Khresmoi

Medical Information Analysis & Retrieval
http://www.khresmoi.eu/

People and contacts

JH = Jan Hajič <hajic (at) ufal.mff.cuni.cz>
PP = Pavel Pecina <pecina (at) ufal.mff.cuni.cz>
JHla = Jaroslava Hlaváčová <hlava (at) ufal.mff.cuni.cz>
JD = Jan Dědek <dedek (at) ksi.mff.cuni.cz>
JB = Jakub Bystroň <jb.elitecode (at) gmail.com>
ZU = Zdeňka Urešová <uresova (at) ufal.mff.cuni.cz>

Data

MT training data available for KHRESMOI

Corpus	Source	Domain	EN-FR	EN-DE	alignment	EN	FR	DE	Note
TDA translation memory	TDA	in	13517 Kw	6797 Kw	sent				DONE
CESTA Evaluation Package	ELRA	in	38 Kw		sent				PROCESSING
EQueR Evaluation Package	ELRA	in					140 MiB		PROCESSING
CESART Evaluation Package	ELRA	in					9000 Kw		PROCESSING
French Gigaword	LDC	news					863 Kw		DVD
Acquis	JRC	law	1,25 Ms	1,33 Ms	sent				JHla (jen FR)
EMEA	European Medicines Agency	in	373 Ks	12 Mw		26.34 Mw	14.9Mw		DONE - i CS
MESH	U.S. National Library of Medicine	in				838 kw			DONE*
OrphaNet	OrphaNet	in	?						Wien will do
Europarl	WMT12	parl	1.8Ms	1.7Ms	sent				DONE
News Commentary	WMT12	news	43ks	60ks	sent				DONE
News monolingual	WMT12	news				181kw	147kw	162kw	DONE
United Nations	WMT12	un	12.3Ms						DONE
French-English 109 corpus	WMT12	web	22.5Ms		sent				DONE
Medpedia wiki	Medpedia	in				?			only EN found
Corpus Of Parallel Patent Applications (Coppa)	WIPO	in	24,8Mw = 1,2Ms		sent				JHla
Corpus Of Parallel Patent Applications (Coppa)	WIPO	in	33,5Mw		par				JHla
Corpus Of Parallel Patent Applications (Coppa)	WIPO	tech	153,8Mw = 7,5Ms		sent				JHla
Corpus Of Parallel Patent Applications (Coppa)	WIPO	tech	178,8Mw		par				JHla
MAREC	Wien TU	in	?	?	?				viz níže
Springer Bilingual Corpus	much.more	in		1.09 Mw	sent				JB
Europarl3	OPUS			1.3 Ms	sent				neni poreba
OpenSubtitles2011	OPUS			5 Ms	sent				JB

k, M … thousand, milion
w, s, f … words, sentences, files (for parallel data only source (English) words are counted)
* viz podrobnější info v podsekcích

Zdroje

JRC Acquis
http://optima.jrc.it/Acquis/JRC-Acquis.3.0/alignmentsHunAlign/index.html

MAREC
A61 (MEDICAL OR VETERINARY SCIENCE; HYGIENE): 1.589,849 files
Nevím, kolik slov, není to v jednolitém balíku.
Na žádost o přístup odpověděli:
the IRF is not granting access to the MAREC collection anymore. However, the access for research purposes should be possible in a foreseeable future via the Vienna University of Technology - Allan will certainly come back to you when the legal status is cleared.

Coppa
IPC: A61, C12N, C12P … medical patents (doporučeno od WIPO)
Patenty členěny podle roku, ve dvou verzích:

segmentované podle vět, ale menší - viz tabulka. Některé patenty chybí zcela, některé jsou zkráceny.
nesegmentované - každý patent má 2 záznamy: jméno a abstract, obojí v EN i FR, tedy alignment podle paragrafu (odhad)

Khresmoi wiki
http://wiki.khresmoi.eu/index.php5/Data_sets_used
http://wiki.khresmoi.eu/index.php5/Data_sets

www stranka WMT workshopu
http://www.statmt.org/wmt12/
http://www.statmt.org/wmt11/translation-task.html … tady je to vsecko pohromade

korpus OPUS
http://opus.lingfil.uu.se/

JRC Acquis
http://langtech.jrc.it/JRC-Acquis.html

ELDA

Objednali jsme několik balíčků s in-domain daty (EN-FR, FR)

TDA

Máme kredit na stažení 1 mld. slov. Zatím stažena EN-FR, EN-DE in-domain data.

LDC