Byte pair encoding (BPE) emerges as an effective tokenization method for tackling the out-of-vocabulary (OOV) challenge in various natural language and speech processing tasks. Recent research highlights the dependency of BPE subword tokenization's efficacy on the morphological nature of the language, particularly in languages rich in inflectional morphology, where fewer BPE merges suffice for generating highly productive tokens. Motivated by this, our study empirically identifies the optimal number of BPE tokens for Bengali, a language known for its morphological complexity, thus enhancing out-of-distribution automatic speech recognition (ASR) performance. Experimental evaluation reveals that an excessively high number of BPE tokens can lead to overfitting, while approximately 500-1000 tokens result in superior OOV performance. Furthermore, we conduct a comparative analysis of BPE with character-based and unigram-based tokenization methods. By introducing BPE tokenization to Bengali ASR, we achieve a substantial reduction in the word error rate (WER) from 66.44% in our character-based baseline system to 63.80% on the LB-ASRTD eval set and from 46.34% to 42.80% on the SHRUTI eval set, both of which include out-of-distribution data.
Dynamic parameterization of acoustic environments has drawn widespread attention in the field of audio processing. Precise representation of local room acoustic characteristics is crucial when designing audio filters for various audio rendering applications. Key parameters in this context include reverberation time (RT60) and geometric room volume. In recent years, neural networks have been extensively applied in the task of blind room parameter estimation. However, there remains a question of whether pure attention mechanisms can achieve superior performance in this task. To address this issue, this study employs blind room parameter estimation based on monaural noisy speech signals. Various model architectures are investigated, including a proposed attention-based model. This model is a convolution-free Audio Spectrogram Transformer, utilizing patch splitting, attention mechanisms, and cross-modality transfer learning from a pretrained Vision Transformer. Experimental results suggest that the proposed attention mechanism-based model, relying purely on attention mechanisms without using convolution, exhibits significantly improved performance across various room parameter estimation tasks, especially with the help of dedicated pretraining and data augmentation schemes. Additionally, the model demonstrates more advantageous adaptability and robustness when handling variable-length audio inputs compared to existing methods.
Automatic speech recognition (ASR) outcomes serve as input for downstream tasks, substantially impacting the satisfaction level of end-users. Hence, the diagnosis and enhancement of the vulnerabilities present in the ASR model bear significant importance. However, traditional evaluation methodologies of ASR systems generate a singular, composite quantitative metric, which fails to provide comprehensive insight into specific vulnerabilities. This lack of detail extends to the post-processing stage, resulting in further obfuscation of potential weaknesses. Despite an ASR model's ability to recognize utterances accurately, subpar readability can negatively affect user satisfaction, giving rise to a trade-off between recognition accuracy and user-friendliness. To effectively address this, it is imperative to consider both the speech-level, crucial for recognition accuracy, and the text-level, critical for user-friendliness. Consequently, we propose the development of an Error Explainable Benchmark (EEB) dataset. This dataset, while considering both speech- and text-level, enables a granular understanding of the model's shortcomings. Our proposition provides a structured pathway for a more `real-world-centric' evaluation, a marked shift away from abstracted, traditional methods, allowing for the detection and rectification of nuanced system weaknesses, ultimately aiming for an improved user experience.
We present Aria Everyday Activities (AEA) Dataset, an egocentric multimodal open dataset recorded using Project Aria glasses. AEA contains 143 daily activity sequences recorded by multiple wearers in five geographically diverse indoor locations. Each of the recording contains multimodal sensor data recorded through the Project Aria glasses. In addition, AEA provides machine perception data including high frequency globally aligned 3D trajectories, scene point cloud, per-frame 3D eye gaze vector and time aligned speech transcription. In this paper, we demonstrate a few exemplar research applications enabled by this dataset, including neural scene reconstruction and prompted segmentation. AEA is an open source dataset that can be downloaded from https://www.projectaria.com/datasets/aea/. We are also providing open-source implementations and examples of how to use the dataset in Project Aria Tools https://github.com/facebookresearch/projectaria_tools.
