preprocess.pl - Split Senseval-2 data file into one file per lexical item (lexelt), and carry out various tokenization and formatting tasks
preprocess.pl [OPTIONS] SOURCE
Takes an xml file in SENSEVAL-2 lexical-sample format and splits it apart into as many files as there are lexical elements in the original file. Each lexical element usually corresponds with a word used in a particular part of speech. It also does other sundry preprocessing tasks with the data such as splitting it into training and test portions, tokenizing, and providing various formatting options. It can also create plain text versions of the xml files, which can be useful when needed as training data.
Senseval2 formatted input file.
Reads tokens from FILE. The context of each instance is broken up into tokens and each pair of consecutive tokens are separated by white space. Non-white space characters which do not belong to any token are put between angular brackets. If this option is not used, the default token definitions of count.pl are assumed.
Removes strings that do not match token file. If not\n"; specified, the non-matching strings are put within angular\n"; brackets, ie, <>
Removes all characters sequences that match Perl regular expressions specified in FILE.
Does not output an xml file.
Outputs the changed xml file to FILE. If this option nor the option --noxml is provided, the file name is derived by concatenating the word in the <lexelt> tag with \".xml\". Note: if this option is used, separate lexelt items will not be split into separate files.
Does not output a NSP-ready file.
Outputs just the part between <context> </context> (after modification) to FILE. FILE can then be used directly with NSP. If this option nor the option --nocount is provided, the file name is derived as in xml above, with a .count extension. Note: if this option is used, separate lexelt items will not be split into separate files.
Includes a tag <lexelt=WORD/> within the <head></head> tags, where WORD is the word in the immediately preceeding <lexelt> tag.
Includes a tag <senseid=XXXXX/> within the <head></head> tags, where XXXXX is the number in the immediately preceeding <answer> tag.
Shuffles the instances in SOURCE and then splits them into two files, a training file and a test file, approximately in the ratio N:(100-N).
Sets the seed for the random number generator used during shuffling. If not used, no seeding is done (except for that provided automatically by perl)
Puts separate lines within the <context> </context> region within <s> </s> pairs of tags. If separate sentences are on seperate lines, these tags effectively denote the start and end of sentences.
Prints the version number.
Prints this help message.
Turns on verbose mode. Silent by default.
1. The modified/processed Input SENSEVAL-2 (*.xml) file, if --noxml option is not specified.
2. The Ngram Statistics Package (NSP) ready (*.count) file, if --nocount is not specified.
3. The *-test and *-training files if the --split option is used.
The following is an example SENSEVAL-2 file that we will refer to later in as example.xml
<corpus lang='english'> <lexelt item="art.n"> <instance id="art.40001"> <answer instance="art.40001" senseid="art~1:06:00::"/> <answer instance="art.40001" senseid="fine_art%1:06:00::"/> <context> <head>Art</head> you can dance to from the creative group called Halo. </context> </instance> <instance id="art.40002"> <answer instance="art.40002" senseid="art_gallery~1:06:00::"/> <context> There's always one to be heard somewhere during the summer in the piazza in front of the <head>art</head>gallery and town hall or in a park. </context> </instance> <instance id="art.40005" docsrc="bnc_ckv_938"> <answer instance="art.40005" senseid="art~1:04:00::"/> <context> Paintings, drawings and sculpture from every period of <head>art</head> during the last 350 years will be on display. </context> </instance> </lexelt> <lexelt item="authority.n"> <instance id="authority.40001"> <answer instance="authority.40001" senseid="authority~1:14:00::"/> <context> Not only is it allowing certain health <head>authorities</head>to waste millions of pounds on computer systems that dont work, it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice. </context> </instance> </lexelt> </corpus>
Here we have two lexelts, "art.n" and "authority.n", where "n" denotes that these are noun senses of the words. We have three instances of art with instance id's art.40001, art.40002 and art.40007 respectively, and one instance of authority with instance id authority.40001. The first instance has two answers, while the others have one each.
preprocess.pl accepts regular expressions from the user and then "tokenizes" the text between the <context> </context> tags. This is done to simplify the construction of regular expressions in program nsp2regex.pl and to achieve optimum regular expression matching in xml2arff.pl. Following is a description of the tokenization process.
The text within the <context> </context> tags is considered as one string, the "input" string. This algorithm takes this input string and creates an "output" string where tokens are identified and separated from each other by a SINGLE space. Regex's provided by the user are checked against the input string to see if a sequence of characters starting with the first character of the string match against any of these regex's. As soon as a we find a regular expression that does match, this checking is halted, the matched sequence of characters is removed from the string and appended to an "output" string with exactly one space to its left and right. If none of the regex's match against the starting characters of the input string, the first character is considered a "non-token". By default this non token is placed in angular brackets (<>) and then put into the output string with one space to its left and right. This process is continued until the input string becomes empty. This process is restarted for the next instance.
