This is the new version of the ELRA Catalogue of Language Resources. If you would like to view the older version, click here

An increasing number of LRs in the various fields of Human Language Technology (see image on the left-hand side) are distributed on behalf of ELRA via its operational body ELDA, thanks to the contribution of various players of the HLT community.
Our aim is to provide Language Resources, by means of this repository, so as to prevent researchers and developers from investing efforts to rebuild resources which already exist as well as help them identify and access those resources.

Latest Resources

• Arabic Speech Corpus

  This speech corpus has been developed as part of a PhD work carried out by Nawar Halabi at the University of Southampton. The corpus was recorded through a Neumann TLM 103 Studio Microphone by one male speaker in South Levantine Arabic (Damascian accent) in a professional studio. The transcript was ...

• Ahoslabi - esophageal speech database

  Ahoslabi was built within the frame of the RESTORE project (“Restauración, almacenamiento y rehabilitación de la voz”) (restrictions apply) and has received funding from Spanish Ministry of Economy and Competitiveness with FEDER support (RESTORE project, TEC2015-67163- C2-1-R), the Basque Government (PIBA-018-0035) and by the European Union's H2020 research and innovation ...

• Japanese Kids Speech database (Lower Grade)

  The Japanese Kids Speech database (Lower Grade) contains the total recordings of 179 Japanese Kids speakers (71 males and 108 females), from 6 to 9 years' old (first, second and third graders in elementary school), recorded in quiet rooms using smartphones. This database may be combined with the Japanese Kids ...

• Japanese Kids Speech database (Upper Grade)

  The Japanese Kids Speech database (Upper Grade) contains the total recordings of 232 Japanese Kids speakers (104 males and 128 females), from 9 to 13 years’ old (fourth, fifth and sixth graders in elementary school), recorded in quiet rooms using smartphones. This database may be combined with the Japanese Kids ...

• CAREGIVER Corpus

  A multi-lingual speech corpus used for modeling language acquisition called CAREGIVER has been designed and recorded within the framework of the EU funded Acquisition of Communication and Recognition Skills (ACORNS) project. The motivation behind the corpus and its design relies on current knowledge regarding infant language acquisition. Instead of recording ...

• MDT Mandarin Chinese Conversational Recognition Corpus – Complete set

  This dataset consists of 4.98 hours of transcribed conversational speech in Mandarin Chinese, where 30 conversations are uttered by 32 speakers (16 males and 16 females). The audios are sampled at 16 kHz and quantized at 16 bits. For each conversation, there are two close-talking channels recorded via the microphones, ...

• MDT Mandarin Chinese Conversational Recognition Corpus – 3 channels

  This dataset consists of 4.98 hours of transcribed conversational speech in Mandarin Chinese, where 30 conversations are uttered by 32 speakers (16 males and 16 females). The audios are sampled at 16 kHz and quantized at 16 bits. For each conversation, there are two close-talking channels recorded via the microphones, ...

• MDT Mandarin Chinese Conversational Recognition Corpus – 1 channel

  This dataset consists of 4.98 hours of transcribed conversational speech in Mandarin Chinese, where 30 conversations are uttered by 32 speakers (16 males and 16 females). The audios are sampled at 16 kHz and quantized at 16 bits. For each conversation, there are two close-talking channels recorded via the microphones, ...

• MDT Mandarin Chinese Conversational Recognition Corpus – 2 channels

  This dataset consists of 4.98 hours of transcribed conversational speech in Mandarin Chinese, where 30 conversations are uttered by 32 speakers (16 males and 16 females). The audios are sampled at 16 kHz and quantized at 16 bits. For each conversation, there are two close-talking channels recorded via the microphones, ...

• Portuguese Speech Recognition Corpus (Desktop)

  This corpus comprises 49,988 entries uttered by 50 speakers (26 males and 24 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 26.41 hours of speech per channel.

• American English Speech Recognition Corpus (Mobile)

  This corpus comprises 14,988 entries uttered by 50 speakers (23 males and 27 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 14.67 hours of speech.

• Spain Spanish Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 9,920 entries uttered by 31 speakers (16 males and 15 females), recorded over 2 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 4.75 hours of speech per channel.

• Italian Speech Recognition Corpus (Desktop)

  This corpus comprises 49,994 entries uttered by 50 speakers (23 males and 27 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 24.21hours of speech per channel.

• Chinese Mandarin Speech Recognition Corpus (Mobile)

  This corpus comprises 60,216 entries uttered by 201 speakers (101 males and 100 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16kHz for a total of 85 hours of speech.

