Wav2vec 2.0 is a recently proposed self-supervised framework for speech representation learning. It follows a two-stage training process of pre-training and fine-tuning, and performs well in speech recognition tasks especially ultra-low resource cases. In this work, we attempt to extend self-supervised framework to speaker verification and language identification. First, we use some preliminary experiments to indicate that wav2vec 2.0 can capture the information about the speaker and language. Then we demonstrate the effectiveness of wav2vec 2.0 on the two tasks respectively. For speaker verification, we obtain a new state-of-the-art result, Equal Error Rate (EER) of 3.61% on the VoxCeleb1 dataset. For language identification, we obtain an EER of 12.02% on 1 second condition and an EER of 3.47% on full-length condition of the AP17-OLR dataset. Finally, we utilize one model to achieve the unified modeling by the multi-task learning for the two tasks.
Automatic language identification is a challenging problem. Discriminating between closely related languages is especially difficult. This paper presents a machine learning approach for automatic language identification for the Nordic languages, which often suffer miscategorisation by existing state-of-the-art tools. Concretely we will focus on discrimination between six Nordic languages: Danish, Swedish, Norwegian (Nynorsk), Norwegian (Bokm{\aa}l), Faroese and Icelandic.
We present several methods to improve the generalisation of language identification (LID) systems to new speakers and to new domains. These methods involve Spectral augmentation, where spectrograms are masked in the frequency or time bands during training and CNN architectures that are pre-trained on the Imagenet dataset. The paper also introduces the novel Triplet Entropy Loss training method, which involves training a network simultaneously using Cross Entropy and Triplet loss. It was found that all three methods improved the generalisation of the models, though not significantly. Even though the models trained using Triplet Entropy Loss showed a better understanding of the languages and higher accuracies, it appears as though the models still memorise word patterns present in the spectrograms rather than learning the finer nuances of a language. The research shows that Triplet Entropy Loss has great potential and should be investigated further, not only in language identification tasks but any classification task.
language identification (LID) is identifing a language in a given spoken utterance. Language segmentation is equally inportant as language identification where language boundaries can be spotted in a multi language utterance. In this paper, we have experimented with two schemes for language identification in Indian regional language context as very few works has been done. Singular value based feature embedding is used for both of the schemes. In first scheme, the singular value decomposition (SVD) is applied to the n-gram utterance matrix and in the second scheme, SVD is applied on the difference supervector matrix space. We have observed that in both the schemes, 55-65% singular value energy is sufficient to capture the language context. In n-gram based feature representation, we have seen that different skipgram models capture different language context. We have observed that for short test duration, supervector based feature representation is better but with a longer duration test signal, n-gram based feature performed better. We have also extended our work to explore language-based segmentation where we have seen that segmentation accuracy of four language group with ten language training model, scheme-1 has performed well but with same four language training model, scheme-2 outperformed scheme-1
Code-switching refers to the usage of two languages within a sentence or discourse. It is a global phenomenon among multilingual communities and has emerged as an independent area of research. With the increasing demand for the code-switching automatic speech recognition (ASR) systems, the development of a code-switching speech corpus has become highly desirable. However, for training such systems, very limited code-switched resources are available as yet. In this work, we present our first efforts in building a code-switching ASR system in the Indian context. For that purpose, we have created a Hindi-English code-switching speech database. The database not only contains the speech utterances with code-switching properties but also covers the session and the speaker variations like pronunciation, accent, age, gender, etc. This database can be applied in several speech signal processing applications, such as code-switching ASR, language identification, language modeling, speech synthesis etc. This paper mainly presents an analysis of the statistics of the collected code-switching speech corpus. Later, the performance results for the ASR task have been reported for the created database.
Natural language processing (NLP) techniques have become mainstream in the recent decade. Most of these advances are attributed to the processing of a single language. More recently, with the extensive growth of social media platforms focus has shifted to code-mixed text. The code-mixed text comprises text written in more than one language. People naturally tend to combine local language with global languages like English. To process such texts, current NLP techniques are not sufficient. As a first step, the text is processed to identify the language of the words in the text. In this work, we focus on language identification in code-mixed sentences for Hindi-English mixed text. The task of language identification is formulated as a token classification task. In the supervised setting, each word in the sentence has an associated language label. We evaluate different deep learning models and input representation combinations for this task. Mainly, character, sub-word, and word embeddings are considered in combination with CNN and LSTM based models. We show that sub-word representation along with the LSTM model gives the best results. In general sub-word representations perform significantly better than other input representations. We report the best accuracy of 94.52% using a single layer LSTM model on the standard SAIL ICON 2017 test set.
