Collaboration diagram for joshua.decoder.JoshuaConfiguration:

[legend]

Classes
class	OOVItem
Public Member Functions
void	reset ()
void	processCommandLineOptions (String[] options)
void	readConfigFile (String configFile) throws IOException
void	sanityCheck ()
Static Public Member Functions
static String	normalize_key (String text)
Public Attributes
ArrayList< String >	tms = new ArrayList<String>()
String	weights_file = ""
String	default_non_terminal = FormatUtils.markup("X")
String	goal_symbol = FormatUtils.markup("GOAL")
ArrayList< OOVItem >	oovList = null
boolean	segment_oovs = false
boolean	lattice_decoding = false
boolean	amortized_sorting = true
boolean	constrain_parse = false
boolean	use_pos_labels = false
boolean	true_oovs_only = false
boolean	filter_grammar = false
int	pop_limit = 100
int	maxlen = 200
boolean	use_unique_nbest = true
boolean	include_align_index = false
int	topN = 1
String	outputFormat = "%i \|\|\| %s \|\|\| %f \|\|\| %c"
int	num_parallel_decoders = 1
String	hypergraphFilePattern = ""
boolean	mark_oovs = false
boolean	parse = false
ArrayList< String >	features = new ArrayList<String>()
ArrayList< String >	weights = new ArrayList<String>()
int	server_port = 0
boolean	rescoreForest = false
float	rescoreForestWeight = 10.0f
String	fragmentMapFile = null
boolean	fuzzy_matching = false
String	search_algorithm = "cky"
int	reordering_limit = 8
int	num_translation_options = 20
boolean	use_dot_chart = true
boolean	moses = false
boolean	show_weights_and_quit = false
String	input_file = null
String	n_best_file = null
boolean	source_annotations = false
String	weight_overwrite = ""
Static Public Attributes
static final String	SOFT_SYNTACTIC_CONSTRAINT_DECODING_PROPERTY_NAME = "fuzzy_matching"
Private Attributes
final Logger	logger = Logger.getLogger(JoshuaConfiguration.class.getName())

Detailed Description

Configuration file for Joshua decoder.

When adding new features to Joshua, any new configurable parameters should be added to this class.

Author:: Zhifei Li, zhife.nosp@m.i.wo.nosp@m.rk@gm.nosp@m.ail..nosp@m.com; Matt Post post@.nosp@m.cs.j.nosp@m.hu.ed.nosp@m.u

Member Function Documentation

static String joshua.decoder.JoshuaConfiguration.normalize_key ( String text ) [static]

Normalizes parameter names by removing underscores and hyphens and lowercasing. This defines equivalence classes on external use of parameter names, permitting arbitrary_under_scores and camelCasing in paramter names without forcing the user to memorize them all. Here are some examples of equivalent ways to refer to parameter names:

{pop-limit, poplimit, PopLimit, popLimit, pop_lim_it} {lmfile, lm-file, LM-FILE, lm_file}

Here is the caller graph for this function:

void joshua.decoder.JoshuaConfiguration.processCommandLineOptions ( String[] options )

To process command-line options, we write them to a file that looks like the config file, and then call readConfigFile() on it. It would be more general to define a class that sits on a stream and knows how to chop it up, but this was quicker to implement.

Here is the call graph for this function:

void joshua.decoder.JoshuaConfiguration.readConfigFile ( String configFile ) throws IOException

PHRASE-BASED PARAMETERS

Here is the call graph for this function:

Here is the caller graph for this function:

void joshua.decoder.JoshuaConfiguration.reset ( )

This method resets the state of JoshuaConfiguration back to the state after initialization. This is useful when for example making different calls to the decoder within the same java program, which otherwise leads to potential errors due to inconsistent state as a result of loading the configuration multiple times without resetting etc.

This leads to the insight that in fact it may be an even better idea to refactor the code and make JoshuaConfiguration an object that is is created and passed as an argument, rather than a shared static object. This is just a suggestion for the next step.

Here is the call graph for this function:

void joshua.decoder.JoshuaConfiguration.sanityCheck ( )

Checks for invalid variable configurations

Here is the caller graph for this function:

Member Data Documentation

boolean joshua.decoder.JoshuaConfiguration.amortized_sorting = true

boolean joshua.decoder.JoshuaConfiguration.constrain_parse = false

String joshua.decoder.JoshuaConfiguration.default_non_terminal = FormatUtils.markup("X")

ArrayList<String> joshua.decoder.JoshuaConfiguration.features = new ArrayList<String>()

boolean joshua.decoder.JoshuaConfiguration.filter_grammar = false

String joshua.decoder.JoshuaConfiguration.fragmentMapFile = null

boolean joshua.decoder.JoshuaConfiguration.fuzzy_matching = false

String joshua.decoder.JoshuaConfiguration.goal_symbol = FormatUtils.markup("GOAL")

String joshua.decoder.JoshuaConfiguration.hypergraphFilePattern = ""

boolean joshua.decoder.JoshuaConfiguration.include_align_index = false

String joshua.decoder.JoshuaConfiguration.input_file = null

boolean joshua.decoder.JoshuaConfiguration.lattice_decoding = false

final Logger joshua.decoder.JoshuaConfiguration.logger = Logger.getLogger(JoshuaConfiguration.class.getName()) [private]

boolean joshua.decoder.JoshuaConfiguration.mark_oovs = false

int joshua.decoder.JoshuaConfiguration.maxlen = 200

boolean joshua.decoder.JoshuaConfiguration.moses = false

String joshua.decoder.JoshuaConfiguration.n_best_file = null

int joshua.decoder.JoshuaConfiguration.num_parallel_decoders = 1

int joshua.decoder.JoshuaConfiguration.num_translation_options = 20

ArrayList<OOVItem> joshua.decoder.JoshuaConfiguration.oovList = null

String joshua.decoder.JoshuaConfiguration.outputFormat = "%i ||| %s ||| %f ||| %c"

This string describes the format of each line of output from the decoder (i.e., the translations). The string can include arbitrary text and also variables. The following variables are available:


i the 0-indexed sentence number
e the source string s the translated sentence
S the translated sentence with some basic capitalization and denormalization
t the synchronous derivation
f the list of feature values (as name=value pairs)
c the model cost
w the weight vector
a the alignments between source and target words (currently unimplemented)
d a verbose, many-line version of the derivation