Can't import NLTK in Jupyter Notebook

Question

I can import the nltk package when running python or ipython from bash. So, nltk is definitely installed somewhere (in python from bash, nltk.__file__ is /home/nadine/anaconda2/lib/python2.7/site-packages/nltk/__init__.pyc )

However, when using Jupyter Notebook (which I installed using Anaconda, with the 2.7 version of python), importing nltk fails:

import nltk
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-2-b06499430ee0> in <module>()
----> 1 import nltk

ImportError: No module named nltk

In Jupyter Notebook, sys.executable yields /home/nadine/.conda/envs/py27/bin/python , while in python from bash it yields /home/nadine/anaconda2/bin/python2.7

What exactly is going wrong here and how can I fix it?

Answer 1

Anaconda uses its own version of Python, and you clearly have installed the nltk in the library for system Python.

But Anaconda normally comes bundled with the nltk -- why is yours absent? Perhaps you installed a minimal version, and the nltk needs to be installed on top of it. Check by running conda list nltk at the (anaconda-aware) bash prompt.

Whatever the reason, it sounds like the nltk is not there. Install it with conda install nltk .

Answer 2

Had the same issue. I resolved it by another jupyter version.

For me it did not work in the classic juypter notebook . I installed jupyterlab -> pip install jupyterlab and start it with: jupyter-lab .

Summary:

pip install jupyterlab
jupyter-lab

Then it worked!

Answer 3

Just a hunch, you can try and see which kernel your Jupyter Notebook is running. Perhaps it is not using the [conda root] and therefore doesn't have access to the package.

You should be able to see it in the top-right corner of the Notebook screen. Otherwise go to Kernel>Change Kernel in the menu and check there.

Answer 4

It's possible that your conda environment is set up with the wrong version of Python. I had this problem and it was resolved by:

• checking the version of python running in the notebook with sys.version
• deactivating the current environment
• creating a new environment with that python package
• activating the new environment
• running the notebook, using import nltk and nltk.download()