46 datasets found

Language models have been commonly adopted to boost the performance of automatic speech recognition (ASR) particularly in domain adaptation tasks. Conventional way of LM...

The local and global features are both essential for automatic speech recognition (ASR). Many recent methods have verified that simply combining local and global features can...

The dataset used in this paper is a Conv1D equipped ASR model deployed on mobile devices.

Libriheavy is a large-scale ASR corpus consisting of 50,000 hours of read English speech derived from LibriVox. To the best of our knowledge, Libriheavy is the largest...

The ATIS dataset is a benchmark dataset for spoken language understanding, consisting of audio recordings and corresponding manual transcripts about humans asking for flight...

The TEDLIUM corpus is a large-volume corpus used for speech recognition and text summarization.

The How2 dataset consists of summarizations of How2 videos taken from YouTube.

Speech summarization, which generates a text summary from speech, can be achieved by combining automatic speech recognition (ASR) and text summarization (TS).

The LRS2 dataset consists of 48,164 video clips from outdoor shows on BBC television. Each video is accompanied by an audio corresponding to a sentence with up to 100 characters.

The LRS3 dataset is a large-scale dataset for visual speech recognition. It consists of thousands of spoken sentences from TED videos.

Open-domain neural dialogue generation (Vinyals and Le, 2015; Sordoni et al., 2015; Li et al., 2016a; Mou et al., 2016; Serban et al., 2016a; Asghar et al., 2016; Mei et al.,...

The dataset used in the paper is a collection of English ASR hypotheses from the eight submissions on the tst2012 test set in the IWSLT 2013 TED talk ASR track, along with...

The dataset used in the paper is the Libri-Light dataset, which is a subset of the LibriSpeech dataset. The authors used this dataset to pre-train their proposed dual-mode ASR...

The HKUST dataset is a large dataset of speech recordings, each containing a single speaker speaking a sentence.

The WSJ dataset is a large dataset of speech recordings, each containing a single speaker speaking a sentence.

The TIMIT acoustic-phonetic continuous speech corpusCD-ROM contains a large collection of speech samples from 250 male and 250 female speakers.

SpecAugment is a data augmentation method for automatic speech recognition, which masks the mel-spectrogram along the time and frequency axes.

MixSpeech is a data augmentation method for automatic speech recognition, which trains an ASR model by taking a weighted combination of two different speech features as the...

The WSJ corpus is a large vocabulary continuous speech recognition dataset. It contains 36416 sequences, representing around 80 hours of speech.

The 80-hour WSJ and 300-hour Switchboard datasets are used for end-to-end speech recognition.