NLPyPort

The NLPy_Port is a pipeline assembled from the NLTK pipeline, adding and changing its elements for better processing the portuguese that were previouslly created for the NLPPort pipeline. It suports at the moment the taks of Tokenization, PoS Tagging , Lemmatization and Named Entity Recognition

Instalation

Installing NLPyPort should be as simple as installing the requirments or installing the module via pip (pip install NLPyPort). However, some other configurations may be necessary.

If your NLTK version is above 3.4.5, install the version 3.4.5 by running:

>>> pip install nltk==3.4.5

If you installed NLTK and do not have downloaded the "Floresta" corpus, run the following commands:

>>> import nltk
>>> nltk.download('floresta')

Usage

In order to simplify the usage of the NLPyPort pipeline, some structural changes were made. The “exemplo.py” file shows exemples os several use cases.

How to use the pipeline

Depending on the planed usage, the pipeline may be called in three different ways:

1 - Default

text = new_full_pipe( your_input_file )

2 - Optional arguments

text = new_full_pipe( your_input_file , options = options )

3 - Optional arguments and pre-load pipeline

config_list = load_congif_to_list() 		# Pre-load the pipeline
text=new_full_pipe( your_input_file , options = options , config_list = config_list)

Available options

"tokenizer" : True -> Perform Tokenization

"pos_tagger" : True -> Perform Pos Tagging

"lemmatizer" : True -> Perform Lemmatization

"entity_recognition" : True -> Perform NER

"np_chunking" : True -> Perform NP Chunking

"pre_load" : False -> Preload the pipeline, needs the additional argument “config_list”

"string_or_array" : True -> Set input as being an array or a string

Returned text

In case of success, the pipeline will return an object of the “Text” class. The properties of this are as follow: text.tokens text.pos_tags text.lemas text.entities text.np_tags

Additionally, there is a method to return the pipeline in the CoNNL Format: text.print_conll()

To separate lines , at the end of each line the additional token EOS is added.

Credits

Tokenizer and Lemmatizer resource files - Rodrigues, Ricardo, Hugo Gonçalo Oliveira, and Paulo Gomes. "NLPPort: A Pipeline for Portuguese NLP (Short Paper)." 7th Symposium on Languages, Applications and Technologies (SLATE 2018). Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik, 2018.

Lemmatizer design - Rodrigues, Ricardo, Hugo Gonçalo Oliveira, and Paulo Gomes. "LemPORT: a high-accuracy cross-platform lemmatizer for portuguese." 3rd Symposium on Languages, Applications and Technologies. Schloss Dagstuhl-Leibniz-Zentrum fuer Informatik, 2014.

PoS trainer (adapted from) - https://github.com/fmaruki/Nltk-Tagger-Portuguese

Named Entity Recognition
CRF suite - Naoaki Okazaki http://www.chokkan.org/software/crfsuite/ sklearn-crfsuite wrapper - https://github.com/TeamHG-Memex/sklearn-crfsuite

Corpus Corpus for PoS tagging training MacMorpho - http://nilc.icmc.usp.br/macmorpho/ Floresta Sintá(c)tica - https://www.linguateca.pt/Floresta/corpus.html

Citations

To cite and give credits to the pipeline please use the following BibText reference:

@inproceedings{ferreira_etal:slate2019, Author = {João Ferreira and Hugo {Gonçalo~Oliveira} and Ricardo Rodrigues}, Booktitle = {Symposium on Languages, Applications and Technologies (SLATE 2019)}, Month = {June}, Note = {In press}, Title = {Improving {NLTK} for Processing {P}ortuguese}, Year = {2019}}