FeatureTypeConcatRegex (Dragontool 1.3.3 API)

java.lang.Object
- dragon.ml.seqmodel.feature.AbstractFeatureType
- - dragon.ml.seqmodel.feature.FeatureTypeConcatRegex

All Implemented Interfaces:

FeatureType
```
public class FeatureTypeConcatRegex
extends AbstractFeatureType
```
ConcatRegexFeatures generates features by matching the token with the character patterns. Character patterns are regular expressions for checking whether the token is capitalized word, a number, small case word, whether the token contains any special characters and like. It uses regular expression to match a sequence of character pattern and generates features accordingly.
The feature generated here is whether a sequence of tokens has a particular sequence of given pattern or not. For example, if a pattern is to mathc a capital word, then for two token context window, various features generated are weither two token (bigram) sequence is having any of the following pattern or not: (1) Capital, Capital (2) Capital, Non-Capital (3) Non-capital, Capital. You can use any window around the current token (segment) for creating regular expression based features. Also, you can define your own patterns, by writing down the regular expression in a file, whose format is specified below.
A token in a token sequence has a index relative to the current token index, which is described below:
```
        x0 x1 x2 x3 x4 x5 x6 x7 .... xn
        -4 -3 -2 -1 0  0  0  1 2 ...
 
```
In above example, the current segment is from postion 4 to 6 with value of pos = 6 and prevPos = 3 in startScanFeaturesAt() call of FeatureGenerator. You can refer to any of the token relative to current position by using the index below the token sequence. Thus, you can create a pattern concat features for any token sequence in the neighbourhood of the current token, using relSegmentStart and relSegmentEnd. For, example to create pattern for two tokens to the left of the current token, following is the parameters to be passed to the constructor of the class:
Author:

Imran Mansuri This FeatureType can not be used independently because the label is not set. This feature type should be wrapped by FeatureTypeStateLoop.

Field Summary

Fields
Modifier and Type	Field and Description
`protected int`	`curId`
`protected DataSequence`	`data`
`protected int`	`idbase`
`protected int`	`index`
`protected int`	`left`
`protected int`	`maxSegmentLength`
`protected int`	`relSegmentEnd`
`protected int`	`relSegmentStart`
`protected int`	`right`
`protected int`	`window`

Fields inherited from class dragon.ml.seqmodel.feature.AbstractFeatureType
idPrefix

Constructor Summary

Constructors
Constructor and Description
`FeatureTypeConcatRegex(int relSegmentStart, int relSegmentEnd)`
`FeatureTypeConcatRegex(int relSegmentStart, int relSegmentEnd, int maxSegmentLength)`
`FeatureTypeConcatRegex(int relSegmentStart, int relSegmentEnd, int maxSegmentLength, String patternFile)` Constructs an object of ConcatRegexFeatures to be used to generate features for the token sequence as specified.
`FeatureTypeConcatRegex(int relSegmentStart, int relSegmentEnd, int maxSegmentLength, String[][] patternString)`
`FeatureTypeConcatRegex(int relSegmentStart, int relSegmentEnd, String patternFile)`

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`boolean`	`hasNext()` Tests if next feature exists
`Feature`	`next()` Gets the next feature
`boolean`	`startScanFeaturesAt(DataSequence data, int startPos, int endPos)` Starts scanning features of the given segment of a sequence

Methods inherited from class dragon.ml.seqmodel.feature.AbstractFeatureType
getTypeID, needTraining, readTrainingResult, saveTrainingResult, setTypeID, startScanFeaturesAt, supportSegment, train

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

- Field Detail
  - data
```
protected transient DataSequence data
```
  - index
```
protected int index
```
  - idbase
```
protected int idbase
```
  - curId
```
protected int curId
```
  - window
```
protected int window
```
  - relSegmentStart
```
protected int relSegmentStart
```
  - relSegmentEnd
```
protected int relSegmentEnd
```
  - maxSegmentLength
```
protected int maxSegmentLength
```
  - left
```
protected int left
```
  - right
```
protected int right
```
- Constructor Detail
  - FeatureTypeConcatRegex
```
public FeatureTypeConcatRegex(int relSegmentStart,
                              int relSegmentEnd,
                              int maxSegmentLength,
                              String patternFile)
```
    Constructs an object of ConcatRegexFeatures to be used to generate features for the token sequence as specified. You can specify the sequence of tokens on which the pattern has to be applied using relSegmentStart and relSegmentEnd, which denotes segment boundries. The maxSegmentLength denotes the maximum segment size, for normal CRF the value of maxSegmentLength is 1. There are certain default patterns defined in the class. You can specify your own pattern in a file, and pass the name of the file in this constructor. The file should begin with integer value for number of pattern in the file. This should be follwoed by one pattern definition on each line. The first word is the name of the pattern and second word is regular expression for the pattern.
    
    Parameters:
    
    fgen - a Model object
    
    relSegmentStart - index of the reltive position for left boundary
    
    relSegmentEnd - index of the reltive position for right boundary
    
    maxSegmentLength - maximum size of a segment
    
    patternFile - file which contains the pattern definition
  - FeatureTypeConcatRegex
```
public FeatureTypeConcatRegex(int relSegmentStart,
                              int relSegmentEnd,
                              int maxSegmentLength,
                              String[][] patternString)
```
  - FeatureTypeConcatRegex
```
public FeatureTypeConcatRegex(int relSegmentStart,
                              int relSegmentEnd,
                              int maxSegmentLength)
```
  - FeatureTypeConcatRegex
```
public FeatureTypeConcatRegex(int relSegmentStart,
                              int relSegmentEnd)
```
  - FeatureTypeConcatRegex
```
public FeatureTypeConcatRegex(int relSegmentStart,
                              int relSegmentEnd,
                              String patternFile)
```
- Method Detail
  - startScanFeaturesAt
```
public boolean startScanFeaturesAt(DataSequence data,
                                   int startPos,
                                   int endPos)
```
    Description copied from interface: FeatureType
    
    Starts scanning features of the given segment of a sequence
    
    Parameters:
    
    data - the sequence
    
    startPos - the start position of the segment
    
    endPos - the end position of the segment
    
    Returns:
    
    true if successfully
  - hasNext
```
public boolean hasNext()
```
    Description copied from interface: FeatureType
    
    Tests if next feature exists
    
    Returns:
    
    true if next feature exists
  - next
```
public Feature next()
```
    Description copied from interface: FeatureType
    
    Gets the next feature
    
    Returns:
    
    the next feature if any

Class FeatureTypeConcatRegex

Field Summary

Fields inherited from class dragon.ml.seqmodel.feature.AbstractFeatureType

Constructor Summary

Method Summary

Methods inherited from class dragon.ml.seqmodel.feature.AbstractFeatureType

Methods inherited from class java.lang.Object

Field Detail

data

index

idbase

curId

window

relSegmentStart

relSegmentEnd

maxSegmentLength

left

right

Constructor Detail

FeatureTypeConcatRegex

FeatureTypeConcatRegex

FeatureTypeConcatRegex

FeatureTypeConcatRegex

FeatureTypeConcatRegex

Method Detail

startScanFeaturesAt

hasNext

next