Papers with Code 아카이브

Speech Recognition

태스크의 하위 분류와 논문·코드·평가·기법·데이터셋을 탐색하세요.

하위 태스크
11
평가표
236
보존된 평가 기록
398

Speech Recognition is the task of converting spoken language into text. It involves recognizing the words spoken in an audio recording and transcribing them into a written format. The goal is to accurately transcribe the speech in real-time or from recorded audio, taking into account factors such as accents, speaking speed, and background noise.

<span style="color:grey; opacity: 0.6">( Image credit: SpecAugment )</span>

데이터셋

출처에 같은 태스크 이름으로 연결된 데이터셋입니다. 동명 태스크가 포함될 수 있으므로, 비교할 때는 평가 결과의 조건도 확인하세요.

SLUE

Spoken Language Understanding Evaluation (SLUE) is a suite of benchmark tasks for spoken language understanding evaluation. It consists of limited-size labeled training sets and corresponding evaluation sets. This resource would allow the research community to track progress, evaluate pre-trained representations for higher-level tasks, and study open questions such as the utility of pipeline versus end-to-end approaches. The first phase of the SLUE benchmark suite consists of named entity recognition (NER), sentiment analysis (SA), and ASR on the corresponding datasets. Corpus includes: - SLUE-VoxPopuli: consists of ASR and NER tasks - CC0 license - SLUE-VoxCeleb: consists of ASR and SA tasks - CCBY 4.0 license

DIRHA

DIRHA-English is a multi-microphone database composed of real and simulated sequences of 1-minute. The overall corpus is composed of different types of sequences including: 1) Phonetically-rich sentences; 2) WSJ 5-k utterances; 3) WSJ 20-k utterances; 4) Conversational speech (also including keywords and commands). The sequences are available for both UK and US English at 48 kHz. The DIRHA-English dataset offers the possibility to work with a very large number of microphone channels, to use of microphone arrays having different characteristics and to work considering different speech recognition tasks (e.g., phone-loop, keyword spotting, ASR with small and very large language models). Source: The DIRHA-English Corpus Image Source: https://arxiv.org/pdf/1710.02560v1.pdf

Taskmaster-1

Taskmaster-1 is a dialog dataset consisting of 13,215 task-based dialogs in English, including 5,507 spoken and 7,708 written dialogs created with two distinct procedures. Each conversation falls into one of six domains: ordering pizza, creating auto repair appointments, setting up ride service, ordering movie tickets, ordering coffee drinks and making restaurant reservations. Image Source: https://arxiv.org/pdf/1909.05358v1.pdf

DiPCo

We present a speech data corpus that simulates a "dinner party" scenario taking place in an everyday home environment. The corpus was created by recording multiple groups of four Amazon employee volunteers having a natural conversation in English around a dining table. The participants were recorded by a single-channel close-talk microphone and by five far-field 7-microphone array devices positioned at different locations in the recording room. The dataset contains the audio recordings and human labeled transcripts of a total of 10 sessions with a duration between 15 and 45 minutes. The corpus was created to advance in the field of noise robust and distant speech processing and is intended to serve as a public research and benchmarking data set.

SPGISpeech

SPGISpeech (pronounced “speegie-speech”) is a large-scale transcription dataset, freely available for academic research. SPGISpeech is a collection of 5,000 hours of professionally-transcribed financial audio. Contrary to previous transcription datasets, SPGISpeech contains global english accents, strongly varying audio quality as well as both spontaneous and presentation style speech. The transcripts have each been cross-checked by multiple professional editors for high accuracy and are fully formatted including sentence structure and capitalization. SPGISpeech consists of 5,000 hours of recorded company earnings calls and associated manual transcription text. The original calls were split based on silences into slices ranging from 5 to 15 seconds to allow easy training of a speech recognition system. The format of each WAV file is single channel, 16kHz, 16 bit audio. Transcription text represents the output of several stages of manual post-processing. As such, the text contains polished English orthography following a detailed style guide, including proper casing, punctuation, and denormalized non-standard words such as numbers or acronyms, making SPGISpeech suited for training fully formatted end-to-end models. In general, the transcriptions aim at professional utility rather than linguistic fidelity, and the correspondence between verbatim speech and finalized text is therefore not exact, resulting in the occasional purposeful omission of meeting operator instructions or certain verbal pleasantries.

