nltk doesn't add $NLTK_DATA to search path?

Question

under linux,I have set env var $NLTK_DATA('/home/user/data/nltk'),and blew test works as expected

>>> from nltk.corpus import brown
>>> brown.words()
['The', 'Fulton', 'County', 'Grand', 'Jury', 'said', ...]

but when running another python script,I got:

LookupError: 
**********************************************************************
Resource u'tokenizers/punkt/english.pickle' not found.  Please
use the NLTK Downloader to obtain the resource:  >>>
nltk.download()
Searched in:
- '/home/user/nltk_data'
- '/usr/share/nltk_data'
- '/usr/local/share/nltk_data'
- '/usr/lib/nltk_data'
- '/usr/local/lib/nltk_data'
- u''

---

As we can see,nltk doesn't add $NLTK_DATA to search path,after appending NLTK_DATA dir manually:

nltk.data.path.append("/NLTK_DATA_DIR");

script runs as expected,question is:

How to make nltk to add $NLTK_DATA to it's search path automatically?

Answer 1

If you don't want to set the $NLTK_DATA before running your scripts, you can do it within the python scripts with:

import nltk
nltk.path.append('/home/alvas/some_path/nltk_data/')

E.g. let's move the the nltk_data to a non-standard path that NLTK won't find it automatically:

alvas@ubi:~$ ls nltk_data/
chunkers  corpora  grammars  help  misc  models  stemmers  taggers  tokenizers
alvas@ubi:~$ mkdir some_path
alvas@ubi:~$ mv nltk_data/ some_path/
alvas@ubi:~$ ls nltk_data/
ls: cannot access nltk_data/: No such file or directory
alvas@ubi:~$ ls some_path/nltk_data/
chunkers  corpora  grammars  help  misc  models  stemmers  taggers  tokenizers

Now, we use the nltk.path.append() hack:

alvas@ubi:~$ python
>>> import os
>>> import nltk
>>> nltk.path.append('/home/alvas/some_path/nltk_data/')
>>> nltk.pos_tag('this is a foo bar'.split())
[('this', 'DT'), ('is', 'VBZ'), ('a', 'DT'), ('foo', 'JJ'), ('bar', 'NN')]
>>> nltk.data
<module 'nltk.data' from '/usr/local/lib/python2.7/dist-packages/nltk/data.pyc'>
>>> nltk.data.path
['/home/alvas/some_path/nltk_data/', '/home/alvas/nltk_data', '/usr/share/nltk_data', '/usr/local/share/nltk_data', '/usr/lib/nltk_data', '/usr/local/lib/nltk_data']
>>> exit()

Let's move it back and see whether it works:

alvas@ubi:~$ ls nltk_data
ls: cannot access nltk_data: No such file or directory
alvas@ubi:~$ mv some_path/nltk_data/ .
alvas@ubi:~$ python
>>> import nltk
>>> nltk.data.path
['/home/alvas/nltk_data', '/usr/share/nltk_data', '/usr/local/share/nltk_data', '/usr/lib/nltk_data', '/usr/local/lib/nltk_data']
>>> nltk.pos_tag('this is a foo bar'.split())
[('this', 'DT'), ('is', 'VBZ'), ('a', 'DT'), ('foo', 'JJ'), ('bar', 'NN')]

---

If you really really want to find nltk_data automagically, use something like:

import scandir
import os, sys
import time

import nltk

def find(name, path):
    for root, dirs, files in scandir.walk(path):
        if root.endswith(name):
            return root

def find_nltk_data():
    start = time.time()
    path_to_nltk_data = find('nltk_data', '/')
    print >> sys.stderr, 'Finding nltk_data took', time.time() - start
    print >> sys.stderr,  'nltk_data at', path_to_nltk_data
    with open('where_is_nltk_data.txt', 'w') as fout:
        fout.write(path_to_nltk_data)
    return path_to_nltk_data

def magically_find_nltk_data():
    if os.path.exists('where_is_nltk_data.txt'):
        with open('where_is_nltk_data.txt') as fin:
            path_to_nltk_data = fin.read().strip()
        if os.path.exists(path_to_nltk_data):
            nltk.data.path.append(path_to_nltk_data)
        else:
            nltk.data.path.append(find_nltk_data())
    else:
        path_to_nltk_data  = find_nltk_data()
        nltk.data.path.append(path_to_nltk_data)


magically_find_nltk_data()
print nltk.pos_tag('this is a foo bar'.split())

Let's call that python script, test.py:

alvas@ubi:~$ ls nltk_data/
chunkers  corpora  grammars  help  misc  models  stemmers  taggers  tokenizers
alvas@ubi:~$ python test.py
Finding nltk_data took 4.27330780029
nltk_data at /home/alvas/nltk_data
[('this', 'DT'), ('is', 'VBZ'), ('a', 'DT'), ('foo', 'JJ'), ('bar', 'NN')]
alvas@ubi:~$ mv nltk_data/ some_path/
alvas@ubi:~$ python test.py
Finding nltk_data took 4.75850391388
nltk_data at /home/alvas/some_path/nltk_data
[('this', 'DT'), ('is', 'VBZ'), ('a', 'DT'), ('foo', 'JJ'), ('bar', 'NN')]

Answer 2

If you are someone who wants to install the NLTK data in a conda environment, and doesn't want to specify the data location in every script, or export the environment variable, you need to do the following:

1. Activate your desired conda environment.
2. Print sys.prefix within your conda environment, and copy this path (let's say /home/dickens/envs/nltk_env.
3. Run nltk.download() within the conda environment, select your desired packages, and append /share/nltk_data to your path from above as the download location. For e.g. in our case, it will become /home/dickens/envs/nltk_env/share/nltk_data.
4. You are now good to go!