Multilingual Speech Recognition for Low-Resource Indian Languages using Multi-Task conformer

Transformers have recently become very popular for sequence-to-sequence applications such as machine translation and speech recognition. In this work, we propose a multi-task learning-based transformer model for low-resource multilingual speech recognition for Indian languages. Our proposed model consists of a conformer [1] encoder and two parallel transformer decoders. We use a phoneme decoder (PHN-DEC) for the phoneme recognition task and a grapheme decoder (GRP-DEC) to predict grapheme sequence. We consider the phoneme recognition task as an auxiliary task for our multi-task learning framework. We jointly optimize the network for both phoneme and grapheme recognition tasks using Joint CTC-Attention [2] training. We use a conditional decoding scheme to inject the language information into the model before predicting the grapheme sequence. Our experiments show that our proposed approach can obtain significant improvement over previous approaches [4]. We also show that our conformer-based dual-decoder approach outperforms both the transformer-based dual-decoder approach and single decoder approach. Finally, We compare monolingual ASR models with our proposed multilingual ASR approach.

A Dual-Decoder Conformer for Multilingual Speech Recognition

Transformer-based models have recently become very popular for sequence-to-sequence applications such as machine translation and speech recognition. This work proposes a dual-decoder transformer model for low-resource multilingual speech recognition for Indian languages. Our proposed model consists of a Conformer [1] encoder, two parallel transformer decoders, and a language classifier. We use a phoneme decoder (PHN-DEC) for the phoneme recognition task and a grapheme decoder (GRP-DEC) to predict grapheme sequence along with language information. We consider phoneme recognition and language identification as auxiliary tasks in the multi-task learning framework. We jointly optimize the network for phoneme recognition, grapheme recognition, and language identification tasks with Joint CTC-Attention [2] training. Our experiments show that we can obtain a significant reduction in WER over the baseline approaches. We also show that our dual-decoder approach obtains significant improvement over the single decoder approach.

Using Large Pre-Trained Models with Cross-Modal Attention for Multi-Modal Emotion Recognition

Recently, self-supervised pre-training has shown significant improvements in many areas of machine learning, including speech and NLP. We propose using large self-supervised pre-trained models for both audio and text modality with cross-modality attention for multimodal emotion recognition. We use Wav2Vec2.0 [1] as an audio encoder base for robust speech features extraction and the BERT model [2] as a text encoder base for better contextual representation of text. These high capacity models trained on large amounts of unlabeled data contain rich feature representations and improve the downstream task's performance. We use the cross-modal attention [3] mechanism to learn alignment between audio and text representations from self-supervised models. Cross-modal attention also helps in extracting interactive information between audio and text features. We obtain utterance-level feature representation from frame-level features using statistics pooling for both audio and text modality and combine them using the early fusion technique. Our experiments show that the proposed approach obtains a 1.88% absolute improvement in accuracy compared to the previous state-of-the-art method [3] on the IEMOCAP dataset [35]. We also conduct unimodal experiments for both audio and text modalities and compare them with previous best methods.

End-to-End Language Identification using Multi-Head Self-Attention and 1D Convolutional Neural Networks

In this work, we propose a new approach for language identification using multi-head self-attention combined with raw waveform based 1D convolutional neural networks for Indian languages. Our approach uses an encoder, multi-head selfattention, and a statistics pooling layer. The encoder learns features directly from raw waveforms using 1D convolution kernels and an LSTM layer. The LSTM layer captures temporal information between the features extracted by the 1D convolutional layer. The multi-head self-attention layer takes outputs of the LSTM layer and applies self-attention mechanisms on these features with M different heads. This process helps the model give more weightage to the more useful features and less weightage to the less relevant features. Finally, the frame-level features are combined using a statistics pooling layer to extract the utterance-level feature vector label prediction. We conduct all our experiments on the 373 hrs of audio data for eight different Indian languages. Our experiments show that our approach outperforms the baseline model by an absolute 3.69% improvement in F1-score and achieves the best F1-score of 95.90%. Our approach also shows that using raw waveform models gets a 1.7% improvement in performance compared to the models built using handcrafted features.

Identification of Indian Languages using Ghost-VLAD pooling

In this work, we propose a new pooling strategy for language identification by considering Indian languages. The idea is to obtain utterance level features for any variable length audio for robust language recognition. We use the GhostVLAD approach to generate an utterance level feature vector for any variable length input audio by aggregating the local frame level features across time. The generated feature vector is shown to have very good language discriminative features and helps in getting state of the art results for language identification task. We conduct our experiments on 635Hrs of audio data for 7 Indian languages. Our method outperforms the previous state of the art x-vector [11] method by an absolute improvement of 1.88% in F1-score and achieves 98.43% F1-score on the held-out test data. We compare our system with various pooling approaches and show that GhostVLAD is the best pooling approach for this task. We also provide visualization of the utterance level embeddings generated using Ghost-VLAD pooling and show that this method creates embeddings which has very good language discriminative features.