Since the first speech recognition systems were built more than 30 years ago, improvement in voice technology has enabled applications such as smart assistants and automated customer support. However, conversation intelligence of the future requires recognizing free-flowing multi-party conversations, which is a crucial and challenging component that still remains unsolved. In this dissertation, we focus on this problem of speaker-attributed multi-talker speech recognition, and propose two perspectives which result from its probabilistic formulation. In the modular perspective, we build a pipeline of sub-tasks involving speaker diarization, target speaker extraction, and speech recognition. Our first contribution is a method to perform overlap-aware diarization by reformulating spectral clustering as a constrained optimization problem. We also describe an algorithm to ensemble diarization outputs, either to combine overlap-aware systems or to perform multi-channel diarization by late fusion. Once speaker segments are identified, we robustly extract single-speaker utterances from the mixture using a GPU-accelerated implementation of guided source separation, which allows us to use an off-the-shelf ASR system to obtain speaker-attributed transcripts. Since the modular approach suffers from error propagation, we propose an alternate "end-to-end" perspective on the problem. For this, we describe the Streaming Unmixing and Recognition Transducer (SURT). We show how to train SURT models efficiently by carefully designing the network architecture, objective functions, and mixture simulation techniques. Finally, we add an auxiliary speaker branch to enable joint prediction of speaker labels synchronized with the speech tokens. We demonstrate that training on synthetic mixtures and adapting with real data helps these models transfer well for streaming transcription of real meeting sessions.
Endpoint (EP) detection is a key component of far-field speech recognition systems that assist the user through voice commands. The endpoint detector has to trade-off between accuracy and latency, since waiting longer reduces the cases of users being cut-off early. We propose a novel two-pass solution for endpointing, where the utterance endpoint detected from a first pass endpointer is verified by a 2nd-pass model termed EP Arbitrator. Our method improves the trade-off between early cut-offs and latency over a baseline endpointer, as tested on datasets including voice-assistant transactional queries, conversational speech, and the public SLURP corpus. We demonstrate that our method shows improvements regardless of the first-pass EP model used.
This paper presents a comprehensive overview of the Comparative Opinion Mining from Vietnamese Product Reviews shared task (ComOM), held as part of the 10$^{th}$ International Workshop on Vietnamese Language and Speech Processing (VLSP 2023). The primary objective of this shared task is to advance the field of natural language processing by developing techniques that proficiently extract comparative opinions from Vietnamese product reviews. Participants are challenged to propose models that adeptly extract a comparative "quintuple" from a comparative sentence, encompassing Subject, Object, Aspect, Predicate, and Comparison Type Label. We construct a human-annotated dataset comprising $120$ documents, encompassing $7427$ non-comparative sentences and $2468$ comparisons within $1798$ sentences. Participating models undergo evaluation and ranking based on the Exact match macro-averaged quintuple F1 score.
In the field of spoken language understanding, systems like Whisper and Multilingual Massive Speech (MMS) have shown state-of-the-art performances. This study is dedicated to a comprehensive exploration of the Whisper and MMS systems, with a focus on assessing biases in automatic speech recognition (ASR) inherent to casual conversation speech specific to the Portuguese language. Our investigation encompasses various categories, including gender, age, skin tone color, and geo-location. Alongside traditional ASR evaluation metrics such as Word Error Rate (WER), we have incorporated p-value statistical significance for gender bias analysis. Furthermore, we extensively examine the impact of data distribution and empirically show that oversampling techniques alleviate such stereotypical biases. This research represents a pioneering effort in quantifying biases in the Portuguese language context through the application of MMS and Whisper, contributing to a better understanding of ASR systems' performance in multilingual settings.
Hate speech is harmful content that directly attacks or promotes hatred against members of groups or individuals based on actual or perceived aspects of identity, such as racism, religion, or sexual orientation. This can affect social life on social media platforms as hateful content shared through social media can harm both individuals and communities. As the prevalence of hate speech increases online, the demand for automated detection as an NLP task is increasing. In this work, the proposed method is using transformer-based model to detect hate speech in social media, like twitter, Facebook, WhatsApp, Instagram, etc. The proposed model is independent of languages and has been tested on Italian, English, German, Bengali. The Gold standard datasets were collected from renowned researcher Zeerak Talat, Sara Tonelli, Melanie Siegel, and Rezaul Karim. The success rate of the proposed model for hate speech detection is higher than the existing baseline and state-of-the-art models with accuracy in Bengali dataset is 89%, in English: 91%, in German dataset 91% and in Italian dataset it is 77%. The proposed algorithm shows substantial improvement to the benchmark method.
As large-scale language models become the standard for text generation, there is a greater need to tailor the generations to be more or less concise, targeted, and informative, depending on the audience/application. Existing control approaches primarily adjust the semantic (e.g., emotion, topics), structural (e.g., syntax tree, parts-of-speech), and lexical (e.g., keyword/phrase inclusion) properties of text, but are insufficient to accomplish complex objectives such as pacing which control the complexity and readability of the text. In this paper, we introduce CEV-LM - a lightweight, semi-autoregressive language model that utilizes constrained edit vectors to control three complementary metrics (speed, volume, and circuitousness) that quantify the shape of text (e.g., pacing of content). We study an extensive set of state-of-the-art CTG models and find that CEV-LM provides significantly more targeted and precise control of these three metrics while preserving semantic content, using less training data, and containing fewer parameters.