Recent innovations in self-supervised representation learning have led to remarkable advances in natural language processing. That said, in the speech processing domain, self-supervised representation learning-based systems are not yet considered state-of-the-art. We propose leveraging recent advances in self-supervised-based speech processing to create a common speech analysis engine. Such an engine should be able to handle multiple speech processing tasks, using a single architecture, to obtain state-of-the-art accuracy. The engine must also enable support for new tasks with small training datasets. Beyond that, a common engine should be capable of supporting distributed training with client in-house private data. We present the architecture for a common speech analysis engine based on the HuBERT self-supervised speech representation. Based on experiments, we report our results for language identification and emotion recognition on the standard evaluations NIST-LRE 07 and IEMOCAP. Our results surpass the state-of-the-art performance reported so far on these tasks. We also analyzed our engine on the emotion recognition task using reduced amounts of training data and show how to achieve improved results.
In this paper, we introduce a novel language identification system based on conformer layers. We propose an attentive temporal pooling mechanism to allow the model to carry information in long-form audio via a recurrent form, such that the inference can be performed in a streaming fashion. Additionally, we investigate two domain adaptation approaches to allow adapting an existing language identification model without retraining the model parameters for a new domain. We perform a comparative study of different model topologies under different constraints of model size, and find that conformer-based models significantly outperform LSTM and transformer based models. Our experiments also show that attentive temporal pooling and domain adaptation improve model accuracy.
To date, efforts in the code-switching literature have focused for the most part on language identification, POS, NER, and syntactic parsing. In this paper, we address machine translation for code-switched social media data. We create a community shared task. We provide two modalities for participation: supervised and unsupervised. For the supervised setting, participants are challenged to translate English into Hindi-English (Eng-Hinglish) in a single direction. For the unsupervised setting, we provide the following language pairs: English and Spanish-English (Eng-Spanglish), and English and Modern Standard Arabic-Egyptian Arabic (Eng-MSAEA) in both directions. We share insights and challenges in curating the "into" code-switching language evaluation data. Further, we provide baselines for all language pairs in the shared task. The leaderboard for the shared task comprises 12 individual system submissions corresponding to 5 different teams. The best performance achieved is 12.67% BLEU score for English to Hinglish and 25.72% BLEU score for MSAEA to English.
We investigate the performance of a state-of-the art (SoTA) architecture T5 (available on the SuperGLUE) and compare with it 3 other previous SoTA architectures across 5 different tasks from 2 relatively diverse datasets. The datasets are diverse in terms of the number and types of tasks they have. To improve performance, we augment the training data by using an autoregressive model. We achieve near-SoTA results on a couple of the tasks - macro F1 scores of 81.66% for task A of the OLID 2019 dataset and 82.54% for task A of the hate speech and offensive content (HASOC) 2021 dataset, where SoTA are 82.9% and 83.05%, respectively. We perform error analysis and explain why one of the models (Bi-LSTM) makes the predictions it does by using a publicly available algorithm: Integrated Gradient (IG). This is because explainable artificial intelligence (XAI) is essential for earning the trust of users. The main contributions of this work are the implementation method of T5, which is discussed; the data augmentation using a new conversational AI model checkpoint, which brought performance improvements; and the revelation on the shortcomings of HASOC 2021 dataset. It reveals the difficulties of poor data annotation by using a small set of examples where the T5 model made the correct predictions, even when the ground truth of the test set were incorrect (in our opinion). We also provide our model checkpoints on the HuggingFace hub1 to foster transparency.
In this paper, the challenges of maintaining a healthy IT operational environment have been addressed by proactively analyzing IT Service Desk tickets, customer satisfaction surveys, and social media data. A Cognitive solution goes beyond the traditional structured data analysis by deep analyses of both structured and unstructured text. The salient features of the proposed platform include language identification, translation, hierarchical extraction of the most frequently occurring topics, entities and their relationships, text summarization, sentiments, and knowledge extraction from the unstructured text using Natural Language Processing techniques. Moreover, the insights from unstructured text combined with structured data allow the development of various classification, segmentation, and time-series forecasting use-cases on the incident, problem, and change datasets. Further, the text and predictive insights together with raw data are used for visualization and exploration of actionable insights on a rich and interactive dashboard. However, it is hard not only to find these insights using traditional structured data analysis but it might also take a very long time to discover them, especially while dealing with a massive amount of unstructured data. By taking action on these insights, organizations can benefit from a significant reduction of ticket volume, reduced operational costs, and increased customer satisfaction. In various experiments, on average, upto 18-25% of yearly ticket volume has been reduced using the proposed approach.
