org.evosuite.lm

Class LangModel

java.lang.Object

org.evosuite.lm.LangModel

public class LangModel
extends Object

Represents a language model, a set of bigrams, unigrams and associated log-probabilities.

Field Summary

Fields

Modifier and Type	Field and Description
static String	END_OF_STRING
static String	START_NEW_WORD
static String	START_OF_STRING

Constructor Summary

Constructors

Constructor and Description
LangModel(String lmFileName) Load the language model.

Method Summary

Modifier and Type	Method and Description
boolean	isEndOfSentence(String character)
boolean	isMagicChar(String character)
double	predict_char_prob(int n) Method which returns the probability of the nth most likley character at the start of a sentence.
double	predict_char_prob(String pre, int n) Method which returns the probability of the nth most likely character, given a preceeding character (pre).
String	predict_char(int n)
String	predict_char(String pre, int n)
double	score(String str) Convenience method for score(String, boolean) with verbose flag set to false.
double	score(String str, boolean verbose) Splits a string into bigrams and calculates the language model score.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

START_OF_STRING

public static final String START_OF_STRING

See Also:

Constant Field Values

END_OF_STRING

public static final String END_OF_STRING

See Also:

Constant Field Values

START_NEW_WORD

public static final String START_NEW_WORD

See Also:

Constant Field Values

Constructor Detail

LangModel

public LangModel(String lmFileName)
          throws IOException

Load the language model.

Parameters:

lmFileName - path to a language model file.

Throws:

IOException - if the model file can't be found or read.

Method Detail

score

public double score(String str,
                    boolean verbose)

Splits a string into bigrams and calculates the language model score. For each bigram, it looks up the probability. The score is the geometric mean of the probability of each bigram in the string according to the model. If a given bigram isn't in the model, unigrams are used to estimate the probability of the bigram instead

Parameters:

str - String for which to compute the score

verbose - whether to print information

Returns:

score

public double score(String str)

Convenience method for score(String, boolean) with verbose flag set to false.

predict_char

public String predict_char(String pre,
                           int n)

Returns:

the nth most likely character to follow pre

predict_char

public String predict_char(int n)

Returns:

the nth most likely character that a string will start with

predict_char_prob

public double predict_char_prob(String pre,
                                int n)

Method which returns the probability of the nth most likely character, given a preceeding character (pre). Use in combination with the predict_char methods.

Returns:

the probability of the nth character that is most likely to appear

predict_char_prob

public double predict_char_prob(int n)

Method which returns the probability of the nth most likley character at the start of a sentence. N.B. Simply calls predict_char_prob/2 with preceeding char set to "".

Returns:

the probability associated with the nth most likely character to start a sentence

isMagicChar

public boolean isMagicChar(String character)

isEndOfSentence

public boolean isEndOfSentence(String character)