Tanl::POS::SuffixGuesser Struct Reference

The task of the suffix guesser is to predict a tag-distribution based on the suffix of the word. More...

#include <SuffixGuesser.h>

List of all members.

Public Member Functions
void	serialize (std::ostream &out)
	Serializes a SuffixGuesser object.
void	serialize (std::istream &in)
	De-Serializes a SuffixGuesser object.
void	add_word (int n, std::string &word, TagID tag, int count)
	Adds a word to the suffix trie.
double	tagprob (std::string &word, int tagid)
	TODO.
double	tagprobs (std::string &word, std::vector< double > &probs)
	TODO.
Static Public Member Functions
static double	calculate_theta (std::vector< double > &apriori_tag_probs)
	TODO.
Public Attributes
double	theta
	Theta used in the interpolation process.
TrieNode	trie
	Trie of suffices.
Counts	empty_counts
	Empty Counts object.

---

Detailed Description

The task of the suffix guesser is to predict a tag-distribution based on the suffix of the word.

In training phase, it calculates for each suffix its count in the corpus, in total and for each tag separately. Let's assume a word ending with ABCDE. During prediction, it linearly interpolates the looked-up predictions for the ABCDE, BCDE, CDE, DE, E, "" suffices. Interpolation is done successively with weights 1 and theta, so weights are basically powers of 1/(1+theta), with the shorter suffix getting the larger weight.

---

Member Function Documentation

void Tanl::POS::SuffixGuesser::add_word	(	int	n,
		std::string &	word,
		TagID	tag,
		int	count
	)

Adds a word to the suffix trie.

Parameters:

	n	Max suffix size.
	word	String to be added to the trie.
	tag	Tag identifier used to tag the word we are trying to add.
	count	Amount of times word was tagged with tag inside the corpus.

void Tanl::POS::SuffixGuesser::serialize

(

std::istream &

in

)

De-Serializes a SuffixGuesser object.

Parameters:

in

The stream from which the object will be read

References Tanl::POS::TrieNode::serialize(), serialize(), theta, and trie.

void Tanl::POS::SuffixGuesser::serialize

(

std::ostream &

out

)

Serializes a SuffixGuesser object.

Parameters:

out

The stream wherein the object will be written

References Tanl::POS::TrieNode::serialize(), theta, and trie.

Referenced by serialize().

---

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

• tag/Tanlpos/SuffixGuesser.h
• tag/Tanlpos/SuffixGuesser.cpp