For example, assume we provide the following regular expressions to preprocess.pl:
<head>\w+</head> \w+
The first regular expression says that a sequence of characters starting with "<head>", having an unbroken sequence of alphanumeric characters and finally ending with a "</head>" is a valid token. Also, an unbroken sequence of alphanum characters makes a token.
Then, assuming that the following text occurs within the <context> </context> tags of an instance:
No, he has no <head>authority</head> on this!
preprocess.pl would then convert this text to:
No <,> he has no <head>authority</head> on this <!>
Observe that "No", "he", "has", "no", "<head>authority</head>", etc are all the tokens, while "," and "!" arent tokens and so have been put into angular brackets. Further observe that each token has exactly one space to its left and right.
One can provide a file containing regular expressions to preprocess.pl using the switch --token. In this file, each regular expression should be on a line of its own and should be preceeded and succeeded with '/' signs. Further these should be perl regular expressions.
Thus our regular expressions above would look like so:
/<head>\w+<\/head>/ /\w+/
We shall call the file these regular expressions lie in "token.txt". Then, we would run preprocess.pl on example.xml with this token file like so:
preprocess.pl example.xml --token token.txt
Although tokenization is best controlled via a user specified tokenization file designated via the --token option, there is also a default definition of tokens that is used in the absence of a tokenization file, which consists of the following:
/w+/ /[\.,;:\?!]/
According to this definition, a token is either a single punctuation mark from the specified class, or it is a string of alpha-numeric characters. Note that this default definition is generally not a good choice for XML data since it does not treat XML tags as tokens and will result in them "breaking apart" during pre-processing. For example, given this default definition, the string :
<head>art</head>
will be represented by preprocess.pl as
<<> head <>> art <<> </> head <>>
which suggests that "<", ">", and "/" are non-tokens, while "art" and "head" are. This is unlikely to provide useful information.
These defaults correspond to those in NSP, which is geared towards plain text. These are provided as a convenience, but in general we recommend against relying upon them when processing XML data.
Assume that the only regular expression in our token file token.txt is /\S+/. This regular expression says that any sequence of non-white-space characters is a token. Now, if we run the program like so:
(where example.xml is the example xml file described in the previous section and token.txt is the file that contains the above regular expressions /\S+/).
We would get all the four files, art.n.xml, art.n.count, authority.n.xml and authority.n.count. From here on we shall show only the "authority" files to save space; it is understood that the art files are also created.
File authority.n.xml:
<corpus lang='english'> <lexelt item="authority.n"> <instance id="authority.40001"> <answer instance="authority.40001" senseid="authority~1:14:00::"/> <context> Not only is it allowing certain health <head>authorities</head>to waste millions of pounds on computer systems that dont work, it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice. </context> </instance> </lexelt> </corpus>
File authority.n.count:
Not only is it allowing certain health <head>authorities</head> to waste millions of pounds on computer systems that dont work, it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice.
Note that every character is a part of some sequence of non-white-space characters, and is therefore part of some token. Hence no character is put into <> brackets. Also, each non-white-space-character-sequence, that is each token, is placed in the output with exactly one space character to its left and right.
On the other hand if our token file token.txt were to contain the following regex which treats every sequence of alpha numeric characters as a token:
/\w+/
... and we were to run the program like so:
... then our authority files would like like so:
<corpus lang='english'> <lexelt item="authority.n"> <instance id="authority.40001"> <answer instance="authority.40001" senseid="authority~1:14:00::"/> <context> Not only is it allowing certain health <<> head <>> authorities <</> head <>> to waste millions of pounds on computer systems that dont work <,> it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice <.> </context> </instance> </lexelt> </corpus>
Not only is it allowing certain health <<> head <>> authorities <</> head <>> to waste millions of pounds on computer systems that dont work <,> it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice <.>
Note again that since the '<' and '>' of the head tags are not alpha-numeric characters they are considered as "non-token" characters, and are put within the <> tags. Further note that if there are more than one such non-token characters one after another, they get put into one pair of diamond brackets '<' and '>'. As mentioned before, the user should include regular expressions that preserve the tags. Thus for the above example, a regular expression like /<head>\w+<\/head>/ would work admirably.
Besides the regular expressions <head>\w+</head> and \w+, we have found the following regular expressions useful too.
/[\.,;:\?!]/ - This states that a single occurrence of one of the puncutation marks in the list is a token. This helps us specify that a puncutation mark is indeed a token and should not be ignored! Further, this allows us to create features consisting of punctuation marks using SenseClusters. /&([^;]+;)+/ - The XML format forces us to replace certain meta symbols in the text by their standard formats. For example, if the '<' symbol occurs in the text, it is replaced with "<". Similarly, '-' is replaced with "‐". This regular expression recognizes these constructs as tokens instead of breaking them up!
