de.up.ling.irtg.automata.coarse_to_fine

Class CoarseToFineParser

java.lang.Object

de.up.ling.irtg.automata.coarse_to_fine.CoarseToFineParser

public class CoarseToFineParser
extends Object

This class implements coarse-to-fine parsing. This builds an intersection automaton for a given input in multiple refinement steps and filters low weight constituents at every step. This can be used to significantly decrease parsing times.

Author:

koller

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
class	CoarseToFineParser.Combination Used to store evaluation Information, not useful for most users.

Field Summary

Fields

Modifier and Type	Field and Description
static boolean	DEBUG Set this variable to true in order for debugging information to be printed to System.err.

Constructor Summary

Constructors

Constructor and Description
CoarseToFineParser(InterpretedTreeAutomaton irtg, String inputInterpretation, FineToCoarseMapping ftc, double theta) Creates a new instance from the given FineToCoarseMapping.

Method Summary

Modifier and Type	Method and Description
static CoarseToFineParser	makeCoarseToFineParser(InterpretedTreeAutomaton irtg, String interpretation, String ftcMap, double theta) This is a convenience method which constructs a coarse to fine parser with the given parameters.
TreeAutomaton	parse(String input) This method is a shorter alias for parseInputObject(java.lang.Object) which also takes care of decoding the string representation of the input into a suitable object.
TreeAutomaton	parseInputObject(Object inputObject) Uses coarse to fine parsing to produce a pruned parse chart for the input object.
CoarseToFineParser.Combination	parseInputObjectTrackSizes(Object inputObject) This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.
CoarseToFineParser.Combination	parseInputObjectTrackTimes(Object inputObject) This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.
TreeAutomaton	parseInputObjectWithSF(Object inputObject) Uses coarse to fine parsing to produce a pruned parse chart for the input object.
CoarseToFineParser.Combination	parseInputObjectWithSFTrackSizes(Object inputObject) This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.
CoarseToFineParser.Combination	parseInputObjectWithSFTrackTimes(Object inputObject) This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.

Methods inherited from class java.lang.Object

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

Field Detail

DEBUG

public static boolean DEBUG

Set this variable to true in order for debugging information to be printed to System.err.

Constructor Detail

CoarseToFineParser

public CoarseToFineParser(InterpretedTreeAutomaton irtg,
                          String inputInterpretation,
                          FineToCoarseMapping ftc,
                          double theta)

Creates a new instance from the given FineToCoarseMapping. each instance will always use the same FineToCoarseMapping. The irtg is equal to the finest level of the mapping. The class also assumes that there is a single interpretation in the IRTG which serves as the source of inputs. This interpretation is specified at this step.

Parameters:

irtg -

inputInterpretation -

ftc -

theta - threshold, if the outside+inside+ruleweight for any given rule is less than the total inside weight of a all parses for a given level of coarseness times theta, then the rule is ignored for the following refinement step.

Method Detail

parse

public TreeAutomaton parse(String input)
                    throws ParserException

This method is a shorter alias for parseInputObject(java.lang.Object) which also takes care of decoding the string representation of the input into a suitable object.

Parameters:

input -

Returns:

Throws:

ParserException

parseInputObject

public TreeAutomaton parseInputObject(Object inputObject)

Uses coarse to fine parsing to produce a pruned parse chart for the input object. This uses the coarse to fine chart that has been passed at construction as well as the given pruning threshold. Note that intersection at the coarsest level is done using the condensed intersection algorithm. For some grammars it might improve performance to use the sibling finder algorithm instead as implemented by parseInputObjectWithSF(java.lang.Object)

Parameters:

inputObject -

Returns:

parseInputObjectWithSF

public TreeAutomaton parseInputObjectWithSF(Object inputObject)

Uses coarse to fine parsing to produce a pruned parse chart for the input object. This uses the coarse to fine chart that has been passed at construction as well as the given pruning threshold. Note that this implementation uses the sibling finder algorithm intersection for the coarsest level of parsing which can be helpful for grammars with more complex homomorphic images.

Parameters:

inputObject -

Returns:

parseInputObjectWithSFTrackSizes

public CoarseToFineParser.Combination parseInputObjectWithSFTrackSizes(Object inputObject)

This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.

Parameters:

inputObject -

Returns:

parseInputObjectWithSFTrackTimes

public CoarseToFineParser.Combination parseInputObjectWithSFTrackTimes(Object inputObject)

This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.

Parameters:

inputObject -

Returns:

parseInputObjectTrackSizes

public CoarseToFineParser.Combination parseInputObjectTrackSizes(Object inputObject)
                                                          throws ClassNotFoundException,
                                                                 InstantiationException,
                                                                 IllegalAccessException

This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.

Parameters:

inputObject -

Returns:

Throws:

ClassNotFoundException

InstantiationException

IllegalAccessException

parseInputObjectTrackTimes

public CoarseToFineParser.Combination parseInputObjectTrackTimes(Object inputObject)
                                                          throws ClassNotFoundException,
                                                                 InstantiationException,
                                                                 IllegalAccessException

This method is used for evaluation of coarse-to-fine parsing with Alto Lab, most users will not need it.

Parameters:

inputObject -

Returns:

Throws:

ClassNotFoundException

InstantiationException

IllegalAccessException

makeCoarseToFineParser

public static CoarseToFineParser makeCoarseToFineParser(InterpretedTreeAutomaton irtg,
                                                        String interpretation,
                                                        String ftcMap,
                                                        double theta)
                                                 throws FileNotFoundException,
                                                        IOException,
                                                        ParseException

This is a convenience method which constructs a coarse to fine parser with the given parameters. The ftc mapping is given in the following format (the outermost level "__" is ignore and only written to ensure that the whole structure forms a tree): __( coarse_nonterminal1 ( fine_nonterminal1_1 ( even_finer_nonterminal1_1_1( ..... ) even_finer_nonterminal1_1_2( ..... ) ) fine_nonterminal1_2 ( .... ) coarse_nonterminal2 ( ... ) ) Note that the default FTC mapping will assume that the value of InsideRuleFactory.NONTERMINAL_SEPARATOR indicates that a nonterminal consists of multiple parts and each part after a separator should be mapped individually.

Parameters:

irtg - grammar corresponding to finest level of analysis

interpretation - interpretation which the objects to be parsed will come from

ftcMap -

theta - pruning threshold as described for the constructor

Returns:

Throws:

FileNotFoundException

IOException

ParseException