• Evaluation Metrics, ROC-Curves and imbalanced datasets (19 Aug 2018)
    This blog post describes some evaluation metrics used in NLP, it points out where we should use each one of them and the advantages and disadvantages of each.

  • Document Classification (01 Apr 2017)
    An introduction to the Document Classification task, in this case in a multi-class and multi-label scenario, proposed solutions include TF-IDF weighted vectors, an average of word2vec words-embeddings and a single vector representation of the document using doc2vec. Includes code using Pipeline and GridSearchCV classes from scikit-learn.

viterbi sequence-prediction pos-tags neural-networks word2vec scikit-learn conditional-random-fields NER word-embeddings syntactic-dependencies gensim fasttext evaluation_metrics document-classification classification SyntaxNet NLTK LSTM tokenization tf-idf stanford-NER seq2seq relationship-extraction recurrent-neural-networks portuguese nlp named-entity-recognition naive-bayes multi-label-classification maximum-entropy-markov-models machine-translation logistic-regression language-models information-extraction imbalanced_data hyperparameter-optimization hidden-markov-models grid-search glove embeddings doc2vec dependency-graph deep-learning data-challenge convolutional-neural-networks conference character-language-models character-embeddings attention RNN PyData KOVENS GRU ELMo BERT