• Japanese Speech Recognition Corpus (Mobile)

  This corpus comprises 16,792 entries uttered by 56 speakers (29 males and 27 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 19.4 hours of speech.

• German Speech Recognition Corpus (Desktop)

  This corpus comprises 51,912 entries uttered by 52 speakers (26 males and 26 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 23.9 hours of speech per channel.

• Canadian English Speech Recognition Corpus (Telephone) - place name

  This corpus comprises 2,400 entries uttered by 150 speakers (106 males and 44 females), recorded over the telephone network. Speech samples are stored as a sequence of 16-bit 8 kHz for a total of 2.24 hours of speech.

• Taiwanese Speech Recognition Corpus (Desktop)

  This corpus comprises 107,924 entries uttered by 54 speakers (27 males and 27 females), recorded over 4 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 43.78 hours of speech per channel.

• Chinese Mandarin Speech Recognition Corpus (Mobile)

  This corpus comprises 91,729 entries uttered by 304 speakers (151 males and 153 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16kHz for a total of 67.4 hours of speech.

• American English Speech Recognition Corpus (Desktop)

  This corpus comprises 49,990 entries uttered by 50 speakers (25 males and 25 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 16kHz for a total of 24.9 hours of speech per channel.

• German Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 9,572 entries uttered by 30 speakers (15 males and 15 females), recorded over 2 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 4.25 hours of speech per channel.

• Korean Speech Recognition Corpus (Desktop+Mobile)

  This corpus comprises 32,247 entries uttered by 52 speakers (26 males and 26 females), recorded over 3 channels (desktop and mobile in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 15.76 hours of speech per channel.

• Brazilian Portuguese Speech Recognition Corpus (Desktop)

  This corpus comprises 99,804 entries uttered by 50 speakers (25 males and 25 females), recorded over 4 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 37.3 hours of speech per channel.

• Hong Kong Cantonese Speech Recognition Corpus (Desktop)

  This corpus comprises 101,964 entries uttered by 51 speakers, recorded over 4 channels (desktop). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 24.18 hours of speech per channel.

• Russian Speech Kids Recognition Corpus (Desktop)

  This corpus comprises 19,164 entries uttered by 30 speakers (16 males and 14 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 4.15 hours of speech per channel.

• Mexican Spanish Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 19,156 entries uttered by 30 speakers (16 males and 14 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 5 hours of speech per channel.

• Japanese Speech Recognition Corpus (Desktop) - sentences (200 people)

  This corpus comprises 7,996 entries uttered by 200 speakers (93 males and 107 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 3.25 hours of speech per channel. This corpus is partly included in ELRA-S0228-54.

• Canadian English Speech Recognition Corpus (Telephone) - person name

  This corpus comprises 2,250 entries uttered by 150 speakers (106 males and 44 females), recorded over the telephone network. Speech samples are stored as a sequence of 16-bit 8 kHz for a total of 2.83 hours of speech.

• Korean Speech Recognition corpus (Desktop) - name, digit string, place, sentences

  This corpus comprises 83,756 entries uttered by 150 speakers (66 males and 84 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 29.65 hours of speech per channel. This set combines ELRA-S0228-50, ELRA-S0228-51, ELRA-S0228-52 and ELRA-S0228-53 ...

• Australian English Speech Recognition Corpus (Mobile)

  This corpus comprises 24,874 entries uttered by 50 speakers (23 males and 27 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 26.7 hours of speech.

• Canadian English Speech Recognition Corpus (Telephone) - sentences

  This corpus comprises 1,500 entries uttered by 150 speakers (106 males and 44 females), recorded over the telephone network. Speech samples are stored as a sequence of 16-bit 8 kHz for a total of 2.09 hours of speech.

• Japanese Speech Recognition Corpus (Desktop)

  This corpus comprises 97,908 entries uttered by 50 speakers (26 males and 23 females), recorded over 4 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 28.31 hours of speech per channel.

• British English Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 19,196 entries uttered by 30 speakers (15 males and 15 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 3.65 hours of speech per channel.

• Chinese Mandarin Speech Recognition Corpus (Mobile)

  This corpus comprises 120,144 entries uttered by 400 speakers (199 males and 201 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 204.2 hours of speech.

• American English Conversational Speech Recognition Corpus (Multi-Channel)

  This corpus was recorded by 20 speakers (10 males and 10 females), over 7 channels (multi-channel in quiet office/home). Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 10 hours of speech per channel.

