I get an ImportError on attempting to import nltk: "ImportError: No module named sparsefuncs"

I am using the model branch of NLTK (installed today with "pip install https://github.com/nltk/nltk/tarball/model".)

Traceback (most recent call last):
File "../MyFile.py", line 3, in
import nltk
File "/usr/local/lib/python2.7/dist-packages/nltk/init.py", line 117, in
from nltk.align import *
File "/usr/local/lib/python2.7/dist-packages/nltk/align/init.py", line 15, in
from nltk.align.ibm1 import IBMModel1
File "/usr/local/lib/python2.7/dist-packages/nltk/align/ibm1.py", line 18, in
from nltk.corpus import comtrans
File "/usr/local/lib/python2.7/dist-packages/nltk/corpus/init.py", line 66, in
from nltk.corpus.reader import *
File "/usr/local/lib/python2.7/dist-packages/nltk/corpus/reader/init.py", line 62, in
from nltk.corpus.reader.chunked import *
File "/usr/local/lib/python2.7/dist-packages/nltk/corpus/reader/chunked.py", line 21, in
from nltk.chunk import tagstr2tree
File "/usr/local/lib/python2.7/dist-packages/nltk/chunk/init.py", line 157, in
from nltk.chunk.api import ChunkParserI
File "/usr/local/lib/python2.7/dist-packages/nltk/chunk/api.py", line 13, in
from nltk.parse import ParserI
File "/usr/local/lib/python2.7/dist-packages/nltk/parse/init.py", line 79, in
from nltk.parse.transitionparser import TransitionParser
File "/usr/local/lib/python2.7/dist-packages/nltk/parse/transitionparser.py", line 17, in
from sklearn.datasets import load_svmlight_file
File "/usr/lib/python2.7/dist-packages/sklearn/datasets/init.py", line 23, in
from .twenty_newsgroups import fetch_20newsgroups
File "/usr/lib/python2.7/dist-packages/sklearn/datasets/twenty_newsgroups.py", line 53, in
from ..feature_extraction.text import CountVectorizer
File "/usr/lib/python2.7/dist-packages/sklearn/feature_extraction/init.py", line 10, in
from . import text
File "/usr/lib/python2.7/dist-packages/sklearn/feature_extraction/text.py", line 29, in
from ..preprocessing import normalize
File "/usr/lib/python2.7/dist-packages/sklearn/preprocessing/init.py", line 6, in
from .data import Binarizer
File "/usr/lib/python2.7/dist-packages/sklearn/preprocessing/data.py", line 20, in
from ..utils.sparsefuncs import inplace_csr_row_normalize_l1
ImportError: No module named sparsefuncs

In case they're relevant, I did get two warnings on my install of NLTK:
warning: no files found matching 'Makefile' under directory '.txt'
warning: no previously-included files matching '~' found anywhere in distribution

It says:

...see the API documentation for <cite>nltk.corpus.CorpusReader</cite>.</p>

But I think it should read:

...see the API documentation for <cite>nltk.corpus.reader</cite>.</p>

Hi,

on this very github project, there is no convenience link to the nltk web page, like the one at the top of the nltk code github project:

It's not related to this repository but to nltk organization on GitHub. It's a security issue in my opinion.

The current setting allows members of the organization give access to its repositories to third-party applications even if the members are giving access to these applications for use in its own repositories (not in nltk's).
The default setting should be to avoid it and, if needed, grant access to certain third-party applications. To configure it an owner should:

• Go to https://github.com/organizations/nltk/settings/oauth_application_policy (Organization page > Settings > Third-party access)
• Click on "Setup application access restrictions"
• Click on "Restrict third-party application acess"

Currently the best practice of installing Python packages is to create separate virtual environments for specific projects. However, http://www.nltk.org/install.html suggests running sudo pip install -U nltk, which IMO is not a piece of good advice. Instructions shouldn't contain sudo in them.

Maybe this is done with the consideration that telling the users to create virtual environments can cause further complications when they're using it, since many are still not familiar with the concept. However, would it be at least possible to suggest this as an alternative? Also I'm not sure if there would be some bugs and confusion for the users if they attempt to use sudo with pip, since there seem to be some problems with this on the latest MacOS.

An example is spaCy's getting started guide: https://spacy.io/usage/

The NLTK FAQ's question #5 ("What data sources does NLTK use?") has in its answer the following stale (404) links:

• http://nltk.org/nltk_data/ ("available datasets")
• http://nltk.org/howto/ ("Corpus Howto")

http://www.nltk.org/ still purports to be "NLTK 3.0 documentation" even though the API docs now reflect docstrings from the latest (3.2.2) release. The title should either be updated with each release, or it should be modified to contain a less precise version number that is not incorrect (e.g. "NLTK 3 documentation")

https://github.com/nltk/nltk.github.com/blob/master/howto/ccg.html#L348
I could not find the correct link or I would have pushed a PR.

NLTK Python 3 support should be updated to 3.4+ as in the github repo: https://github.com/nltk/nltk/blob/develop/web/install.rst

Currently on the page: http://www.nltk.org/install.html and https://github.com/nltk/nltk.github.com/blob/master/install.html it's showing 3.2+

Visiting https://www.nltk.org results in a browser warning due to a certificate error:

This is caused by the server presenting a certificate issued to github.com instead of nltk.org:

I understand this is because www.nltk.org is hosted on Github. However, having to click through a serious security warning is an absolutely awful experience for first-time users. Please consider either getting a correct certificate or disabling HTTPS access to the page.