PELCRA Spoken Offline Corpora
PELCRA Spoken Offline Corpora of conversational Polish (a.k.a SpokesMix) were collected as part of the CLARIN-PL project.
Each corpus consists of speech recordings (in WAV format) and word-by-word transcriptions, which also include some non-speech events. The transcriptions (in EAF format) are complemented with words and phones annotations ( _out.eaf files), and, if available, with video content (MP4 format) and PDF transcripts.
Metadata are provided in XML files listing information about the recordings (titles, topics, dates, and URLs), media available (audio, video, pdf), and annotation details (file, date, annotator, place, duration, with additional information about the speakers whenever such data was available). A Document Type Definition specifying the structure of the elements and attributes of an XML document is included in each of the corpora.
Most of these offline corpora are indexed in Spokes and Spokes2. A subset of them can be obtained by filling out this form. Once the form is submitted you will get a password necessary to download the corpora.
| corpus | description | recordings | speakers | word count | voice activity time (hh:mm) | total duration (hh:mm) |
|---|---|---|---|---|---|---|
| PELCRA_EMO | A corpus of focused interviews (people reflecting upon their emotions). | 40 | 80 | 252,000 | 26:53 | 28:12 |
| PELCRA_LUZ | A corpus of open interviews. | 21 | 42 | 213,000 | 20:14 | 19:58 |
| PELCRA_EMI | A corpus of Polish emmigrants to Scotland. | 22 | 44 | 96,000 | 09:36 | 18:07 |
| PELCRA_PARL | Samples of spoken parliamentary data. | 48 | 241 | 99,000 | 12:22 | 14:13 |
| PELCRA_YT_1 | Samples of Polish YouTubers' videos. | 25 | 106 | 49,000 | 04:56 | 06:39 |
| PELCRA_YT_2 | Second part of Polish YouTubers' videos. | 23 | 45 | 49,000 | 05:10 | 05:46 |
| MMW_1 | A corpus of Polish conversations recorded in Wrocław in the 1980s. | 14 | 65 | 60,000 | 7:02 | 8:33 |
| MMW_2 | Second part of the conversations recorded in Wrocław in the 1980s. | 14 | 38 | 70,000 | 7:31 | 7:50 |
| MMK | A corpus of Polish conversations recorded in Kraków in the 1980s. | 4 | 11 | 15,900 | 1:46 | 1:49 |
| PELCRA_IDIO | A corpus of open interviews in Polish. | 146 | 148 | 327,500 | : | 38:51 |
| TOTAL | 357 | 820 | 1,220,400 | 100:47 | 149:58 |
The following paper should be cited to fulfill the CC attribution condition of the license for these resources:
- Pęzik, Piotr. “Increasing the Accessibility of Time-Aligned Speech Corpora with Spokes Mix,” 4297–4300. Miyazaki, Japan, 2018. http://www.lrec-conf.org/proceedings/lrec2018/pdf/888.pdf.