Code-Switching (CS) is a common linguistic phenomenon in multilingual communities that consists of switching between languages while speaking. This paper presents our investigations on end-to-end speech recognition for Mandarin-English CS speech. We analyse different CS specific issues such as the properties mismatches between languages in a CS language pair, the unpredictable nature of switching points, and the data scarcity problem. We exploit and improve the state-of-the-art end-to-end system by merging nonlinguistic symbols, by integrating language identification using hierarchical softmax, by modeling sub-word units, by artificially lowering the speaking rate, and by augmenting data using speed perturbed technique and several monolingual datasets to improve the final performance not only on CS speech but also on monolingual benchmarks in order to make the system more applicable on real life settings. Finally, we explore the effect of different language model integration methods on the performance of the proposed model. Our experimental results reveal that all the proposed techniques improve the recognition performance. The best combined system improves the baseline system by up to 35% relatively in terms of mixed error rate and delivers acceptable performance on monolingual benchmarks.
This work aims to evaluate the ability that both probabilistic and state-of-the-art vector space modeling (VSM) methods provide to well known machine learning algorithms to identify social network documents to be classified as aggressive, gender biased or communally charged. To this end, an exploratory stage was performed first in order to find relevant settings to test, i.e. by using training and development samples, we trained multiple algorithms using multiple vector space modeling and probabilistic methods and discarded the less informative configurations. These systems were submitted to the competition of the ComMA@ICON'21 Workshop on Multilingual Gender Biased and Communal Language Identification.
Over the recent years, various deep learning-based methods were proposed for extracting a fixed-dimensional embedding vector from speech signals. Although the deep learning-based embedding extraction methods have shown good performance in numerous tasks including speaker verification, language identification and anti-spoofing, their performance is limited when it comes to mismatched conditions due to the variability within them unrelated to the main task. In order to alleviate this problem, we propose a novel training strategy that regularizes the embedding network to have minimum information about the nuisance attributes. To achieve this, our proposed method directly incorporates the information bottleneck scheme into the training process, where the mutual information is estimated using the main task classifier and an auxiliary normalizing flow network. The proposed method was evaluated on different speech processing tasks and showed improvement over the standard training strategy in all experimentation.
This paper presents XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0. We train models with up to 2B parameters on nearly half a million hours of publicly available speech audio in 128 languages, an order of magnitude more public data than the largest known prior work. Our evaluation covers a wide range of tasks, domains, data regimes and languages, both high and low-resource. On the CoVoST-2 speech translation benchmark, we improve the previous state of the art by an average of 7.4 BLEU over 21 translation directions into English. For speech recognition, XLS-R improves over the best known prior work on BABEL, MLS, CommonVoice as well as VoxPopuli, lowering error rates by 14-34% relative on average. XLS-R also sets a new state of the art on VoxLingua107 language identification. Moreover, we show that with sufficient model size, cross-lingual pretraining can outperform English-only pretraining when translating English speech into other languages, a setting which favors monolingual pretraining. We hope XLS-R can help to improve speech processing tasks for many more languages of the world.
With the fast growth of mobile computing and Web technologies, offensive language has become more prevalent on social networking platforms. Since offensive language identification in local languages is essential to moderate the social media content, in this paper we work with three Dravidian languages, namely Malayalam, Tamil, and Kannada, that are under-resourced. We present an evaluation task at FIRE 2020- HASOC-DravidianCodeMix and DravidianLangTech at EACL 2021, designed to provide a framework for comparing different approaches to this problem. This paper describes the data creation, defines the task, lists the participating systems, and discusses various methods.
This paper contains a post-challenge performance analysis on cross-lingual speaker verification of the IDLab submission to the VoxCeleb Speaker Recognition Challenge 2021 (VoxSRC-21). We show that current speaker embedding extractors consistently underestimate speaker similarity in within-speaker cross-lingual trials. Consequently, the typical training and scoring protocols do not put enough emphasis on the compensation of intra-speaker language variability. We propose two techniques to increase cross-lingual speaker verification robustness. First, we enhance our previously proposed Large-Margin Fine-Tuning (LM-FT) training stage with a mini-batch sampling strategy which increases the amount of intra-speaker cross-lingual samples within the mini-batch. Second, we incorporate language information in the logistic regression calibration stage. We integrate quality metrics based on soft and hard decisions of a VoxLingua107 language identification model. The proposed techniques result in a 11.7% relative improvement over the baseline model on the VoxSRC-21 test set and contributed to our third place finish in the corresponding challenge.