Diacritic characters can be considered as a unique set of characters providing us with adequate and significant clue in identifying a given language with considerably high accuracy. Diacritics, though associated with phonetics often serve as a distinguishing feature for many languages especially the ones with a Latin script. In this proposed work, we aim to identify language of text in images using the presence of diacritic characters in order to improve Optical Character Recognition (OCR) performance in any given automated environment. We showcase our work across 13 Latin languages encompassing 85 diacritic characters. We use an architecture similar to Squeezedet for object detection of diacritic characters followed by a shallow network to finally identify the language. OCR systems when accompanied with identified language parameter tends to produce better results than sole deployment of OCR systems. The discussed work apart from guaranteeing an improvement in OCR results also takes on-device (mobile phone) constraints into consideration in terms of model size and inference time.
This paper describes the WLV-RIT entry to the Hate Speech and Offensive Content Identification in Indo-European Languages (HASOC) shared task 2020. The HASOC 2020 organizers provided participants with annotated datasets containing social media posts of code-mixed in Dravidian languages (Malayalam-English and Tamil-English). We participated in task 1: Offensive comment identification in Code-mixed Malayalam Youtube comments. In our methodology, we take advantage of available English data by applying cross-lingual contextual word embeddings and transfer learning to make predictions to Malayalam data. We further improve the results using various fine tuning strategies. Our system achieved 0.89 weighted average F1 score for the test set and it ranked 5th place out of 12 participants.
Large text corpora are increasingly important for a wide variety of Natural Language Processing (NLP) tasks, and automatic language identification (LangID) is a core technology needed to collect such datasets in a multilingual context. LangID is largely treated as solved in the literature, with models reported that achieve over 90% average F1 on as many as 1,366 languages. We train LangID models on up to 1,629 languages with comparable quality on held-out test sets, but find that human-judged LangID accuracy for web-crawl text corpora created using these models is only around 5% for many lower-resource languages, suggesting a need for more robust evaluation. Further analysis revealed a variety of error modes, arising from domain mismatch, class imbalance, language similarity, and insufficiently expressive models. We propose two classes of techniques to mitigate these errors: wordlist-based tunable-precision filters (for which we release curated lists in about 500 languages) and transformer-based semi-supervised LangID models, which increase median dataset precision from 5.5% to 71.2%. These techniques enable us to create an initial data set covering 100K or more relatively clean sentences in each of 500+ languages, paving the way towards a 1,000-language web text corpus.
Deep neural networks have been employed for various spoken language recognition tasks, including tasks that are multilingual by definition such as spoken language identification. In this paper, we present a neural model for Slavic language identification in speech signals and analyze its emergent representations to investigate whether they reflect objective measures of language relatedness and/or non-linguists' perception of language similarity. While our analysis shows that the language representation space indeed captures language relatedness to a great extent, we find perceptual confusability between languages in our study to be the best predictor of the language representation similarity.
Sign Language helps people with Speaking and Hearing Disabilities communicate with others efficiently. Sign Language identification is a challenging area in the field of computer vision and recent developments have been able to achieve near perfect results for the task, though some challenges are yet to be solved. In this paper we propose a novel machine learning based pipeline for American Sign Language identification using hand track points. We convert a hand gesture into a series of hand track point coordinates that serve as an input to our system. In order to make the solution more efficient, we experimented with 28 different combinations of pre-processing techniques, each run on three different machine learning algorithms namely k-Nearest Neighbours, Random Forests and a Neural Network. Their performance was contrasted to determine the best pre-processing scheme and algorithm pair. Our system achieved an Accuracy of 95.66% to identify American sign language gestures.
With the popularity of social media, communications through blogs, Facebook, Twitter, and other plat-forms have increased. Initially, English was the only medium of communication. Fortunately, now we can communicate in any language. It has led to people using English and their own native or mother tongue language in a mixed form. Sometimes, comments in other languages have English transliterated format or other cases; people use the intended language scripts. Identifying sentiments and offensive content from such code mixed tweets is a necessary task in these times. We present a working model submitted for Task2 of the sub-track HASOC Offensive Language Identification- DravidianCodeMix in Forum for Information Retrieval Evaluation, 2020. It is a message level classification task. An embedding model-based classifier identifies offensive and not offensive comments in our approach. We applied this method in the Manglish dataset provided along with the sub-track.
Spoken language Identification (LID) systems are needed to identify the language(s) present in a given audio sample, and typically could be the first step in many speech processing related tasks such as automatic speech recognition (ASR). Automatic identification of the languages present in a speech signal is not only scientifically interesting, but also of practical importance in a multilingual country such as India. In many of the Indian cities, when people interact with each other, as many as three languages may get mixed. These may include the official language of that province, Hindi and English (at times the languages of the neighboring provinces may also get mixed during these interactions). This makes the spoken LID task extremely challenging in Indian context. While quite a few LID systems in the context of Indian languages have been implemented, most such systems have used small scale speech data collected internally within an organization. In the current work, we perform spoken LID on three Indian languages (Gujarati, Telugu, and Tamil) code-mixed with English. This task was organized by the Microsoft research team as a spoken LID challenge. In our work, we modify the usual spectral augmentation approach and propose a language mask that discriminates the language ID pairs, which leads to a noise robust spoken LID system. The proposed method gives a relative improvement of approximately 3-5% in the LID accuracy over a baseline system proposed by Microsoft on the three language pairs for two shared tasks suggested in the challenge.
