de.up.ling.irtg.codec

Class BolinasHrgInputCodec

java.lang.Object

de.up.ling.irtg.codec.InputCodec<InterpretedTreeAutomaton>

de.up.ling.irtg.codec.BolinasHrgInputCodec

public class BolinasHrgInputCodec
extends InputCodec<InterpretedTreeAutomaton>

An input codec for reading hyperedge replacement grammars (HRGs) in the input format for the Bolinas parser. The codec reads a monolingual graph grammar and converts it into an IRTG with a single interpretation, called "Graph", over the GraphAlgebra .

Because the graph algebra only represents graphs (and not hypergraphs), the conversion will only be successful if every ordinary hyperedge in the rules (i.e., every hyperedge that is not labeled with a nonterminal) has one or two endpoints. These hyperedges are translated as follows:

• Hyperedges with two endpoints are translated into ordinary labeled edges.
• By default, hyperedges with a single endpoint (and label L) are translated into node labels (i.e., the source node of the edge is taken to carry the label L).
• You can call setConvertUnaryEdgesToNodeLabels(boolean) to switch to a behavior where hyperedges with a single endpoint are translated into loops, i.e. into edges from the source node to itself with the given edge label.

Whether you want the loop encoding of unary edges or the node label encoding depends on how you represent node labels in the graphs you're trying to parse. The unmodified AMR-Bank uses node labels, which is why the node-label encoding is the default behvior of the codec.

Nonterminal hyperedges are treated differently, and may still have an arbitrary number of endpoints.

The codec allows you to specify external nodes of the graph on the right-hand side of a rule either with an anonymous marker (".*") or with an explicit marker ("*.2"). The Bolinas documentation does not specify precisely how anonymous and explicit markers can be mixed, so we recommend against mixing both kinds in the same rule. Anonymous markers are translated into external nodes in ascending order, from left to right in the rule. Explicit markers are translated into external nodes in ascending order; note that it is okay to use e.g. *.1 and *.3 but not *.2. The root of the RHS graph is always translated into the first external node.

A note of caution: This class is not thread-safe. If you want to use it in a multi-threaded environment, you should make a separate codec object for each thread.

Author:

koller

Constructor Summary

Constructors

Constructor and Description
BolinasHrgInputCodec()

Method Summary

Modifier and Type	Method and Description
boolean	isConvertUnaryEdgesToNodeLabels() Returns the current behavior with respect to encoding non-nonterminal unary hyperedges.
InterpretedTreeAutomaton	read(InputStream is) Reads an object from an input stream.
void	setConvertUnaryEdgesToNodeLabels(boolean convertUnaryEdgesToNodeLabels) Select how the codec should encode non-nonterminal hyperedges with single endpoints.

Methods inherited from class de.up.ling.irtg.codec.InputCodec

addOptions, getAllInputCodecs, getInputCodecByExtension, getInputCodecByName, getInputCodecByNameOrExtension, getInputCodecs, getMetadata, getOption, hasTrueOption, main, read, setOption, setProgressListener

Methods inherited from class java.lang.Object

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

Constructor Detail

BolinasHrgInputCodec

public BolinasHrgInputCodec()

Method Detail

isConvertUnaryEdgesToNodeLabels

public boolean isConvertUnaryEdgesToNodeLabels()

Returns the current behavior with respect to encoding non-nonterminal unary hyperedges.

Returns:

See Also:

setConvertUnaryEdgesToNodeLabels(boolean)

setConvertUnaryEdgesToNodeLabels

public void setConvertUnaryEdgesToNodeLabels(boolean convertUnaryEdgesToNodeLabels)

Select how the codec should encode non-nonterminal hyperedges with single endpoints. If the argument is "true" (the default), unary hyperedges are encoded as node labels. If the argument is "false", unary hyperedges are encoded as labeled loops.

Parameters:

convertUnaryEdgesToNodeLabels -

read

public InterpretedTreeAutomaton read(InputStream is)
                              throws CodecParseException,
                                     IOException

Description copied from class: InputCodec

Reads an object from an input stream.

Specified by:

read in class InputCodec<InterpretedTreeAutomaton>

Returns:

Throws:

CodecParseException - if an error occurred while decoding the input stream into an object

IOException - if an error occurred while reading data from the input stream