Motivated by recent innovations in biologically-inspired neuromorphic hardware, this article presents a novel unsupervised machine learning algorithm named Hyperseed that draws on the principles of Vector Symbolic Architectures (VSA) for fast learning of a topology preserving feature map of unlabelled data. It relies on two major operations of VSA, binding and bundling. The algorithmic part of Hyperseed is expressed within Fourier Holographic Reduced Representations model, which is specifically suited for implementation on spiking neuromorphic hardware. The two primary contributions of the Hyperseed algorithm are, few-shot learning and a learning rule based on single vector operation. These properties are empirically evaluated on synthetic datasets as well as on illustrative benchmark use-cases, IRIS classification, and a language identification task using n-gram statistics. The results of these experiments confirm the capabilities of Hyperseed and its applications in neuromorphic hardware.
Social media has become a bedrock for people to voice their opinions worldwide. Due to the greater sense of freedom with the anonymity feature, it is possible to disregard social etiquette online and attack others without facing severe consequences, inevitably propagating hate speech. The current measures to sift the online content and offset the hatred spread do not go far enough. One factor contributing to this is the prevalence of regional languages in social media and the paucity of language flexible hate speech detectors. The proposed work focuses on analyzing hate speech in Hindi-English code-switched language. Our method explores transformation techniques to capture precise text representation. To contain the structure of data and yet use it with existing algorithms, we developed MoH or Map Only Hindi, which means "Love" in Hindi. MoH pipeline consists of language identification, Roman to Devanagari Hindi transliteration using a knowledge base of Roman Hindi words. Finally, it employs the fine-tuned Multilingual Bert and MuRIL language models. We conducted several quantitative experiment studies on three datasets and evaluated performance using Precision, Recall, and F1 metrics. The first experiment studies MoH mapped text's performance with classical machine learning models and shows an average increase of 13% in F1 scores. The second compares the proposed work's scores with those of the baseline models and offers a rise in performance by 6%. Finally, the third reaches the proposed MoH technique with various data simulations using the existing transliteration library. Here, MoH outperforms the rest by 15%. Our results demonstrate a significant improvement in the state-of-the-art scores on all three datasets.
Code-switching (CS) is common in daily conversations where more than one language is used within a sentence. The difficulties of CS speech recognition lie in alternating languages and the lack of transcribed data. Therefore, this paper uses the recently successful self-supervised learning (SSL) methods to leverage many unlabeled speech data without CS. We show that hidden representations of SSL models offer frame-level language identity even if the models are trained with English speech only. Jointly training CTC and language identification modules with self-supervised speech representations improves CS speech recognition performance. Furthermore, using multilingual speech data for pre-training obtains the best CS speech recognition.
Language Identification (LID) is a crucial preliminary process in the field of Automatic Speech Recognition (ASR) that involves the identification of a spoken language from audio samples. Contemporary systems that can process speech in multiple languages require users to expressly designate one or more languages prior to utilization. The LID task assumes a significant role in scenarios where ASR systems are unable to comprehend the spoken language in multilingual settings, leading to unsuccessful speech recognition outcomes. The present study introduces convolutional recurrent neural network (CRNN) based LID, designed to operate on the Mel-frequency Cepstral Coefficient (MFCC) characteristics of audio samples. Furthermore, we replicate certain state-of-the-art methodologies, specifically the Convolutional Neural Network (CNN) and Attention-based Convolutional Recurrent Neural Network (CRNN with attention), and conduct a comparative analysis with our CRNN-based approach. We conducted comprehensive evaluations on thirteen distinct Indian languages and our model resulted in over 98\% classification accuracy. The LID model exhibits high-performance levels ranging from 97% to 100% for languages that are linguistically similar. The proposed LID model exhibits a high degree of extensibility to additional languages and demonstrates a strong resistance to noise, achieving 91.2% accuracy in a noisy setting when applied to a European Language (EU) dataset.
