Any help on finding the biggest freely available English corpus that can be used on research?
So far I have found OANC with 15 M words.
-
Does it matter what language it's in?– user6726Commented May 10, 2015 at 21:26
-
1Take a look at the Global Web-Based English Corpus. 1.9 billion words.– robertCommented May 10, 2015 at 22:42
-
Yes English, we want to use it for letter frequency studies. The online ones usually don't let you search on letters– Tt22Commented May 11, 2015 at 4:03
-
Are you just trying to estimate letter frequencies? If so, I think a much smaller corpus will be fine.– Russell RichieCommented Feb 25, 2016 at 5:59
-
@robert While that corpus is searchable, the content is not freely available. Each corpus at that site costs hundreds of dollars.– MarkCommented Jan 27, 2018 at 0:15
Add a comment
|
7 Answers
Sketch Engine, a corpus manager and text analysis software, provides a few corpora with open access for research on https://app.sketchengine.eu/#open The largest English corpus (freely available) is ACL Anthology Reference Corpus with 62 million words.
On the other hand, you can try 30-day free trial of Sketch Engine and search one of the biggest English corpus which currently exists with over 35 billion words, see more at https://www.sketchengine.eu/timestamped-english-corpus/
Can't beat the Global Web-Based English Corpus proposed by robert---but here are is another big one:
- British national corpus (BNC) http://corpus.byu.edu/bnc/ 100 mio words
A Wikipedia dump is also huge ...
answered May 18, 2015 at 8:32
Westbury labs provides a ~1 billion word Wikipedia dump of all articles with greater than 1000 words from 2010: http://www.psych.ualberta.ca/~westburylab/downloads/westburylab.wikicorp.download.html
BYU has a larger dump (1.9 bn) from 2014, but it's not available for download.
I found the Exquisite Corpus and it's freely avalaible. A detail of the sources can be seen here. I really don't know the exact size, but it's on the billions' scale.
The COBUILD corpus (18M tokens) is available through WebCelex, if the arcane user interface isn't a deal-breaker. It's valuable more for its extensive manual annotations than its size, with quite a lot of morphological and phonological information available.
(It's smaller than most of the others listed here, but seems worth mentioning, since it's larger than the OANC mentioned in the question and is well-annotated.)
answered Jul 23, 2020 at 4:02
CommonCrawl
crawls the web and freely provides its archives and datasets to the public. Common Crawl's web archive consists of petabytes of data collected since 2011. It completes crawls generally every month.
It's a few billion pages (petabytes of data).
You can find versions of it that are already cleaned, de-duplicated and split by language.
answered Feb 23, 2021 at 13:00
What about 1 Billion Word Language Model Benchmark? It is freely available for download.
Also you might find this Reddit thread useful for other corpus' links.