Recall that at every point of the "input string", the matching mechanism marches down the regular expressions in the order they are provided in the input regular expression file, and stops at the FIRST regular expression that matches. Thus the order of the regular expression makes a difference. For example, say our regular expression file has the following regular expressions in this order:
/he/ /hear/ /\w+/
and our input text is "hear me"
Then our output text is " he ar me "
On the other hand, if we reverse the first two regular expressions
/hear/ /he/ /\w+/
we get as output " hear me "
Thus as expected, the order of the regular expressions define how the output will look.
Consider the following regular expressions:
/\S+/ /\w+/
As should be obvious, every token that matches the second regular expression matches the first one too. We say that the first regular expression "subsumes" the second one, and the second regular expression is redundant. This is because the matching mechanism will always stop at the first regular expression, and never get an opportunity to exercise the second one. Note of course that this does not adversely affect anything.
Recall that characters in the input string that do not match any regular expression as defined in token are put into angular (<>) brackets. You may, if you wish, remove these "non tokens", that is not have them appear in the output xml and count files, by using the switch --removeNotToken.
Thus, for the following text:
No, he has no <head>authority</head> on me!
and with regular expressions
and if we were to run the program with the switch --removeNotToken, preprocess.pl would convert the text into:
No he has no <head>authority</head> on me
The --nontoken option allows a user to specify a list of regular expressions. Any strings in the input file that match this list are removed from the file prior to tokenization.
It's important to note the order in which tokenization occurs. First, those strings that match the regexes defined in nontoken are removed. Then the strings that match the regexes defined in token are matched. Those tokens that do not match the token regexes are then removed. Thus, the "order" of precedence during tokenization is:
-nontoken -token -removeNotToken
By default, for each lexical element "word" in the training or test file (in the lexical sample of SENSEVAL-2), preprocess.pl creates a file of the name "word".xml. For example for the file example.xml, preprocess.pl will create files art.n.xml and authority.n.xml if it is run as follows:
File art.n.xml:
<corpus lang='english'> <lexelt item="art.n"> <instance id="art.40001"> <answer instance="art.40001" senseid="art~1:06:00::"/> <answer instance="art.40001" senseid="fine_art%1:06:00::"/> <context> <head>Art</head> you can dance to from the creative group called Halo <.> </context> </instance> <instance id="art.40002"> <answer instance="art.40002" senseid="art_gallery~1:06:00::"/> <context> There <'> s always one to be heard somewhere during the summer in the piazza in front of the <head>art</head> gallery and town hall or in a park <.> </context> </instance> <instance id="art.40005" docsrc="bnc_ckv_938"> <answer instance="art.40005" senseid="art~1:04:00::"/> <context> Paintings <,> drawings and sculpture from every period of <head>art</head> during the last 350 years will be on display <.> </context> </instance> </lexelt> </corpus>
<corpus lang='english'> <lexelt item="authority.n"> <instance id="authority.40001"> <answer instance="authority.40001" senseid="authority~1:14:00::"/> <context> Not only is it allowing certain health <head>authorities</head> to waste millions of pounds on computer systems that dont work <,> it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice <.> </context> </instance> </lexelt> </corpus>
Observe of course that the text within the <context> </context> region has been tokenized as described previously according to the regular expressions in file token.txt.
This default behavior can be stopped either by using the switch --xml FILE, by which only one FILE is created, or by using the switch --noxml, by which no xml file is created.
Besides creating xml output, this program also creates output that can be used directly with the program count.pl (from the Ngram Statistics Package). After tokenizing the region within the <context> </context> tags of each instance, the program puts together ONLY these pieces of text to create "count.pl ready" output. This is because count.pl assumes that all tokens in the input file needs to be "counted" and generally we are only interested in the "contextual" material provided in each instance, and not the tags that occur outside the <context> </context> region of text.
By default, for each lexical element "word", this program creates a file of the name word.count. For example, for the file example.xml, we would get the files art.n.count and authority.n.count.
File art.n.count:
<head>Art</head> you can dance to from the creative group called Halo <.> There <'> s always one to be heard somewhere during the summer in the piazza in front of the <head>art</head> gallery and town hall or in a park <.> Paintings <,> drawings and sculpture from every period of <head>art</head> during the last 350 years will be on display <.>
Not only is it allowing certain health <head>authorities</head> to waste millions of pounds on computer systems that dont work <,> it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice <.>
This default behavior can be stopped either by using the switch --count FILE, by which only one FILE is created, or by using the switch --nocount, by which no count file is created.
Note that the --xml/--noxml switches and the --count/--nocount switches are independant of each other. Thus, although providing --xml FILE or --noxml switchs produces a single xml FILE or no xml file at all, you will still get all the count files, unless you also give the --count FILE or --nocount switches. Similarly, providing the --count FILE or --nocount switches does not affect the production of the xml files.
