nltk_cli

This repo contains the TL;DR solutions to using third-party tools (e.g. Stanford NLP, SENNA NLP, etc.) with NLTK.

Disclaimers: It comes with unexpected caveats that are idiosyncratic to (i) the data you are processing and (ii) how NLTK API and Stanford NLP tools work.

Installation

cd $HOME

# Download the Stanford NLP tools
wget http://nlp.stanford.edu/software/stanford-ner-2015-04-20.zip
wget http://nlp.stanford.edu/software/stanford-postagger-full-2015-04-20.zip
wget http://nlp.stanford.edu/software/stanford-parser-full-2015-04-20.zip
# Extract the zip file.
unzip stanford-ner-2015-04-20.zip 
unzip stanford-parser-full-2015-04-20.zip 
unzip stanford-postagger-full-2015-04-20.zip
# Change to a shorter path.
mv stanford-postagger-full-2015-04-20 stanford-postagger
mv stanford-parser-full-2015-04-20 stanford-parser
mv stanford-ner-2015-04-20 stanford-ner

# Download the Stanford Spanish models.
wget http://nlp.stanford.edu/software/stanford-spanish-corenlp-2015-01-08-models.jar
unzip stanford-spanish-corenlp-2015-01-08-models.jar -d stanford-spanish
cp stanford-spanish/edu/stanford/nlp/models/ner/* stanford-ner/classifiers/

# Download the Stanford German models.
wget http://nlp.stanford.edu/software/stanford-german-2015-01-30-models.jar
unzip stanford-german-2015-01-30-models.jar -d stanford-german
cp stanford-german/edu/stanford/nlp/models/ner/* stanford-ner/classifiers/

# Download the Stanford Chinese models.
wget http://nlp.stanford.edu/software/stanford-chinese-corenlp-2015-04-20-models.jar
unzip stanford-chinese-corenlp-2015-04-20-models.jar -d stanford-chinese-ner
cp stanford-chinese-ner/edu/stanford/nlp/models/ner/* stanford-ner/classifiers/

# Download the SENNA tools.
wget http://ronan.collobert.com/senna/senna-v3.0.tgz
tar zxvf senna-v3.0.tgz
mv senna-v3.0 senna

# Install NLTK
pip install -U nltk
# Git clone this repository.
git clone https://github.com/alvations/nltk_cli.git

Usage

cd nltk_cli

###############################################################################
# Stanford NLP Tools
###############################################################################

# Using Stanford LexParser
python3 stanford.py --tool=lexparser \
--jar=$HOME/stanford-parser/stanford-parser.jar \
--modeljar=$HOME/stanford-parser/stanford-parser-3.5.2-models.jar \
--model=edu/stanford/nlp/models/lexparser/englishPCFG.ser.gz \
--input=test.txt

# Using Stanford POSTagger
python3 stanford.py --tool=postagger \
--jar=$HOME/stanford-postagger/stanford-postagger.jar \
--model=$HOME/stanford-postagger/models/english-bidirectional-distsim.tagger \
--input=test.txt

# Using Stanford NERTagger
python3 stanford.py --tool=nertagger \
--jar=$HOME/stanford-ner/stanford-ner.jar \
--model=$HOME/stanford-ner/classifiers/english.all.3class.distsim.crf.ser.gz \
--input=test.txt

# TL;DR way to use Stanford LexParser, make sure your `stanford-parser` directory
# is in your $HOME directory and you have installed as per the installation
# instruction above, otherwise these TL;DR commands won't work.
python3 stanford.py --postag test.txt
python3 stanford.py --nertag test.txt
python3 stanford.py --lexparser test.txt

# TL;DR with model option.
python3 stanford.py --lexparse test.txt \
--model=edu/stanford/nlp/models/lexparser/wsjPCFG.ser.gz 
python3 stanford.py --postag test.txt \
--model=$HOME/stanford-postagger/models/english-bidirectional-distsim.tagger 
python3 stanford.py --nertag test.txt \
--model=$HOME/stanford-ner/classifiers/english.all.3class.distsim.crf.ser.gz

