The Design for the Wall Street Journal-based CSR Corpus

By D. Paul, J. Baker. 1992

The DARPA Spoken Language System (SLS) community has long taken a leadership position in designing, implementing, and globally distributing significant speech corpora widely used for advancing speech recognition research. The Wall Street Journal (WSJ) CSR Corpus described here is the newest addition to this valuable set of resources. In contrast to previous corpora, the WSJ corpus will provide DARPA its first general-purpose English, large vocabulary, natural language, high perplexity, corpus containing significant quantities of both speech data (400 hrs.) and text data (47M words), thereby providing a means to integrate speech recognition and natural language processing in application domains with high potential practical value. This paper presents the motivating goals, acoustic data design, text processing steps, lexicons, and testing paradigms incorporated into the multi-faceted WSJ CSR Corpus.

Subfields: Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

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

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

Imputer: Sequence Modelling via Imputation and Dynamic Programming (2020, methodology): “For WSJ, our Imputer model achieves 12.7 WER, while our non-autoregressive CTC baseline achieves 15.2 WER.”
Concrete Problems in AI Safety (2016, background): “a speech system that \knows when it is uncertain" could be one possible demonstration project. To be specic, the challenge could be: train a state-of-the-art speech system on a standard dataset [116] that gives well-calibrated results (if not necessarily good results) on a range of other test s”
Toward Domain-Invariant Speech Recognition via Large Scale Training (2018, methodology): “In [7], transfer learning is used to adapt a model trained on switchboard [15] to improve performance on domains with smaller training sets like WSJ [16] and AMI [17].”
Librispeech: An ASR corpus based on public domain audio books (2015, methodology): “The WSJ results we present in Table 2 are for the " open-vocabulary " (60K) test condition, using not the standard 60K word dictionary supplied with WSJ but an extended version that we built to cover more of the words that appear in the WSJ language models.”
Learning Robust and Multilingual Speech Representations (2020, positive): “We compare ASR systems on all of the Wall Street Journal and LibriSpeech corpora with the same optimization as explained above and evaluate word error rate on different evaluation sets, such as phone call conversations (Switchboard).”
An Unsupervised Autoregressive Model for Speech Representation Learning (2019, methodology): “To examine the robustness to shift in domains, rather than on the LibriSpeech test set, we conduct phone classification and speaker verification on the Wall Street Journal (WSJ) [28] and TIMIT corpora.”
Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention (2020, methodology): “We use the 80 hour WSJ dataset (Paul & Baker, 1992) with 40-dimensional mel-scale filterbanks without temporal differences as features.”
Fully Convolutional Speech Recognition (2018, methodology): “We evaluate our approach on the large vocabulary task of the Wall Street Journal (WSJ) dataset [26], which contains 80 hours of clean read speech, and Librispeech [27], which contains 1000 hours with separate train/dev/test splits for clean and noisy speech.”
Libriheavy: A 50,000 Hours ASR Corpus with Punctuation Casing and Context (2023, background): “A lot of work has been done on publishing open source datasets, for example, the Wall Street Journal corpus [5], SwitchBoard [6], Fisher [7] and the famous LibriSpeech corpus [8].”
Exploring Speech Enhancement with Generative Adversarial Networks for Robust Speech Recognition (2017, methodology): “We use the Wall Street Journal (WSJ) corpus [26] as our source of clean speech data.”
Unsupervised Learning of Disentangled and Interpretable Representations from Sequential Data (2017, background): “(2) Aurora-4 [32], a broadband corpus designed for noisy speech recognition tasks based on the Wall Street Journal corpus (WSJ0) [31].”
Achieving Human Parity in Conversational Speech Recognition (2016, methodology): “In the area of speech recognition, much of the pioneering early work was driven by a series of carefully designed tasks with DARPA-funded datasets publicly released by the LDC and NIST [6]: first simple ones like the “resource management” task [7] with a small vocabulary and carefully controlled gra”
ESPnet: End-to-End Speech Processing Toolkit (2018, methodology): “For example, the WER of the hybrid HMM/DNN systems for the WSJ task is below 5%, and this degradation probably comes from the lack of the amount of training data.”
Online and Linear-Time Attention by Enforcing Monotonic Alignments (2017, methodology): “We tested our approach on two datasets: TIMIT (Garofolo et al., 1993) and the Wall Street Journal corpus (Paul & Baker, 1992).”
Generative Pre-Training for Speech with Autoregressive Predictive Coding (2019, methodology): “We use WSJ for our SID experiments.”
Optimal Completion Distillation for Sequence Learning (2018, methodology): “We conduct our experiments on speech recogntion on the Wall Street Journal (WSJ) (Paul and Baker, 1992) and Librispeech (Panayotov et al., 2015) benchmarks.”