HUGE-COUNT(1) User Contributed Perl Documentation HUGE-COUNT(1)NAMEhuge-count.pl - Count all the bigrams in a huge text without using huge
amounts of memory.
SYNOPSIShuge-count.pl--tokenlist --split 100 destination-dir input
DESCRIPTION
Runs count.pl efficiently on large amounts of data by splitting the
data into separate files, and counting up each file separately, and
then merging them to get overall results.
Two output files are created. destination-dir/huge-count.output
contains the bigram counts after applying --remove and --remove.
destination-dir/complete-huge-count.output provides the bigram counts
as if no --uremove or --remove cutoff were provided.
USAGEhuge-count.pl [OPTIONS] DESTINATION [SOURCE]+
INPUT
Required Arguments:
[SOURCE]+
Input to huge-count.pl should be a -
1. Single plain text file
Or
2. Single flat directory containing multiple plain text files
Or
3. List of multiple plain text files
DESTINATION
A complete path to a writable directory to which huge-count.pl can
write all intermediate and final output files. If DESTINATION does not
exist, a new directory is created, otherwise, the current directory is
simply used for writing the output files.
NOTE: If DESTINATION already exists and if the names of some of the
existing files in DESTINATION clash with the names of the output files
created by huge-count, these files will be over-written w/o prompting
user.
--tokenlist
This parameter is required. huge-count will call count.pl and print out
all the bigrams count.pl can find out.
Optional Arguments:
--split N
This parameter is required. huge-count will divide the output bigrams
tokenlist generated by count.pl, sort on each part and recombine the
bigram counts from all these intermediate result files into a single
bigram output that shows bigram counts in SOURCE.
Each part created with --split N will contain N lines. Value of N
should be chosen such that huge-sort.pl can be efficiently run on any
part containing N lines from the file contains all bigrams file.
We suggest that N is equal to the number of KB of memory you have. If
the computer has 8 GB RAM, which is 8,000,000 KB, N should be set to
8000000. If N is set too small, split output file suffixes exhausted.
--token TOKENFILE
Specify a file containing Perl regular expressions that define the
tokenization scheme for counting. This will be provided to count.pl's
--token option.
--nontoken NOTOKENFILE
Specify a file containing Perl regular expressions of non-token
sequences that are removed prior to tokenization. This will be provided
to the count.pl's --nontoken option.
--stop STOPFILE
Specify a file of Perl regex/s containing the list of stop words to be
omitted from the output BIGRAMS. Stop list can be used in two modes -
AND mode declared with '@stop.mode = AND' on the 1st line of the
STOPFILE
or
OR mode declared using '@stop.mode = OR' on the 1st line of the
STOPFILE.
In AND mode, bigrams whose both constituent words are stop words are
removed while, in OR mode, bigrams whose either or both constituent
words are stopwords are removed from the output.
--window W
Tokens appearing within W positions from each other (with at most W-2
intervening words) will form bigrams. Same as count.pl's --window
option.
--remove L
Bigrams with counts less than L in the entire SOURCE data are removed
from the sample. The counts of the removed bigrams are not counted in
any marginal totals. This has same effect as count.pl's --remove
option.
--uremove L
Bigrams with counts more than L in the entire SOURCE data are removed
from the sample. The counts of the removed bigrams are not counted in
any marginal totals. This has same effect as count.pl's --uremove
option.
--frequency F
Bigrams with counts less than F in the entire SOURCE are not displayed.
The counts of the skipped bigrams ARE counted in the marginal totals.
In other words, --frequency in huge-count.pl has same effect as the
count.pl's --frequency option.
--ufrequency F
Bigrams with counts more than F in the entire SOURCE are not displayed.
The counts of the skipped bigrams ARE counted in the marginal totals.
In other words, --frequency in huge-count.pl has same effect as the
count.pl's --ufrequency option.
--newLine
Switches ON the --newLine option in count.pl. This will prevent bigrams
from spanning across the lines.
Other Options :
--help
Displays this message.
--version
Displays the version information.
PROGRAM LOGIC
· STEP 1
# create output dir
if(!-e DESTINATION) then
mkdir DESTINATION;
· STEP 2
1. If SOURCE is a single plain file -
huge-count.pl with --tokenlist option call count.pl and run on
the single plain file and print out all bigrams into one file.
The count outputs are also created in DESTINATION.
2. SOURCE is a single flat directory containing multiple plain
files -
huge-count.pl with --tokenlist option call count.pl and run on
each file present in the SOURCE directory. All files in SOURCE
are treated as the data files. If SOURCE contains sub-
directories, these are simply skipped. Intermediate bigram
outputs are written in DESTINATION.
3. SOURCE is a list of multiple plain files -
If #arg > 2, all arguments specified after the first argument
are considered as the SOURCE file names. count.pl is separately
run on each of the SOURCE files specified by argv[1], argv[2],
... argv[n] (skipping argv[0] which should be DESTINATION).
Intermediate results are created in DESTINATION.
In summary, a large datafile can be provided to huge-count in the
form of
a. A single plain file
b. A directory containing several plain files
c. Multiple plain files directly specified as command line
arguments
In all these cases, count.pl with --tokenlist is separately run on
SOURCE files or parts of SOURCE file and intermediate results are
written in DESTINATION dir.
· STEP 3
Split the output file generate by count.pl with --tokenlist into
smaller files by the number of bigrams N.
· STEP 4
huge-sort.pl counts the unique bigrams and sort them in alphabetic
order.
· STEP 5
huge-merge.pl merge the bigrams of each sorted bigrams file.
OUTPUT
After huge-count finishes successfully, DESTINATION will contain -
· Final bigram count file (huge-count.output) showing bigram counts
in the entire SOURCE after --remove and --uremove applied.
· Final bigram count file (complete-huge-count.output) showing bigram
counts in the entire SOURCE without --remove and --uremove.
BUGShuge-count.pl doesn't consider bigrams at file boundaries. In other
words, the result of count.pl and huge-count.pl on the same data file
will differ if --newLine is not used, in that, huge-count.pl runs
count.pl on multiple files separately and thus looses the track of the
bigrams on file boundaries. With --window not specified, there will be
loss of one bigram at each file boundary while its W bigrams with
--window W.
Functionality of huge-count with --tokenlist is same as count only if
--newLine is used and all files start and end on sentence boundaries.
In other words, there should not be any sentence breaks at the start or
end of any file given to huge-count.
AUTHOR
Amruta Purandare, University of Minnesota, Duluth
Ted Pedersen, University of Minnesota, Duluth tpederse at umn.edu
Ying Liu, University of Minnesota, Twin Cities liux0395 at umn.edu
COPYRIGHT
Copyright (c) 2004-2010, Amruta Purandare, Ted Pedersen, and Ying Liu
This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to
The Free Software Foundation, Inc., 59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA.
perl v5.20.2 2011-03-31 HUGE-COUNT(1)
