Active learning for automatic speech recognition

By Dilek Z. Hakkani-Tür, Giuseppe Riccardi, Allen L. Gorin. 2002

159 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, Meta & Continual Learning

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

1. Li Deng
2. Dong Yu
3. Jinyu Li
4. Jianfeng Gao
5. Yoshua Bengio
6. Mark John Francis Gales
7. Dilek Z. Hakkani-Tür
8. Xiaodong He

In Speech & Audio

1. Dong Yu
2. Jinyu Li
3. Li Deng
4. Mark John Francis Gales
5. Hermann Ney
6. Ralf Schlüter
7. Brian Kingsbury
8. Yifan Gong

In Meta & Continual Learning

1. Li Deng
2. Xiao Li
3. Gökhan Tür
4. Robert E. Schapire
5. Dilek Z. Hakkani-Tür
6. Bernard Ghanem

Citations in the corpus

• Libri-Light: A Benchmark for ASR with Limited or No Supervision (2019, neutral): “Active learning [25] could further select useful parts of the dataset (we have provided SNR data to facilitate this effort).”
• Active Learning for Efficient Audio Annotation and Classification with a Large Amount of Unlabeled Data (2019, background): “The effectiveness of AL has been shown in research areas such as natural language processing [9, 10], computer vision [11, 12], and speech recognition [13, 14].”
• Phrase-Based Statistical Language Generation Using Graphical Models and Active Learning (2010, background): “…parsing and information extraction (Thompson et al., 1999; Tang et al., 2002), spoken language understanding (Tur et al., 2003), speech recognition (Hakkani-Tür et al., 2002), word alignment (Sassano, 2002), and more recently for statistical machine translation (Bloodgood and Callison-Burch,…”
• Machine Learning Paradigms for Speech Recognition: An Overview (2013, methodology): “First, confidence measures is used as the criterion as in the standard uncertainty sampling method discussed earlier [180]–[182].”
• Multilingual representations for low resource speech recognition and keyword search (2015, background): “However, participants are allowed to automatically select two additional hours of audio using the one-hour set as a seed, for which manual transcripts will be provided [32–37].”
• What Do I Annotate Next? An Empirical Study of Active Learning for Action Localization (2018, background): “In the last decade, several active learning strategies have been proposed [27,42,63] and applied to several research fields, including speech recognition [32], natural language processing [62], chemistry [18], just to name a few.”
• Combining active and semi-supervised learning for spoken language understanding (2005, methodology): “In our previous work, we presented certainty-based active learning approaches for reducing the amount of labeled data needed for spoken language understanding (Tur et al., 2003); these techniques were also applied to automatic speech recognition by Hakkani-Tü r et al. (2002).”
• Active Learning for Sound Event Detection (2020, background): “In many cases, uncertainty sampling does not outperform random sampling when the labeling budget is low [10], [11].”
• Semi-supervised GMM and DNN acoustic model training with multi-system combination and confidence re-calibration (2013, background): “From the active learning point of view, these data are usually less valuable [9, 10, 11].”