Companies / Data vendor
M47 AI Software
Spanish data vendor (M47 Labs) that wrote and annotated Catalan training data to order for the Barcelona Supercomputing Center. Recorded here: 5 public prices.
m47labs.comLatest record 21 Dec 2022
Public prices involving M47 AI Software
- Segmentation, alignment and translation of a speech corpus, to order, for language-model training (Barcelona Supercomputing Center, Aina project, lot 9)
A public research centre's award to M47 AI Software for speech-corpus work; three tenders. Neither the notice nor a published BSC dataset card names the hours, so there is no per-hour figure.
The payer's own page, TED (EU Official Journal), contract award notice 2022/S 250-731229: Barcelona Supercomputing Center, creation and annotation of datasets to train and evaluate language models (CONSER02022005OP)
EUR 264,000 (lot 9; estimate EUR 295,200)per lot (4 months, one 4-month extension possible)
- CoQCat: 89,364 Catalan conversational question-answer pairs over 6,000 text passages, written to order on the CoQA model (Barcelona Supercomputing Center, Aina project)
A public research centre paid M47 AI Software for annotators working in pairs (one asks, one answers) plus a third person for extra answers on the dev and test splits; the card says the annotation was entrusted to M47 labs through a public tender process. Per text passage the same award is EUR 14.00 (84,000 / 6,000).
The payer's own page, TED (EU Official Journal), contract award notice 2022/S 250-731229: Barcelona Supercomputing Center, creation and annotation of datasets to train and evaluate language models (CONSER02022005OP)
EUR 84,000 (lot 3; estimate EUR 90,000)per lot (89,364 question-answer pairs)
- Annotation and linking of named entities (NERC and entity linking) in Catalan text for language models (Barcelona Supercomputing Center, Aina project, lot 2)
Won by M47 AI Software against 3 other tenders. The notice states no count of documents or entities.
The payer's own page, TED (EU Official Journal), contract award notice 2022/S 250-731229: Barcelona Supercomputing Center, creation and annotation of datasets to train and evaluate language models (CONSER02022005OP)
EUR 126,300 (lot 2; estimate EUR 139,200)per lot (4 months, one 4-month extension possible)
- NLUCat: nearly 12,000 Catalan voice-assistant instructions written to order and annotated with intents and spans, for training and evaluating language models (Barcelona Supercomputing Center, Aina project)
A public research centre paid M47 AI Software (M47 Labs) to write and annotate the examples; three tenders, won about 6% under the lot estimate. The dataset card's split table totals 11,997 examples (9,117 train, 1,440 dev, 1,440 test), so the per-example price is all-in for writing, correction and annotation.
The payer's own page, TED (EU Official Journal), contract award notice 2022/S 250-731229: Barcelona Supercomputing Center, creation and annotation of datasets to train and evaluate language models (CONSER02022005OP)
EUR 18,836.49 (lot 5; estimate EUR 20,113.38)per lot (11,997 written and annotated examples)
- Annotation of polarity, emotions and opinion in Catalan text for language models (Barcelona Supercomputing Center, Aina project, lot 4)
Won by M47 AI Software; three tenders. The notice states no count. The BSC later published three Catalan sentiment and emotion sets (CaSSA 6,400 texts, CaSET about 11,000 sentences, CaSERa 15,782 sentences) whose cards name no vendor, so no per-unit figure is drawn.
The payer's own page, TED (EU Official Journal), contract award notice 2022/S 250-731229: Barcelona Supercomputing Center, creation and annotation of datasets to train and evaluate language models (CONSER02022005OP)
EUR 89,000 (lot 4; estimate EUR 93,000)per lot (4 months, one 4-month extension possible)
Cite this
AIData.Broker, “M47 AI Software in the AI training data market”, reviewed 21 Dec 2022. https://aidata.broker/companies/m47-ai-software
Free to reuse under CC BY 4.0 if you credit AIData.Broker and link to this page. Each row names the publication that stated it, and words quoted from a source keep that source's own terms.