In this paper, we describe the team \textit{BRUMS} entry to OffensEval 2: Multilingual Offensive Language Identification in Social Media in SemEval-2020. The OffensEval organizers provided participants with annotated datasets containing posts from social media in Arabic, Danish, English, Greek and Turkish. We present a multilingual deep learning model to identify offensive language in social media. Overall, the approach achieves acceptable evaluation scores, while maintaining flexibility between languages.
Offensive content is pervasive in social media and a reason for concern to companies and government organizations. Several studies have been recently published investigating methods to detect the various forms of such content (e.g. hate speech, cyberbulling, and cyberaggression). The clear majority of these studies deal with English partially because most annotated datasets available contain English data. In this paper, we take advantage of English data available by applying cross-lingual contextual word embeddings and transfer learning to make predictions in languages with less resources. We project predictions on comparable data in Bengali, Hindi, and Spanish and we report results of 0.8415 F1 macro for Bengali, 0.8568 F1 macro for Hindi, and 0.7513 F1 macro for Spanish. Finally, we show that our approach compares favorably to the best systems submitted to recent shared tasks on these three languages, confirming the robustness of cross-lingual contextual embeddings and transfer learning for this task.
In a multilingual or sociolingual configuration Intra-sentential Code Switching (ICS) or Code Mixing (CM) is frequently observed nowadays. In the world, most of the people know more than one language. CM usage is especially apparent in social media platforms. Moreover, ICS is particularly significant in the context of technology, health, and law where conveying the upcoming developments are difficult in one's native language. In applications like dialog systems, machine translation, semantic parsing, shallow parsing, etc. CM and Code Switching pose serious challenges. To do any further advancement in code-mixed data, the necessary step is Language Identification. In this paper, we present a study of various models - Nave Bayes Classifier, Random Forest Classifier, Conditional Random Field (CRF), and Hidden Markov Model (HMM) for Language Identification in English - Telugu Code Mixed Data. Considering the paucity of resources in code mixed languages, we proposed the CRF model and HMM model for word level language identification. Our best performing system is CRF-based with an f1-score of 0.91.
This paper describes Galileo's performance in SemEval-2020 Task 12 on detecting and categorizing offensive language in social media. For Offensive Language Identification, we proposed a multi-lingual method using Pre-trained Language Models, ERNIE and XLM-R. For offensive language categorization, we proposed a knowledge distillation method trained on soft labels generated by several supervised models. Our team participated in all three sub-tasks. In Sub-task A - Offensive Language Identification, we ranked first in terms of average F1 scores in all languages. We are also the only team which ranked among the top three across all languages. We also took the first place in Sub-task B - Automatic Categorization of Offense Types and Sub-task C - Offence Target Identification.
In this paper, we describe the PUM team's entry to the SemEval-2020 Task 12. Creating our solution involved leveraging two well-known pretrained models used in natural language processing: BERT and XLNet, which achieve state-of-the-art results in multiple NLP tasks. The models were fine-tuned for each subtask separately and features taken from their hidden layers were combined and fed into a fully connected neural network. The model using aggregated Transformer features can serve as a powerful tool for offensive language identification problem. Our team was ranked 7th out of 40 in Sub-task C - Offense target identification with 64.727% macro F1-score and 64th out of 85 in Sub-task A - Offensive language identification (89.726% F1-score).
Slot-filling, Translation, Intent classification, and Language identification, or STIL, is a newly-proposed task for multilingual Natural Language Understanding (NLU). By performing simultaneous slot filling and translation into a single output language (English in this case), some portion of downstream system components can be monolingual, reducing development and maintenance cost. Results are given using the multilingual BART model (Liu et al., 2020) fine-tuned on 7 languages using the MultiATIS++ dataset. When no translation is performed, mBART's performance is comparable to the current state of the art system (Cross-Lingual BERT by Xu et al. (2020)) for the languages tested, with better average intent classification accuracy (96.07% versus 95.50%) but worse average slot F1 (89.87% versus 90.81%). When simultaneous translation is performed, average intent classification accuracy degrades by only 1.7% relative and average slot F1 degrades by only 1.2% relative.
Identification of the languages written using cuneiform symbols is a difficult task due to the lack of resources and the problem of tokenization. The Cuneiform Language Identification task in VarDial 2019 addresses the problem of identifying seven languages and dialects written in cuneiform; Sumerian and six dialects of Akkadian language: Old Babylonian, Middle Babylonian Peripheral, Standard Babylonian, Neo-Babylonian, Late Babylonian, and Neo-Assyrian. This paper describes the approaches taken by SharifCL team to this problem in VarDial 2019. The best result belongs to an ensemble of Support Vector Machines and a naive Bayes classifier, both working on character-level features, with macro-averaged F1-score of 72.10%.