This paper describes the system submitted to Dravidian-Codemix-HASOC2021: Hate Speech and Offensive Language Identification in Dravidian Languages (Tamil-English and Malayalam-English). This task aims to identify offensive content in code-mixed comments/posts in Dravidian Languages collected from social media. Our approach utilizes pooling the last layers of pretrained transformer multilingual BERT for this task which helped us achieve rank nine on the leaderboard with a weighted average score of 0.61 for the Tamil-English dataset in subtask B. After the task deadline, we sampled the dataset uniformly and used the MuRIL pretrained model, which helped us achieve a weighted average score of 0.67, the top score in the leaderboard. Furthermore, our approach to utilizing the pretrained models helps reuse our models for the same task with a different dataset. Our code and models are available in https://github.com/seanbenhur/tanglish-offensive-language-identification
We summarize the results of a host of efforts using giant automatic speech recognition (ASR) models pre-trained using large, diverse unlabeled datasets containing approximately a million hours of audio. We find that the combination of pre-training, self-training and scaling up model size greatly increases data efficiency, even for extremely large tasks with tens of thousands of hours of labeled data. In particular, on an ASR task with 34k hours of labeled data, by fine-tuning an 8 billion parameter pre-trained Conformer model we can match state-of-the-art (SoTA) performance with only 3% of the training data and significantly improve SoTA with the full training set. We also report on the universal benefits gained from using big pre-trained and self-trained models for a large set of downstream tasks that cover a wide range of speech domains and span multiple orders of magnitudes of dataset sizes, including obtaining SoTA performance on many public benchmarks. In addition, we utilize the learned representation of pre-trained networks to achieve SoTA results on non-ASR tasks.
Language identification is a critical component of language processing pipelines (Jauhiainen et al.,2019) and is not a solved problem in real-world settings. We present a lightweight and effective language identifier that is robust to changes of domain and to the absence of copious training data. The key idea for classification is that the reciprocal of the rank in a frequency table makes an effective additive feature score, hence the term Reciprocal Rank Classifier (RRC). The key finding for language classification is that ranked lists of words and frequencies of characters form a sufficient and robust representation of the regularities of key languages and their orthographies. We test this on two 22-language data sets and demonstrate zero-effort domain adaptation from a Wikipedia training set to a Twitter test set. When trained on Wikipedia but applied to Twitter the macro-averaged F1-score of a conventionally trained SVM classifier drops from 90.9% to 77.7%. By contrast, the macro F1-score of RRC drops only from 93.1% to 90.6%. These classifiers are compared with those from fastText and langid. The RRC performs better than these established systems in most experiments, especially on short Wikipedia texts and Twitter. The RRC classifier can be improved for particular domains and conversational situations by adding words to the ranked lists. Using new terms learned from such conversations, we demonstrate a further 7.9% increase in accuracy of sample message classification, and 1.7% increase for conversation classification. Surprisingly, this made results on Twitter data slightly worse. The RRC classifier is available as an open source Python package (https://github.com/LivePersonInc/lplangid).
Identifying check-worthy claims is often the first step of automated fact-checking systems. Tackling this task in a multilingual setting has been understudied. Encoding inputs with multilingual text representations could be one approach to solve the multilingual check-worthiness detection. However, this approach could suffer if cultural bias exists within the communities on determining what is check-worthy.In this paper, we propose a language identification task as an auxiliary task to mitigate unintended bias.With this purpose, we experiment joint training by using the datasets from CLEF-2021 CheckThat!, that contain tweets in English, Arabic, Bulgarian, Spanish and Turkish. Our results show that joint training of language identification and check-worthy claim detection tasks can provide performance gains for some of the selected languages.
The first edition of the IberLEF 2021 shared task on automatic detection of borrowings (ADoBo) focused on detecting lexical borrowings that appeared in the Spanish press and that have recently been imported into the Spanish language. In this work, we tested supplementary training on intermediate labeled-data tasks (STILTs) from part of speech (POS), named entity recognition (NER), code-switching, and language identification approaches to the classification of borrowings at the token level using existing pre-trained transformer-based language models. Our extensive experimental results suggest that STILTs do not provide any improvement over direct fine-tuning of multilingual models. However, multilingual models trained on small subsets of languages perform reasonably better than multilingual BERT but not as good as multilingual RoBERTa for the given dataset.