Skip to content

Latest commit

 

History

History
156 lines (108 loc) · 6.55 KB

scripts.rst

File metadata and controls

156 lines (108 loc) · 6.55 KB

Standalone Scripts

:mod:`nlpnet` includes standalone scripts that may be called from a command line. They are copied to the scripts subdirectory of your Python installation, which can be included in the system PATH variable. There are three such scripts:

nlpnet-train
Script to train a new model or further train an existing one.
nlpnet-test
Script to measure the performance of a model against a gold data set.
nlpnet-tag
Script to call a model and tag some given text.

Each of them is explained below.

nlpnet-tag

This is the simplest :mod:`nlpnet` script. It simply runs the system for a given text input. It should be called with the following syntax:

$ nlpnet-tag.py TASK DATA_DIRECTORY

Where TASK is either pos or srl and DATA_DIRECTORY is the directory with the trained models. It has also the following command line options:

-v Verbose mode
--no-repeat Forces the classification step to avoid repeated argument labels (SRL only).

For example:

$ nlpnet-tag.py pos /path/to/nlpnet-data/
O rato roeu a roupa do rei de Roma.
O_ART rato_N roeu_V a_ART roupa_N do_PREP+ART rei_N de_PREP Roma_NPROP ._PU

Or with semantic role labeling:

$ nlpnet-tag.py srl /path/to/nlpnet-data/
O rato roeu a roupa do rei de Roma.
O rato roeu a roupa do rei de Roma .
roeu
    A1: a roupa do rei de Roma
    A0: O rato
    V: roeu

The first line was typed by the user, and the second one is the result of tokenization.

nlpnet-train

There are a lot of training parameters that can be supplied to :mod:`nlpnet`. Some of them depend on the task that the network is being trained for, as the type of network can be a simple MLP for POS tagging and a convolutional network for SRL.

General Options

These options can be used in either POS or SRL training.

-w NUMBER The size of the word window. For SRL, the supplied model used 3, and for POS, 5. It is important to have a reasonably large window in POS so the tagger can analyze the context.
-n NUMBER Number of hidden neurons.
-f NUMBER Generates feature vectors randomly with the given number of dimensions for words. Ignore it if you supply pre-initialized representations.
--load_features
 Loads the features vectors representing words. The file containing the data must be set in config.py and be in the data/ directory. Nlpnet uses numpy files for storing representations as 2-dimensional arrays.
-e NUMBER Number of epochs to train the network.
-l NUMBER The learning rate for network weights.
--lf NUMBER The learning rate for features (including extra features like the ones from --caps).
--lt NUMBER The learning rate for the tag transition scores.
--caps NUMBER Include capitalization as a feature. If a number is given, determine the number of features (default 5).
--suffix NUMBER
 Same as --caps, but for suffixes. It will search a file named suffixes.txt in the data/ directory, and read each line as suffix.
-a NUMBER Stop training when the network achieves this accuracy. Useful to avoid divergence when the learning rate is high.
-v Verbose mode, it will output more information about what is happening internally.
--load_network Loads a previously saved network. The file name must be set in config.py and be in the data/ directory.
--task TASK Task to train for. It must be either srl or pos.
--data DIRECTORY
 The directory containing the model files. If a new model is being trained, everything is saved to that dir.
--gold FILE A file containing the gold data used for training.

Data files must be in the format used by :mod:`nlpnet`. A POS file must have one sentence per line, each sentence containing tokens in the format token_tag and separated by whitespace. SRL files must be in the CoNLL format.

SRL

-c NUMBER Number of neurons in the convolution layer.
--pos NUMBER Uses POS as a feature. Currently, it must read the tags from the training data. Works same as --caps.
--chunk NUMBER Uses syntactic chunks as a feature. Same as --pos.
--use_lemma Reads word lemmas instead of surface forms. It needs to read them from the training data.
--id Train for argument boundary identification only.
--class Train for previously identified argument classification only. (if neither this or --id is supplied, trains a network that does both in a single step)
--pred Train for predicate recognizing only.
--max_dist NUMBER
 The maximum distance (to predicates and target words) to have an own feature vector. Any distance greater than this will be mapped to a single vector.
--target_features NUMBER
 Number of features for vectors representing distance to the target word.
--pred_features NUMBER
 Same as --target_features for the predicate.

nlpnet-test

This script is much simpler. It evaluates the system performance against a gold standard.

General options

The arguments below are valid for both tasks.

--task TASK Task for which the network should be used. Either pos or srl.
-v Verbose mode
--gold FILE File with gold standard data
--data DIRECTORY
 Directory with trained models

POS

--oov FILE Analyze performance on the words described in the given file.

The --oov option requires a UTF-8 file containing one word per line. Actually, this option is not exclusive for OOV (out-of-vocabulary) words, but rather any word list you want to evaluate.

SRL

SRL evaluation is performed in different ways, depending on whether it is aimed at argument identification, classification, predicate detection or all of them. In the future, there may be a more standardized version for this test.

--id Evaluate only argument identification (SRL only). The script will output the score.
--class Evaluate only argument classification (SRL only). The script will output the score.
--preds Evaluate only predicate identification (SRL only). The script will output the score.
--2steps Execute SRL with two separate steps. The script will output the results in CoNLL format.
--no-repeat Forces the classification step to avoid repeated argument labels (2 step SRL only)
--auto-pred Determines SRL predicates automatically. Only used when evaluating the full process (identification + classification)

The CoNLL output can be evaluated against a gold file using the official SRL eval script (see http://www.lsi.upc.edu/~srlconll/soft.html).