MaSS

MaSS (Multilingual corpus of Sentence-aligned Spoken utterances) is an extension of the CMU Wilderness Multilingual Speech Dataset, a speech dataset based on recorded readings of the New Testament. MaSS extends it by providing a large and clean dataset of 8,130 parallel spoken utterances across 8 languages (56 language pairs). The covered languages are: Basque, English, Finnish, French, Hungarian, Romanian, Russian and Spanish. Source: MaSS: A Large and Clean Multilingual Corpus of Sentence-aligned Spoken Utterances Extracted from the Bible Image Source: https://arxiv.org/pdf/1907.12895v3.pdf

SMS-WSJ

Spatialized Multi-Speaker Wall Street Journal (SMS-WSJ) consists of artificially mixed speech taken from the WSJ database, but unlike earlier databases this one considers all WSJ0+1 utterances and takes care of strictly separating the speaker sets present in the training, validation and test sets. Source: SMS-WSJ: Database, performance measures, and baseline recipe for multi-channel source separation and recognition

MagicData-RAMC

The MagicData-RAMC corpus contains 180 hours of conversational speech data recorded from native speakers of Mandarin Chinese over mobile phones with a sampling rate of 16 kHz. The dialogs in the dialogs are classified into 15 diversified domains and tagged with topic labels, ranging from science and technology to ordinary life. Accurate transcription and precise speaker voice activity timestamps are manually labeled for each sample. Speakers' detailed information is also provided.

CALLHOME American English Speech

The CALLHOME English Corpus is a collection of unscripted telephone conversations between native speakers of English. Here are the key details: Participants: 120 individuals. Type of Study: Naturalistic. Location: USA. Media Type: Audio. DOI: doi:10.21415/T5KP54. Contents: The corpus contains 120 telephone conversations, each lasting up to 30 minutes. Origins: All calls originated in North America, with 90 calls placed to various locations overseas and 30 calls within North America. Transcripts: The transcripts cover contiguous 5 or 10-minute segments from recorded conversations. Speaker Awareness: All speakers were aware that they were being recorded. Topics: Participants had no guidelines on what to talk about; most called family members or close friends overseas. Purpose: Collected primarily to support the project on Large Vocabulary Conversational Speech Recognition (LVCSR), sponsored by the U.S. Department of Defense.

SOMOS

The SOMOS dataset is a large-scale mean opinion scores (MOS) dataset consisting of solely neural text-to-speech (TTS) samples. It can be employed to train automatic MOS prediction systems focused on the assessment of modern synthesizers, and can stimulate advancements in acoustic model evaluation. It consists of 20K synthetic utterances of the LJ Speech voice, a public domain speech dataset which is a common benchmark for building neural acoustic models and vocoders. Utterances are generated from 200 TTS systems including vanilla neural acoustic models as well as models which allow prosodic variations.

TUDA

Overall duration per microphone: about 36 hours (31 hrs train / 2.5 hrs dev / 2.5 hrs test) Count of microphones: 3 (Microsoft Kinect, Yamaha, Samson) Count of wave-files per microphone: about 14500 Overall count of participations: 180 (130 male / 50 female)

VoxForge

VoxForge is an open speech dataset that was set up to collect transcribed speech for use with Free and Open Source Speech Recognition Engines (on Linux, Windows and Mac). Image Source: http://www.voxforge.org/home

GRID Dataset

The QMUL underGround Re-IDentification (GRID) dataset contains 250 pedestrian image pairs. Each pair contains two images of the same individual seen from different camera views. All images are captured from 8 disjoint camera views installed in a busy underground station. The figures beside show a snapshot of each of the camera views of the station and sample images in the dataset. The dataset is challenging due to variations of pose, colours, lighting changes; as well as poor image quality caused by low spatial resolution. Source: https://personal.ie.cuhk.edu.hk/~ccloy/downloads_qmul_underground_reid.html

SpeakingFaces

SpeakingFaces is a publicly-available large-scale dataset developed to support multimodal machine learning research in contexts that utilize a combination of thermal, visual, and audio data streams; examples include human-computer interaction (HCI), biometric authentication, recognition systems, domain transfer, and speech recognition. SpeakingFaces is comprised of well-aligned high-resolution thermal and visual spectra image streams of fully-framed faces synchronized with audio recordings of each subject speaking approximately 100 imperative phrases. Source: SpeakingFaces: A Large-Scale Multimodal Dataset of Voice Commands with Visual and Thermal Video Streams

CAS-VSR-W1k (LRW-1000)

LRW-1000 has been renamed as CAS-VSR-W1k.* It is a naturally-distributed large-scale benchmark for word-level lipreading in the wild, including 1000 classes with about 718,018 video samples from more than 2000 individual speakers. There are more than 1,000,000 Chinese character instances in total. Each class corresponds to the syllables of a Mandarin word which is composed by one or several Chinese characters. This dataset aims to cover a natural variability over different speech modes and imaging conditions to incorporate challenges encountered in practical applications. Source: VIPL Image Source: https://arxiv.org/pdf/1810.06990v6.pdf

