Word embeddings for speech recognition

By Samy Bengio, Georg Heigold. 2014

153 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Natural Language Processing, Representation & Self-Supervised Learning, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Representation & Self-Supervised Learning

In Speech & Audio

In Signal Processing

Citations in the corpus

Effectiveness of self-supervised pre-training for speech recognition (2019, background): “The success of BERT [19] and Word2Vec [24] for NLP tasks motivated more research on self-supervised approaches for acoustic word embedding and unsupervised acoustic feature representation [39, 40, 41, 42, 10, 43, 12, 11, 44, 1], either by predicting masked discrete or continuous input, or by contras”
Speech2Vec: A Sequence-to-Sequence Framework for Learning Word Embeddings from Speech (2018, background): “Researchers have also explored the concept of learning vector representations from speech [9, 10, 11, 12, 13, 14].”
Audio Word2Vec: Unsupervised Learning of Audio Segment Representations Using Sequence-to-Sequence Autoencoder (2016, methodology): “Different from the previous work [12, 13, 11], here learning SA does not need any supervision.”
Unsupervised Cross-Modal Alignment of Speech and Text Embedding Spaces (2018, background): “speech [18, 19, 20, 21, 22, 23], they are primarily based on acoustic-phonetic similarity, and aim to represent the way a word sounds rather than its meaning.”
Combining Residual Networks with LSTMs for Lipreading (2017, background): “Nevertheless, recent advances in speech recognition and natural language processing show that direct modeling of words is feasible even for LVCSR, [7] [8] [9].”
Discriminative acoustic word embeddings: Tecurrent neural network-based approaches (2016, methodology): “Therehas been some, thus far limited, work on acoustic word embeddings, focused on a number of embedding models, training approaches, and tasks [10, 11, 12, 13, 14, 15, 16, 17].”
Word-level Speech Recognition with a Dynamic Lexicon (2019, result): “Similar to our work is that of Bengio and Heigold [6], who learn to predict word embeddings by using a sub-word level embedding model.”
Deep convolutional acoustic word embeddings using word-pair side information (2015, methodology): “Bengio and Heigold [4] similarly used whole-word ﬁxed-dimensional representations in a segmental ASR lattice rescoring system.”
Phonetic Context Embeddings for DNN-HMM Phone Recognition (2016, background): “, [10]), and in language modeling [11, 12, 13] and acoustic modeling/lattice re-scoring [14] for ASR.”
Acoustically Grounded Word Embeddings for Improved Acoustics-to-word Speech Recognition (2019, background): “Other work has considered acoustically grounded word embed-dings , that is embeddings of written words that encode their acoustic/phonetic content [12, 24, 27, 28].”
Word-Level Speech Recognition With a Letter to Word Encoder (2019, result): “Similar to our work is that of Bengio & Heigold (2014), who learn to predict word embeddings by using a sub-word level embedding model.”
Learning-Based Methods for Comparing Sequences, with Applications to Audio-to-MIDI Alignment and Matching (2016, background): “For example, (Kamper et al., 2016; Bengio and Heigold, 2014) learn to embed spoken utterances of a fixed length in a Euclidean space where utterances of the same word are close together.”
Cross-Loss Influence Functions to Explain Deep Network Representations (2020, background): “Deep learning has become a ubiquitous tool in domains from speech-recognition (Bengio and Heigold, 2014) to knowledge-discovery (Tshitoyan et al., 2019), owing to large datasets, increasing computation power, and the development of powerful unsupervised and semi-supervised learning techniques.”
Effectiveness of Self-Supervised Pre-Training for ASR (2020, background): “The practical need for building ASR systems for new conditions with limited resources spurred a lot of work focused on unsupervised speech recognition and representation learning [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14], in addition to semi- and weakly-supervised learning techniques to reduce th”
Unspeech: Unsupervised Speech Context Embeddings (2018, background): “In [9], Bengio and Heigold proposed supervised word embeddings for speech recognition, where words are nearby in the vector space if they sound alike.”
Semantic Query-by-example Speech Search Using Visual Grounding (2019, methodology): “Various acoustic embedding functions have been proposed, with neural models used in several studies [16, 18, 23–30].”
Multi-view Recurrent Neural Acoustic Word Embeddings (2016, methodology): “Such embeddings could be useful for tasks like spoken term detection (Fiscus et al., 2007), spoken query-by-example search (Anguera et al., 2014), or even speech recognition using a whole-word approach (Gemmeke et al., 2011; Bengio & Heigold, 2014).”
Deep multimodal semantic embeddings for speech and images (2015, methodology): “Standard CNNs expect their inputs to be of a fixed size, so in order to accommodate our variable duration words we follow [12] and choose to embed their spectrograms in a fixed duration window, applying zero-padding and truncation when necessary.”
Audio Word2vec: Sequence-to-Sequence Autoencoding for Unsupervised Learning of Audio Segmentation and Representation (2019, methodology): “Recently, deep learning has been used to encode acoustic information as vectors [36]–[44].”