Config (Stanford JavaNLP API)

java.lang.Object
- edu.stanford.nlp.parser.nndep.Config

```
public class Config
extends java.lang.Object
```
Defines configuration settings for training and testing the neural-network dependency parser.

Author:

Danqi Chen, Jon Gauthier

See Also:

DependencyParser

Field Summary

Fields
Modifier and Type	Field and Description
`double`	`adaAlpha` Initial global learning rate for AdaGrad training
`double`	`adaEps` An epsilon value added to the denominator of the AdaGrad expression for numerical stability
`int`	`batchSize` Size of mini-batch for training.
`int`	`clearGradientsPerIter` During training, clear AdaGrad gradient histories after every `clearGradientsPerIter` iterations.
`boolean`	`cPOS` Use coarse POS instead of fine-grained POS if cPOS = true.
`boolean`	`doWordEmbeddingGradUpdate` Update word embeddings when performing gradient descent.
`double`	`dropProb` Dropout probability.
`int`	`embeddingSize` Dimensionality of the word embeddings used
`java.util.function.Function<java.util.List<HasWord>,java.util.List<HasWord>>`	`escaper` Defines a word-escaper to use when parsing raw sentences.
`int`	`evalPerIter` During training, run a full UAS evaluation after every `evalPerIter` iterations.
`int`	`hiddenSize` Size of the neural network hidden layer.
`double`	`initRange` Model weights will be initialized to random values within the range `[-initRange, initRange]`.
`Language`	`language` The language being parsed.
`int`	`maxIter` Maximum number of iterations for training
`static int`	`NONEXIST` Represent a non-existent token.
`boolean`	`noPunc` Exclude punctuations in evaluation if noPunc = true.
`static java.lang.String`	`NULL` Non-existent token string.
`int`	`numCached` Number of hidden layer activations to cache.
`int`	`numPreComputed` Number of input tokens for which we should compute hidden-layer unit activations.
`static int`	`numTokens` Total number of tokens provided as input to the classifier.
`boolean`	`preTokenized` Provided text is tokenized by whitespace.
`double`	`regParameter` Regularization parameter.
`static java.lang.String`	`ROOT` Root token string.
`boolean`	`saveIntermediate` Save an intermediate model file whenever we see an improved UAS evaluation.
`java.lang.String`	`sentenceDelimiter` If non-null, when parsing raw text assume sentences have already been split and are separated by the given delimiter.
`static java.lang.String`	`SEPARATOR` For printing messages.
`java.lang.String`	`tagger` Path to a tagger file compatible with `MaxentTagger`.
`TreebankLanguagePack`	`tlp` Describes language-specific properties necessary for training and testing.
`int`	`trainingThreads` Number of threads to use during training.
`static java.lang.String`	`UNKNOWN` Out-of-vocabulary token string.
`boolean`	`unlabeled` Train a labeled parser if labeled = true, and a unlabeled one otherwise.
`int`	`wordCutOff` Refuse to train on words which have a corpus frequency less than this number.

Constructor Summary

Constructors
Constructor and Description

Config(java.util.Properties properties)

Constructors
Constructor and Description
`Config(java.util.Properties properties)`

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`static Language`	`getLanguage(java.lang.String languageStr)` Get the `Language` object corresponding to the given language string.
`void`	`printParameters()`

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Field Detail
  - UNKNOWN
```
public static final java.lang.String UNKNOWN
```
    Out-of-vocabulary token string.
    
    See Also:
    
    Constant Field Values
  - ROOT
```
public static final java.lang.String ROOT
```
    Root token string.
    
    See Also:
    
    Constant Field Values
  - NULL
```
public static final java.lang.String NULL
```
    Non-existent token string.
    
    See Also:
    
    Constant Field Values
  - NONEXIST
```
public static final int NONEXIST
```
    Represent a non-existent token.
    
    See Also:
    
    Constant Field Values
  - SEPARATOR
```
public static final java.lang.String SEPARATOR
```
    For printing messages.
    
