A 100+ million word corpus of American English freely available

laohong

管理员
Staff member
A new 100+ million word corpus of American English (1920s-2000s) is now
freely available
at:

http://corpus.byu.edu/time/

The corpus is based on more than 275,000 articles in TIME magazine from
1923 to 2006, and it contains articles on a wide range of topics -
domestic and international, sports, financial, cultural, entertainment,
personal interest, etc.

The architecture and interface is similar to the one that we have
created for our version of the British National Corpus (see
http://corpus.byu.edu/bnc), and it allows users to:

-- Find the frequency of particular words, phrases, substrings
(prefixes, suffixes, roots) in each decade from the 1920s-2000s. Users
can also limit the results by frequency in any set of years or decades.
They can also see charts that show the totals for all matching strings
in each decade (1920s-2000s), as well as each year within a given
decade.

-- Study changes in syntax since the 1920s. The corpus has been tagged
for part of speech with CLAWS (the same tagger used for the BNC), and
users can easily carry out searches like the following (from among
endless possibilities): changes in the overall frequency of "going + to
+ V", or "end up V-ing", or preposition stranding (e.g. "[VV*] with ."),
or phrasal verbs (1920s-1940s vs 1980s-2000s).

-- Look at changes in collocates to investigate semantic shifts during
the past 80 years. Users can find collocates up to 10 words to left or
right of node word, and sort and limit by frequency in any set of years
or decades.

-- As mentioned, the interface is designed to easily permit comparisons
between different sets of decades or years. For example, with one simple
query users could find words ending in -dom that are much more frequent
1920s-40s than 1980s-1990s, nouns occurring with "hard" in 1940s-50s but
not in the 1960s, adjectives that are more common 2003-06 than 2000-02,
or phrasal verbs whose usage increases markedly after the 1950s, etc.

-- Users can easily create customized lists (semantically-related words,
specialized part of speech category, morphologically-related words,
etc), and then use these lists directly as part of the query syntax.

----------

For more information, please contact Mark Davies
(http://davies-linguistics.byu.edu), or visit:

http://corpus.byu.edu/

for information and links to related corpora, including the upcoming BYU
American National Corpus [BANC] (350+ million words, 1990-2007+).

============================================
Mark Davies
Professor of (Corpus) Linguistics
Brigham Young University
(phone) 801-422-9168 / (fax) 801-422-0906
Web: davies-linguistics.byu.edu

** Corpus design and use // Linguistic databases **
** Historical linguistics // Language variation **
** English, Spanish, and Portuguese **
============================================
 
回复: A 100+ million word corpus of American English freely available

Thanks. It is better than the previous corpus "www.view.byu.edu".
 
回复: A 100+ million word corpus of American English freely available

terrific
thanks, laohong

BTW,how to understand insert tag, for instance, difference between "noun.ALL" and "noun.ALL+"?
 
Last edited:
回复: A 100+ million word corpus of American English freely available

The corpus was tagged with CLAWS8 tagset, see here for a complete list:
http://www.comp.lancs.ac.uk/ucrel/claws8tags.pdf

When you insert "noun.ALL", the wildcard is "[nn*]", and it searches POS tags as NN, NN1, NN2, NNA, NNB, etc. However, it cannot get ND1, NP, NP1, etc.

On the countrary, when you insert "noun.ALL+", the wildcard is [n*], and it can get all the POS tags which start with a N.
 
回复: A 100+ million word corpus of American English freely available

Thank you for sharing, laohong!
 
Back
顶部