• Australian English Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 9,596 entries uttered by 30 speakers (15 males and 15 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 5 hours of speech per channel.

• American/Canadian English Speech Recognition Corpus (headset+mobile)

  This corpus comprises 12,974 entries uttered by 30 speakers (15 males and 15 females), recorded over 2 channels (headset and mobile in noisy restaurant/shopping mall/info center/hospital/station/car). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 12 hours of speech per channel.

• Canadian English Speech Recognition Corpus (Desktop)

  This corpus comprises 6,976 entries uttered by 150 speakers (80 males and 70 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 3.86 hours of speech per channel.

• Spain Spanish Speech Recognition Corpus (Desktop)

  This corpus comprises 49,998 entries uttered by 50 speakers (28 males and 22 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 23.64 hours of speech per channel.

• British English Speech Recognition Corpus (Mobile)

  This corpus comprises 63,495 entries uttered by 54 speakers (27 males and 27 females), recorded over 3 channels (mobile in noisy café/restaurant/street). Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 22.3 hours of speech per channel.

• British English Speech Recognition Corpus (Desktop)

  This corpus comprises 50,858 entries uttered by 51 speakers (28 males and 23 females), recorded over 2 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 29.7 hours of speech per channel.

• Canadian English Speech Recognition Corpus (Telephone) - spell words

  This corpus comprises 1,500 entries uttered by 150 speakers (106 males and 44 females), recorded over the telephone network. Speech samples are stored as a sequence of 16-bit 8 kHz for a total of 3.6 hours of speech.

• Italian Kids Speech Recognition Corpus (Desktop)

  This corpus comprises 19,788 entries uttered by 31 speakers (15 males and 16 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 4.9 hours of speech per channel.

• American Spanish Recognition Corpus (Desktop+Mobile)

  This corpus comprises 33,527 entries uttered by 40 speakers (21 males and 19 females), recorded over 2 channels (desktop in quiet office and mobile in noisy restaurant). Speech samples are stored as a sequence of 16-bit 16kHz for a total of 14.7 hours of speech per channel.

• Canadian French Speech Recognition Corpus (Mobile)

  This corpus comprises 75,147 entries uttered by 50 speakers (25 males and 25 females), recorded over 3 channels (mobile quiet office). Speech samples are stored as a sequence of 16-bit 16kHz for a total of 25.67 hours of speech per channel.

• Japanese English Speech Recognition corpus(Mobile)

  This corpus comprises 219,139 entries uttered by 402 speakers (200 males and 202 females), recorded over the mobile telephone network over 2 channels. Speech samples are stored as a sequence of 16-bit 16kHz for a total of 172.9 hours of speech per channel.

• Russian Speech Recognition Corpus (Desktop)

  This corpus comprises 59,968 entries uttered by 50 speakers (25 males and 25 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 25.85 hours of speech per channel.

• Chinese English Speech Recognition Corpus (Desktop)

  This corpus comprises 30,076 entries uttered by 100 speakers (48 males and 52 females), recorded over desktop in quiet office. Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 60.28 hours of speech per channel.

• Australian English Speech Recognition Corpus (Desktop)

  This corpus comprises 99,624 entries uttered by 51 speakers (21 males and 30 females), recorded over 4 channels (desktop in quiet office/home). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 27 hours of speech per channel.

• Korean English Speech Recognition Corpus (Mobile)

  This corpus comprises 139011 entries uttered by 116 speakers (63 males and 53 females), recorded over 3 channels (mobile). Speech samples are stored as a sequence of 16-bit 16 kHz for a total of 68.9 hours of speech per channel.

• France French Speech Recognition Corpus (Desktop)

  This corpus comprises 49,982 entries uttered by 50 speakers (28 males and 22 females), recorded over 2 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 48kHz for a total of 22.06 hours of speech per channel.

• Russian Speech Recognition Corpus (Desktop)

  This corpus comprises 99,940 entries uttered by 50 speakers (25 males and 25 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 32.13 hours of speech per channel.

• Japanese English Speech Recognition Corpus (Desktop)

  This corpus comprises 240,296 entries uttered by 201 speakers (100 males and 101 females), recorded over 4 channels (desktop in quiet office). Speech samples are stored as a sequence of 16-bit 44.1kHz for a total of 95.63 hours of speech per channel.

• American English Speech Recognition Corpus (Mobile)

  This corpus comprises 39,243 entries uttered by 151 speakers (74 males and 77 females), recorded over the mobile telephone network. Speech samples are stored as a sequence of 16-bit 16kHz for a total of 19.4 hours of speech.