    See Also:
    
    Constant Field Values
  - language
```
public Language language
```
    The language being parsed.
  - trainingThreads
```
public int trainingThreads
```
    Number of threads to use during training. Also indirectly controls how mini-batches are partitioned (more threads => more partitions => smaller partitions).
  - wordCutOff
```
public int wordCutOff
```
    Refuse to train on words which have a corpus frequency less than this number.
  - initRange
```
public double initRange
```
    Model weights will be initialized to random values within the range [-initRange, initRange].
  - maxIter
```
public int maxIter
```
    Maximum number of iterations for training
  - batchSize
```
public int batchSize
```
    Size of mini-batch for training. A random subset of training examples of this size will be used to train the classifier on each iteration.
  - adaEps
```
public double adaEps
```
    An epsilon value added to the denominator of the AdaGrad expression for numerical stability
  - adaAlpha
```
public double adaAlpha
```
    Initial global learning rate for AdaGrad training
  - regParameter
```
public double regParameter
```
    Regularization parameter. All weight updates are scaled by this single parameter.
  - dropProb
```
public double dropProb
```
    Dropout probability. For each training example we randomly choose some amount of units to disable in the neural network classifier. This probability controls the proportion of units "dropped out."
  - hiddenSize
```
public int hiddenSize
```
    Size of the neural network hidden layer.
  - embeddingSize
```
public int embeddingSize
```
    Dimensionality of the word embeddings used
  - numTokens
```
public static final int numTokens
```
    Total number of tokens provided as input to the classifier. (Each token is provided in word embedding form.)
    
    See Also:
    
    Constant Field Values
  - numPreComputed
```
public int numPreComputed
```
    Number of input tokens for which we should compute hidden-layer unit activations. If zero, the parser will skip the pre-computation step.
  - numCached
```
public int numCached
```
    Number of hidden layer activations to cache. Only applies at test time.
  - evalPerIter
```
public int evalPerIter
```
    During training, run a full UAS evaluation after every evalPerIter iterations.
  - clearGradientsPerIter
```
public int clearGradientsPerIter
```
    During training, clear AdaGrad gradient histories after every clearGradientsPerIter iterations. (If zero, never clear gradients.)
  - saveIntermediate
```
public boolean saveIntermediate
```
    Save an intermediate model file whenever we see an improved UAS evaluation. (The frequency of these evaluations is configurable as well; see evalPerIter.)
  - unlabeled
```
public boolean unlabeled
```
    Train a labeled parser if labeled = true, and a unlabeled one otherwise.
  - cPOS
```
public boolean cPOS
```
    Use coarse POS instead of fine-grained POS if cPOS = true.
  - noPunc
```
public boolean noPunc
```
    Exclude punctuations in evaluation if noPunc = true.
  - doWordEmbeddingGradUpdate
```
public boolean doWordEmbeddingGradUpdate
```
    Update word embeddings when performing gradient descent. Set to false if you provide embeddings and do not want to finetune.
  - tlp
```
public TreebankLanguagePack tlp
```
    Describes language-specific properties necessary for training and testing. By default, PennTreebankLanguagePack will be used.
  - sentenceDelimiter
```
public java.lang.String sentenceDelimiter
```
    If non-null, when parsing raw text assume sentences have already been split and are separated by the given delimiter. If null, the parser splits sentences automatically.
  - escaper
```
public java.util.function.Function<java.util.List<HasWord>,java.util.List<HasWord>> escaper
```
    Defines a word-escaper to use when parsing raw sentences. As a command-line option, you should provide the fully qualified class name of a valid escaper (that is, a class which implements Function<List<HasWord>, List<HasWord>>).
  - tagger
```
public java.lang.String tagger
```
    Path to a tagger file compatible with MaxentTagger.
  - preTokenized
```
public boolean preTokenized
```
    Provided text is tokenized by whitespace.
- Constructor Detail
  - Config
```
public Config(java.util.Properties properties)
```
- Method Detail
  - getLanguage
```
public static Language getLanguage(java.lang.String languageStr)
```
    Get the Language object corresponding to the given language string.
    
    Returns:
    
    A Language or null if no instance matches the given string.
  - printParameters
```
public void printParameters()
```

Class Config

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

UNKNOWN

ROOT

NULL

NONEXIST

SEPARATOR

language

trainingThreads

wordCutOff

initRange

maxIter

batchSize

adaEps

adaAlpha

regParameter

dropProb

hiddenSize

embeddingSize

numTokens

numPreComputed

numCached

evalPerIter

clearGradientsPerIter

saveIntermediate

unlabeled

cPOS

noPunc

doWordEmbeddingGradUpdate

tlp

sentenceDelimiter

escaper

tagger

preTokenized

Constructor Detail

Config

Method Detail

getLanguage

printParameters