SPEECH-COCO

SPEECH-COCO contains speech captions that are generated using text-to-speech (TTS) synthesis resulting in 616,767 spoken captions (more than 600h) paired with images. Source: SPEECH-COCO: 600k Visually Grounded Spoken Captions Aligned to MSCOCO Data Set

ADIMA

ADIMA is a novel, linguistically diverse, ethically sourced, expert annotated and well-balanced multilingual profanity detection audio dataset comprising of 11,775 audio samples in 10 Indic languages spanning 65 hours and spoken by 6,446 unique users.

Europarl-ASR

Europarl-ASR (EN) is a 1300-hour English-language speech and text corpus of parliamentary debates for (streaming) Automatic Speech Recognition training and benchmarking, speech data filtering and speech data verbatimization, based on European Parliament speeches and their official transcripts (1996-2020). Includes dev-test sets for streaming ASR benchmarking, made up of 18 hours of manually revised speeches. The availability of manual non-verbatim and verbatim transcripts for dev-test speeches makes this corpus also useful for the assessment of automatic filtering and verbatimization techniques. The corpus is released under an open licence at https://www.mllp.upv.es/europarl-asr/ Europarl-ASR CONTENTS: [Speech data] 1300 hours of English-language annotated speech data, 3 full sets of timed transcriptions (official non-verbatim, automatically noise-filtered, automatically verbatimized), 18 hours of speech data with both manually revised verbatim transcriptions and official non-verbatim transcriptions, split in 2 independent validation- evaluation partitions for 2 realistic ASR tasks (with vs. without previous knowledge of the speaker); [Text data] 70 million tokens of English-language text data; [Pretrained language models] the Europarl-ASR English-language n-gram language model and vocabulary.

JSUT Corpus

JSUT Corpus is a free large-scale speech corpus that can be shared between academic institutions and commercial companies has an important role. However, such a corpus for Japanese speech synthesis does not exist. Source: JSUT corpus: free large-scale Japanese speech corpus for end-to-end speech synthesis

ToN_IoT

The TON_IoT datasets are new generations of Internet of Things (IoT) and Industrial IoT (IIoT) datasets for evaluating the fidelity and efficiency of different cybersecurity applications based on Artificial Intelligence (AI). The datasets have been called ‘ToN_IoT’ as they include heterogeneous data sources collected from Telemetry datasets of IoT and IIoT sensors, Operating systems datasets of Windows 7 and 10 as well as Ubuntu 14 and 18 TLS and Network traffic datasets. The datasets were collected from a realistic and large-scale network designed at the IoT Lab of the UNSW Canberra Cyber, the School of Engineering and Information technology (SEIT), UNSW Canberra @ the Australian Defence Force Academy (ADFA). The datasets were gathered in a parallel processing to collect several normal and cyber-attack events from IoT networks. A new testbed was developed at the IoT lab to connect many virtual machine, physical systems, hacking platforms, cloud and fog platforms, IoT and IIoT sensors to mimic the complexity and scalability of industrial IoT and Industry 4.0 networks. Different hacking techniques, such as DoS, DDoS and ransomware against, were launched against web applications, IoT gateways and computer systems across the IIoT network.

Gumar Corpus

A large-scale corpus of Gulf Arabic consisting of 110 million words from 1,200 forum novels. Source: A Large Scale Corpus of Gulf Arabic

VIVOS

VIVOS is a free Vietnamese speech corpus consisting of 15 hours of recording speech prepared for Automatic Speech Recognition task. The corpus was prepared by AILAB, a computer science lab of VNUHCM - University of Science, with Prof. Vu Hai Quan is the head of. We publish this corpus in hope to attract more scientists to solve Vietnamese speech recognition problems. The corpus should only be used for academic purposes.

