Case tagging and python

[Fred Mangusta]

Hi,

I'm relatively new to programming in general, and totally new to python,
and I've been told that this language is particularly good for what I
need to do. Let me explain.
I have a large corpus of English text, in the form of several files.

First of all I would like to scan each file. Then, for each word I find,
I'd like to examine its case status, and write the (lower case) word back
to another text file - with, appended, a tag stating the case it had in
the original file.

An example. Suppose we have three possible "case conditions"
-all lowercase
-all uppercase
-initial uppercase only

Three corresponding tags for each of these might be, respectively:
-nocap
-allcaps
-cap

Therefore, given the string

"The Chairman of BP was asleep"

I would like to produce

"the/cap chairman/cap of/nocap /bp/allcaps was/nocap /asleep/nocap"

and writing this into a file.
I have the following algorithm in mind:

-open input file
-open output file
-get line of text
-split line into words
-for each word
-tag = checkCase(word)
-newword = lowercase(word) + append(tag)
rejoin words into line
write line into output file

Now, I managed to write the following initial code

for s in file:
lines += 1
if lines % 1000 == 0:
print '%d lines' % We print the total lines
sent = s.split() #split string by spaces
#...
But then I don't quite know what would be the fastest/best way to do
this. Could I use the join function to reform the string? And, regarding
the casetest() function, what do you suggest to do? Should I test each
character of each word or there are faster methods?

Thanks very much,

F.

[bearophileHUGS]

Fred Mangusta:

Could I use the join function to reform the string?

You can write a function to split the words, for example taking in
account the points too, etc.

And, regarding the casetest() function, what do you suggest to do?

Python strings have isupper, islower, istitle methods, they may be
enough for your purposes.

-open input file
-open output file
-get line of text
-split line into words
-for each word
-tag = checkCase(word)
-newword = lowercase(word) + append(tag)
rejoin words into line
write line into output file

It seems good. To join the words of a line there's str.join. Now you
can write a function that splits lines, and another to check the case,
then you can show them to us.

Yet, I don't see how much use can have your output file :-)

Bye,
bearophile

[Fred Mangusta]

Hi, I came up with the following procedure

ALLCAPS = "|ALLCAPS"
NOCAPS = "|NOCAPS"
MIDCAPS = "|MIDCAPS"
CAPS = "|CAPS"
DIGIT = "|DIGIT"

def test_case(w):

w_out = ''

if w.isalpha(): #se la virgola non ci entra
if w.isupper():
w_out = w.lower() + ALLCAPS
return w_out
elif w.islower():
w_out = w + NOCAPS
return w_out
else:
m = re.match("^[A-Z]",w)
if m:
w_out = w.lower() + CAPS #notsure about this..
return w_out
else:
w_out = w.lower() + MIDCAPS
return w_out
elif w.isdigit():
w_out = w + DIGIT
return w_out

Called in here:
#=========================
lines = 0
for s in file:
lines += 1
if lines % 1000 == 0:
print '%d lines' % lines
#sent = sent.replace(",","")
sent = s.split() #split string by spaces
for w in sent:
wout= test_case(w)
#==========================

But I don't know if I'm doing something sensible? Moreover:

- test_case has problems, cause whenever It finds some punctuation
character attached to some word, doesn't tag it. I was thinking of
cleaning the line of the punctuation before using split on it (see
commented row) but I don't know if I have to call that replace() once
for every punctuation char?
-Is there a way to reprint the tagged text in a file including punctuation?
-Is my test_case a good start? Would you use regular expressions?

Thanks very much!
F.

[chrispoliquin]

I second the idea of just using the islower(), isupper(), and
istitle() methods.
So, you could have a function - let's call it checkCase() - that
returns a string with the tag you want...

def checkCase(word):

if word.islower():
tag = 'nocap'
elif word.isupper():
tag = 'allcaps'
elif word.istitle():
tag = 'cap'

return tag

Then let's take an input file and pass every word through the
function...

f = open(path:to:file, 'r')
corpus_text = f.read()
f.close()

tagged_corpus = ''
all_words = corpus_text.split()

for w in all_words:
tagtext = checkCase(w)
tagged_corpus = tagged_corpus + ' ' + w + '/' + tagtext

output_file = open(path:to:file, 'w')
output_file.write(tagged_corpus)
print 'All Done!'

Also, if you're doing natural language processing in Python, you
should get NLTK.