Skip to content

Latest commit

 

History

History
79 lines (59 loc) · 5.05 KB

Training.md

File metadata and controls

79 lines (59 loc) · 5.05 KB
Raw

Training

Training new models is simple. Look at Train_POS.py and Train_Chunking.py for examples.

Place new datasets in the folder data. The system expects three files train.txt, dev.txt and test.txt in a CoNLL format. I.e. each token is in a new line, different columns are seperated by a white space (either a space or a tab). Sentences are seperated by an empty line.

For an example look at data/conll2000_chunking/train.txt. Files with multiple columns, like data/unidep_pos/train.txt are no problem, as we will specify later which columns should be used for training.

Specifying datasets

The dataset is specified in the datasets variable (Train_POS.py):

datasets = {
    'unidep_pos':                            #Name of the dataset
        {'columns': {1:'tokens', 3:'POS'},   #CoNLL format for the input data. Column 1 contains tokens, column 3 contains POS information
         'label': 'POS',                     #Which column we like to predict
         'evaluate': True,                   #Should we evaluate on this task? Set true always for single task setups
         'commentSymbol': None}              #Lines in the input data starting with this string will be skipped. Can be used to skip comments
}

unidep_pos defines the name of the dataset. The code looks for the train.txt/dev.txt/test.txt in the folder data/unidep_pos. If your data folder has a different name, you must change the value of unidep_pos.

columns specifies the columns that should be read from the CoNLL file, in this case the column 1 and the column 3 should be read. The counting starts at 0. Column 1 contains the tokens, and the column 3 the POS-tag. Note, that we must always specify a 'tokens' column. The other columns can be named arbitrarily. If your data format changes, you must change the columns value.

label will specify which column should serve as label, in this case we want to predict the POS column. The name must match with the name specified in the dictionary dataColumns.

evaluate defines if we want to evaluate on this task. Must be true for single-task learning.

commentSymbol - Lines starting with commentSymbol will be skipped when the data files are read. We set it here to None as our input data does not have comment lines.

Word Embeddings

The path to pre-trained word embeddings must be specified in the variable embeddingsPath:

embeddingsPath = 'komninos_english_embeddings.gz'

The format for the pre-trained embeddings file must be text-based, i.e. each line contains the embedding for a word. The first column in that line is the word, followed by the dense vector. Our script will automatically download the embeddings komninos_english_embeddings.gz if they are not present in the current folder. The data can be compressed using .gz.

Training the network

The network is trained using the following lines of code (Train_POS.py):

params = {'classifier': ['CRF'], 'LSTM-Size': [100], 'dropout': (0.25, 0.25)}

model = BiLSTM(params)
model.setMappings(mappings, embeddings)
model.setDataset(datasets, data)
model.storeResults('results/unidep_pos_results.csv') #Path to store performance scores for dev / test
model.modelSavePath = "models/[ModelName]_[DevScore]_[TestScore]_[Epoch].h5" #Path to store models
model.fit(epochs=25)

params defines the hyperparameters of the network. See Hyperparameters.md for more details.

model.modelSavePath defines the path where the trained models should be stored. [ModelName] is replaced by the name of your dataset, [DevScore] with the score on the development set, [TestScore] with the score on the test set and [Epoch] is replaced by the epoch.

Storing performance Scores

By calling the model.storeResults() we specify the path where the performance scores during training should be stored. The file contains for each training epoch a line that contains the following information:

  • epoch
  • dataset name
  • Performance on development set
  • Performance on test set
  • Highest development set performance so far
  • Test performance for epoch with highest development score

Training BIO-Encoded Labels

If you want to perform chunking instead of POS-tagging, simple change the datasets variable (Train_Chunking.py):

datasets = {
    'conll2000_chunking':                                   #Name of the dataset
        {'columns': {0:'tokens', 1:'POS', 2:'chunk_BIO'},   #CoNLL format for the input data. Column 0 contains tokens, column 2 contains POS and column 2 contains chunk information using BIO encoding
         'label': 'chunk_BIO',                                #Which column we like to predict
         'evaluate': True,                                  #Should we evaluate on this task? Set true always for single task setups
         'commentSymbol': None}                             #Lines in the input data starting with this string will be skipped. Can be used to skip comments
}

Note: By appending a _BIO to a column name, we indicate that this column is BIO encoded. The system will then compute the F1-score instead of the accuracy.