amazing-nlp | Eventually we 'll build an universal deep NLP toolkits

amazing-nlp is a Python library. amazing-nlp has no bugs, it has no vulnerabilities, it has build file available and it has low support. You can download it from GitHub.

use simple linear regression to fit a line. apply logistic regression to MNIST digit recognition. apply convolution neural network to MNIST digit recognition. apply convolution neural network to CIFAR-10 object recognition. use recurrent neural network to reconstruct sine wave. apply multi-layers LSTM to milk production time-series prediction. use autoencoder as PCA to perform dimension reduction from 3D to 2D. use autoencoder as PCA to perform dimension reduction from 30D to 2D. use autoencoder to generate new training data for MNIST. use genarative adversarial network to generate new training data for MNIST. apply fully-connected network to wine classification with sparse categorical cross entropy. apply fully-connected network to iris classification with categorical cross entropy. investigate the relations between output, hidden states, cell states of LSTM and BiLSTM and their shapes. investigate the relations between output and hidden states of GRU and BiGRU and their shapes. load pre-trained word2vec model into gensim. load pre-trained Glove model into numpy array. use bag-of-words and pre-trained Glove to perform text classification. train a new word2vec model using Tensorflow's nce-loss. train a new word2vec model using Tensorflow, but doesn't use nce-loss, do the negative sampling ourselves instead. use gensim API to train a new word2vec model. implement and train a new Glove model using Tensorflow. perform name-entity recognition using Tensorflow. perform name-entity recognition using Keras. train a language model based on Moby Dick, and generate new text sequences using this model the input length of the model is fixed, so when we do the sampling, we need 0 ~ T inputs to generate T + 1 output. use LSTM and pre-trained Glove to perform text classification. prove that not only RNN, CNN can be used on time-series data and perform text classification, too. apply a simple memory network to bAbI dataset, build a model can give a yes/no answer based on the given story and question. prove that not only CNN, RNN can be used on image data and perform digit recognition, too. train a language model based on Robert Frost poetry, and generate new peotry sequences using this model we create a model with input length 1 besides the trained model (which has input length T) so when we do the sampling, we only need T input to generate T + 1 output.