EmoDB Dataset

The EMODB database is the freely available German emotional database. The database is created by the Institute of Communication Science, Technical University, Berlin, Germany. Ten professional speakers (five males and five females) participated in data recording. The database contains a total of 535 utterances. The EMODB database comprises of seven emotions: 1) anger; 2) boredom; 3) anxiety; 4) happiness; 5) sadness; 6) disgust; and 7) neutral. The data was recorded at a 48-kHz sampling rate and then down-sampled to 16-kHz. Citation: Felix Burkhardt, Astrid Paeschke, Miriam Rolfes, Walter Sendlmeier und Benjamin Weiss A Database of German Emotional Speech Proceedings Interspeech 2005, Lissabon, Portugal

GLips

The German Lipreading dataset consists of 250,000 publicly available videos of the faces of speakers of the Hessian Parliament, which was processed for word-level lip reading using an automatic pipeline. The format is similar to that of the English language Lip Reading in the Wild (LRW) dataset, with each H264-compressed MPEG-4 video encoding one word of interest in a context of 1.16 seconds duration, which yields compatibility for studying transfer learning between both datasets. Choosing video material based on naturally spoken language in a natural environment ensures more robust results for real-world applications than artificially generated datasets with as little noise as possible. The 500 different spoken words ranging between 4-18 characters in length each have 500 instances and separate MPEG-4 audio- and text metadata-files, originating from 1018 parliamentary sessions. Additionally, the complete TextGrid files containing the segmentation information of those sessions are also included. The size of the uncompressed dataset is 15GB.

Open Images V7

Open Images is a computer vision dataset covering ~9 million images with labels spanning thousands of object categories. A subset of 1.9M includes diverse annotations types. - 15,851,536 boxes on 600 classes - 2,785,498 instance segmentations on 350 classes - 3,284,280 relationship annotations on 1,466 relationships - 675,155 localized narratives (synchronized voice, mouse trace, and text caption) - 66,391,027 point-level annotations on 5,827 classes - 61,404,966 image-level labels on 20,638 classes Images are under a CC BY 2.0 license, annotations under CC BY 4.0 license.

AccentDB

AccentDB is a database that contains samples of 4 Indian-English accents, and a compilation of samples from 4 native-English, and a metropolitan Indian-English accent.

ADL Piano MIDI

The ADL Piano MIDI is a dataset of 11,086 piano pieces from different genres. This dataset is based on the Lakh MIDI dataset, which is a collection on 45,129 unique MIDI files that have been matched to entries in the Million Song Dataset. Most pieces in the Lakh MIDI dataset have multiple instruments, so for each file the authors of ADL Piano MIDI dataset extracted only the tracks with instruments from the "Piano Family" (MIDI program numbers 1-8). This process generated a total of 9,021 unique piano MIDI files. Theses 9,021 files were then combined with other approximately 2,065 files scraped from publicly-available sources on the internet. All the files in the final collection were de-duped according to their MD5 checksum. Source: ADL Piano MIDI

Artie Bias Corpus

Artie Bias Corpus is an open dataset for detecting demographic bias in speech applications.

Blizzard Challenge 2013

The English data for voice building was obtained, prepared and provided the the challenge by Lessac Technologies Inc., having originally came from the publishers Voice Factory International Inc. It comprises speech from one female professional narrator & actress, Catherine ‘Bobbie’ Byers, reading the text of a collection of classic novels. These had been divided by the publishers of the original audiobooks into a number of genres, such as “Classic Novels”, “Women’s Classics”, “Young Readers” and so on.

ClovaCall

ClovaCall is a new large-scale Korean call-based speech corpus under a goal-oriented dialog scenario from more than 11,000 people. The raw dataset of ClovaCall includes approximately 112,000 pairs of a short sentence and its corresponding spoken utterance in a restaurant reservation domain. Source: https://github.com/ClovaAI/ClovaCall

CSRC

CSRC is a collection of data for Children Speech Recognition. The data for this challenge is divided into 3 datasets, referred to as A (Adult speech training set), C1 (Children speech training set) and C2 (Children conversation training set). All dataset combined amount to 400 hours of Mandarin speech data.

Libri-Adapt

Libri-Adapt aims to support unsupervised domain adaptation research on speech recognition models. Source: Libri-Adapt: A New Speech Dataset for Unsupervised Domain Adaptation

MediaSpeech

MediaSpeech is a media speech dataset (you might have guessed this) built with the purpose of testing Automated Speech Recognition (ASR) systems performance. The dataset consists of short speech segments automatically extracted from media videos available on YouTube and manually transcribed, with some pre- and post-processing. The dataset contains 10 hours of speech for each language provided. This release contains audio datasets in French, Arabic, Turkish and Spanish, and is a part of a larger private dataset. Source: MediaSpeech: Multilanguage ASR Benchmark and Dataset

OLR 2021

The OLR 2021 dataset contains the data for the Oriental Language Recognition (OLR) 2021 Challenge, which intends to improve the performance of language recognition systems and speech recognition systems within multilingual scenarios.

출처: Papers with Code · CC BY-SA 4.0. 아카이브 자료와 새로 수집한 논문은 별도로 관리합니다.