###############################################################################
# SENNA NLP Tools
###############################################################################

# Using SENNA POSTagger.
python senna.py --sennadir $HOME/senna --postag --input test.txt

# Using SENNA NERTagger.
python senna.py --sennadir $HOME/senna --nertag --input test.txt

# Using SENNA ChunkTagger.
python senna.py --sennadir $HOME/senna --chunktag --input test.txt

# TL;DR NP Chunk Extractor using SENNA ChunkTagger.
python senna.py --sennadir $HOME/senna --chunk NP test.txt
python senna.py --sennadir $HOME/senna --np test.txt

# Tl;DR way to use SENNA Taggers, make sure your `senna` directory is in your
# $HOME directory and you have installed as per the installation instruction 
# above, otherwise, these Tl;DR might not work.
python senna.py --postag --input test.txt
python senna.py --nertag --input test.txt
python senna.py --chunktag --input test.txt

# Tl;DR way to use SENNA ChunkTagger to extract NPs, make sure your `senna` 
# directory is in your $HOME directory and you have installed as per the 
# installation instruction above, otherwise, these Tl;DR might not work.
python senna.py --np test.txt
python senna.py --vp test.txt
python senna.py --chunk NP test.txt
python senna.py --chunk VP test.txt

##############################################################################
# Terminator Term Extract filters.
##############################################################################

# Extracts NPs using SENNA ChunkTagger.
python3 senna.py --np test.txt --output test.np
# Filter NPs using Terminator semi-rule based cleaner. Input can be either 
# one NP per line or multiple NPs per line separated by pipe, i.e. "|"
python3 clean_np.py test.np
# If you would like to output the filtered NPs:
python3 clean_np.py test.np --output test.filtered.np
# To remove empty lines from the filtered NPs list:
python3 clean_np.py test.np | sed '/^$/d'
# To get unique list of NPs
python3 clean_np.py test.np | sed '/^$/d' | sort | uniq

Note: The test.txt file is the fish-head-curry file from the NTU-Multilingual Corpus

FAQ

UnicodeEncodeError: 'ascii' codec can't encode character...

If any of your output from the commands above cannot be output using >, e.g.

$ python senna.py --np test.txt > test.np
Traceback (most recent call last):
  File "senna.py", line 114, in <module>
    print(processed_sent)
UnicodeEncodeError: 'ascii' codec can't encode character u'\u2019' in position 9: ordinal not in range(128)

Then try the --output parameter, e.g.:

$ python senna.py --np test.txt --output test.np

Alternatively, you can set your STDOUT encoding, e.g.:

$ export PYTHONIOENCODING=utf-8
$ python senna.py --np test.txt > test.np

---

IndexError: Misalignment error occurred at sentence number ...

If you get an error as such:

$ python senna.py --np test.txt --output test.np
Traceback (most recent call last):
  File "senna.py", line 110, in <module>
    for processed_sent in process(sentences, tool, arguments['--chunk']):
  File "senna.py", line 83, in senna_extract_chunks
    tagged_sents = chunker.tag_sents(sentences)
  File "/usr/local/lib/python2.7/dist-packages/nltk/tag/senna.py", line 70, in tag_sents
    tagged_sents = super(SennaChunkTagger, self).tag_sents(sentences)
  File "/usr/local/lib/python2.7/dist-packages/nltk/classify/senna.py", line 167, in tag_sents
    % sentence_index)
IndexError: Misalignment error occurred at sentence number 11. Possible reason is that the sentence size exceeded the maximum size. Check the documentation of Senna class for more information.

First check that you have no empty lines in your input file, then remove the empty line, you can also do:

$ sed '/^$/d' test.txt > test.noempty.txt
$ python senna.py --np test.noempty.txt --output test.np