Tanl::SST::SstFeatureExtractor Class Reference

Extract features for SST. More...

#include <SstFeatureExtractor.h>

Inheritance diagram for Tanl::SST::SstFeatureExtractor:

List of all members.

Public Member Functions
	SstFeatureExtractor (Resources &resources)
void	analyze (Sentence *sent, int zone)
	Set the sentence from which to extract features.
void	extract (Classifier::Features &feats, const int &pos)
	Extract features at position pos in sentence and put them into .
void	reset ()
	Reset to initial state.
void	classified (int position, char const *className)
	Record that a token at given position has been classified in the given class.
Protected Attributes
Resources &	resources
bool	insideQuotes
TokenCategorizer	tokenCategorizer
	extracts token type
std::vector< EntityType >	tokenTypes
	types of sentence tokens
Sentence *	sentence
	sentence being analyzed
unordered_map< string, bool >	capitalized
	Words that appeared previously as capitalized.
Tanl::Text::NormWordSet	prevClass [NUM_CLASSES]
	List of words previously designated as given class.
Tanl::Text::NormWordSet	otherLast [NUM_CLASSES]
	other word in Cap sequence is in Last Words
Tanl::Text::NormWordSet	acronyms
	List of previously found acronyms.

---

Detailed Description

Extract features for SST.

---

Member Function Documentation

void Tanl::SST::SstFeatureExtractor::analyze	(	Sentence *	sent,
		int	zone
	)

Set the sentence from which to extract features.

References Tanl::SST::TokenCategorizer::analyze(), Tanl::Token::form, sentence, tokenCategorizer, and tokenTypes.

Referenced by Tanl::SST::SstEventStream::analyze().

void Tanl::SST::SstFeatureExtractor::extract	(	Classifier::Features &	feats,
		const int &	pos
	)			[virtual]

Extract features at position pos in sentence and put them into .

Local features include:

1. lexical word features
2. features that depend on surrounding words
3. features from dictionary lookup of word.

Implements Tanl::Classifier::FeatureExtractor< Classifier::Features, const int >.

References acronyms, Tanl::Token::attrIndex(), capitalized, Tanl::Text::NormWordSet::contains(), Tanl::Token::form, Tanl::Token::get(), Tanl::Text::NormWordSet::insert(), Tanl::SST::Resources::nClasses, otherLast, prevClass, Tanl::Text::RegExp::Pattern::replace(), sentence, and Tanl::Text::to_upper().

Referenced by Tanl::SST::SstEventStream::next().

void Tanl::SST::SstFeatureExtractor::reset

(

)

[virtual]

Reset to initial state.

Useful when reading several documents with the same tagger instance.

Reimplemented from Tanl::Classifier::FeatureExtractor< Classifier::Features, const int >.

References acronyms, capitalized, Tanl::SST::Resources::nClasses, otherLast, and prevClass.

Referenced by Tanl::SST::SstEventStream::reset().

---

Member Data Documentation

unordered_map<string, bool> Tanl::SST::SstFeatureExtractor::capitalized [protected]

Words that appeared previously as capitalized.

Referenced by extract(), and reset().

---

The documentation for this class was generated from the following files:

• tag/SST/SstFeatureExtractor.h
• tag/SST/SstFeatureExtractor.cpp