The lexelt information and the senseId information are outside the <context> </context> region. This program gives you the capability to bring these pieces of information inside the context.
Switch --useLexelt puts the tag <lexelt=WORD/> within the <head></head> tags, where WORD is the word in the immediately preceding <lexelt> tag.
Switch --useSenseid puts the tag <senseid=XXXXX/> within the <head></head> tags, where XXXXX is the number in the immediately preceding <answer> tag.
For example, running the program like so:
preprocess.pl example.xml --useLexelt --useSenseid --token token.txt
produces this for authority.n.xml:
<corpus lang='english'> <lexelt item="authority.n"> <instance id="authority.40001"> <answer instance="authority.40001" senseid="authority~1:14:00::"/> <context> Not only is it allowing certain health <head> authorities <lexelt=authority.n/><senseid=authority~1:14:00::/></head> to waste millions of pounds on computer systems that dont work , it also allowed the London ambulance service to put lives at risk with a system that had not been fully proven in practice . </context> </instance> </lexelt> </corpus>
Note that the extra information is put inside the <head> </head> region. Hence the user has to provide a token file that will preserve these <head> </head> tags. For instance, as shown in the previous section, if one were to rely on the default regex's, these tags would not be preserved (the '<' and '>' would be considered non-token symbols) and the lexelt and senseid information would not be included within the tags.
So for example, the following regular expression file is adequate:
The english lexical sample data available from SENSEVAL-2 is such that each sentence within the <context> </context> tags is on a line of its own. This human-detected sentence boundary information is usually lost in preprocess.pl, but can be preserved using the switch --putSentenceTags. This puts each line within <s> and </s> tags. Assuming that each sentence was originally on a line of its own, then <s> marks the start of a sentence and </s> marks its end. Note that no sentence boundary detection is done: if the end of line character (\n) does not match the end of a sentence, then the <s> </s> tags will not be indicative of a sentence boundary either.
For example, assume the following is our source xml file, source.xml:
<corpus lang='english'> <lexelt item="word"> <instance id="word.1"> <answer instance="word.1" senseid="1"/> <context> This is the first line This is the second line This is the last line for <head>word</head> </context> </instance> </lexelt> </corpus>
Further assume our token file is this:
/<head>\w+</head>/ /<s>/ /<\/s>/ /\w+/
Running preprocess.pl like so:
preprocess.pl --token token.txt source.xml
Produces the following word.xml file:
and the following word.count file:
This is the first line This is the second line This is the last line for <head>word</head>
However, running preprocess.pl like so:
preprocess.pl --token token.txt --putSentenceTags source.xml
<corpus lang='english'> <lexelt item="word"> <instance id="word.1"> <answer instance="word.1" senseid="1"/> <context> <s> This is the first line </s> <s> This is the second line </s> <s> This is the last line for <head>word</head> </s> </context> </instance> </lexelt> </corpus>
<s> This is the first line </s> <s> This is the second line </s> <s> This is the last line for <head>word</head> </s>
Note that the <s> and </s> tags are placed into the data BEFORE the tokenization process. Hence a token regular expression that preserves these tags is required! The token file shown above is adequate for this.
Besides splitting the lexical elements into separate files, preprocess.pl also allows you to split the instances of a single lexical element into separate "training" and "test" files.
If one has a corpus of sense-tagged text, it is often desirable to divide that sense tagged text into training and test portions in order to develop or tune a methodology. This is the intention of the --split option.
The --split option of preprocess.pl allows you to specify an integer N... the instances of each lexical element in the input XML SOURCE file are split into two files approximately in the ratio N:(100-N).
If an output XML file "foo" is specified through the switch --xml then two files, foo-training.xml and foo-test.xml are created.
If an output count file "foo" is specified through the switch --count then two files, foo-training.count and foo-test.count are created.
Creation of Xml and count output files can be suppressed by using the --noxml and --nocount switches respectively.
If neither --noxml nor --xml switches are used, then files of the type word-training.xml, word-test.xml are created.
If neither --nocount nor --count switches are used, then files of the type word-training.count, word-test.count are created.
The instances are shuffled before being put into training and test files. Perl automatically seeds the randomizing process... but you can specify your own seed using the switch --seed.
Satanjeev Banerjee, Carnegie-Mellon University Ted Pedersen, University of Minnesota, Duluth tpederse at d.umn.edu
Copyright (c) 2001-2008, Satanjeev Banerjee and Ted Pedersen
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.
To install Text::SenseClusters, copy and paste the appropriate command in to your terminal.
cpanm
cpanm Text::SenseClusters
CPAN shell
perl -MCPAN -e shell install Text::SenseClusters
For more information on module installation, please visit the detailed CPAN module installation guide.