Measuring automatic speech recognition (ASR) system quality is critical for creating user-satisfying voice-driven applications. Word Error Rate (WER) has been traditionally used to evaluate ASR system quality; however, it sometimes correlates poorly with user perception/judgement of transcription quality. This is because WER weighs every word equally and does not consider semantic correctness which has a higher impact on user perception. In this work, we propose evaluating ASR output hypotheses quality with SemDist that can measure semantic correctness by using the distance between the semantic vectors of the reference and hypothesis extracted from a pre-trained language model. Our experimental results of 71K and 36K user annotated ASR output quality show that SemDist achieves higher correlation with user perception than WER. We also show that SemDist has higher correlation with downstream Natural Language Understanding (NLU) tasks than WER.

Word vector representations enable machines to encode human language for spoken language understanding and processing. Confusion2vec, motivated from human speech production and perception, is a word vector representation which encodes ambiguities present in human spoken language in addition to semantics and syntactic information. Confusion2vec provides a robust spoken language representation by considering inherent human language ambiguities. In this paper, we propose a novel word vector space estimation by unsupervised learning on lattices output by an automatic speech recognition (ASR) system. We encode each word in Confusion2vec vector space by its constituent subword character n-grams. We show that the subword encoding helps better represent the acoustic perceptual ambiguities in human spoken language via information modeled on lattice-structured ASR output. The usefulness of the proposed Confusion2vec representation is evaluated using analogy and word similarity tasks designed for assessing semantic, syntactic and acoustic word relations. We also show the benefits of subword modeling for acoustic ambiguity representation on the task of spoken language intent detection. The results significantly outperform existing word vector representations when evaluated on erroneous ASR outputs, providing improvements up-to 13.12% relative to previous state-of-the-art in intent detection on ATIS benchmark dataset. We demonstrate that Confusion2vec subword modeling eliminates the need for retraining/adapting the natural language understanding models on ASR transcripts.

Purpose of research.Emotions play one of the key roles in the regulation of human behaviour. Solving the problem of automatic recognition of emotions makes it possible to increase the effectiveness of operation of a whole range of digital systems such as security systems, human-machine interfaces, e-commerce systems, etc. At the same time, the low efficiency of modern approaches to recognizing emotions in speech can be noted. This work studies automatic recognition of emotions in speech applying machine learning methods.Methods.The article describes and tests an approach to automatic recognition of emotions in speech based on multitask learning of deep convolution neural networks of AlexNet and VGG architectures using automatic selection of the weight coefficients for each task when calculating the final loss value during learning. All the models were trained on a sample of the IEMOCAP dataset with four emotional categories of ‘anger’, ‘happiness’, ‘neutral emotion’, ‘sadness’. The log-mel spectrograms of statements processed by a specialized algorithm are used as input data.Results.The considered models were tested on the basis of numerical metrics: the share of correctly recognized instances, accuracy, completeness, f-measure. For all of the above metrics, an improvement in the quality of emotion recognition by the proposed model was obtained in comparison with the two basic single-task models as well as with known solutions. This result is achieved through the use of automatic weighting of the values of the loss functions from individual tasks when forming the final value of the error in the learning process.Conclusion.The resulting improvement in the quality of emotion recognition in comparison with the known solutions confirms the feasibility of applying multitask learning to increase the accuracy of emotion recognition models. The developed approach makes it possible to achieve a uniform and simultaneous reduction of errors of individual tasks, and is used in the field of emotions recognition in speech for the first time.

The history of humanity has reached Industry 4.0 that aims to the integration of information technologies and especially artificial intelligence with all life-sustaining mechanisms in the 21st century, and consecutively, the transformation of Society 5.0 has begun. Society 5.0 means a smart society in which humans share life with physical robots and software robots as well as smart devices based on augmented reality. Industry 4.0 contains main structures such as the internet of things, big data analytics, digital transformation, cyber-physical systems, artificial intelligence, and business processes optimization. It is impossible to consider the machines to be without emotions and emotional intelligence within the transformation of smart tools and artificial intelligence, in addition, while it is planned to give most of the commands with voice and speaking, it became more important to develop algorithms that can detect emotions. In the smart society, new and rapid methods are needed for speech recognition, emotion recognition, and speech emotion recognition areas to maximize human-computer (HCI) or human-robot interaction (HRI) and collaboration. In this study, speech recognition and speech emotion recognition studies in robot technology are investigated and developments are revealed.