Source Code for Module jazzparser.settings

  1  """Global settings for the Jazz Parser. 
  2   
  3  This module is imported by other modules to access global settings. 
  4  """ 
  5   
  6  import os 
  7  PROJECT_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")) 
  8  # Where all the source code lives (absolute path) 
  9  SOURCE_DIR = os.path.join(PROJECT_ROOT, "src") 
 10  # Where the statistical model data lives 
 11  MODEL_DATA_DIR = os.path.join(PROJECT_ROOT, "etc", "data") 
 12  # Model data for PCFG models 
 13  PCFG_MODEL_DATA_DIR = os.path.join(MODEL_DATA_DIR, "pcfg") 
 14  # A slightly different location for backoff model data 
 15  BACKOFF_MODEL_DATA_DIR = os.path.join(PROJECT_ROOT, "etc", "backoff_data") 
 16  # Where the XML grammar definitions live 
 17  GRAMMAR_DATA_DIR = os.path.join(PROJECT_ROOT, "grammars") 
 18  # Where config files and local data are stored 
 19  LOCAL_DATA_DIR = os.path.join(PROJECT_ROOT, "etc", "local") 
 20  # Generic temporary directory for any purpose 
 21  TEMP_DIR = os.path.join(PROJECT_ROOT, "etc", "tmp") 
 22  # Where external corpora are stored within the project 
 23  CORPORA_DIR = os.path.join(PROJECT_ROOT, "input", "corpora") 
 24  # Where tonal space analysis sets live 
 25  ANALYSIS_DATA_DIR = os.path.join(PROJECT_ROOT, "etc", "analyses") 
 26  # Where data for unit tests is stored 
 27  TEST_DATA_DIR = os.path.join(PROJECT_ROOT, "etc", "test") 
 28  # Where shell states should be stored 
 29  SHELL_STATE_DIR = os.path.join(PROJECT_ROOT, "etc", "shell_state") 
 30  # Where releases are built 
 31  RELEASE_BUILD_DIR = os.path.join(TEMP_DIR, "release") 
 32   
 33  # The version ID of the software currently 
 34  from . import __version__ 
 35  CURRENT_VERSION = __version__ 
 36   
 37  # Sets the name of the default grammar that will be used if none other is 
 38  #  specified on the command line. 
 39  DEFAULT_GRAMMAR = "jazz3.0" 
 40   
 41  # The type of supertagger that should be used by default 
 42  DEFAULT_SUPERTAGGER = 'full' 
 43   
 44  # The grammar infrastructure to use by default 
 45  # This doesn't usually make a difference, since it's specified by the grammar (in the XML) 
 46  # If it's omitted from the XML this will be used, and some tools will assume  
 47  #  the default formalism if no other information is available 
 48  DEFAULT_FORMALISM = 'music_halfspan' 
 49   
 50  # The parser algorithm module to use by default 
 51  DEFAULT_PARSER = 'cky' 
 52   
 53  # Substrings that are stripped from input sequences given in plain text format 
 54  IGNORED_INPUT_STRINGS = [ "|", ] 
 55   
 56  # Output warnings during derivation if there are free variables in the semantics 
 57  WARN_ABOUT_FREE_VARS = False 
 58   
 59  # File to save the interactive shell's history to 
 60  SHELL_HISTORY_FILE = os.path.join(LOCAL_DATA_DIR, "shell-history") 
 61  # File to save the input prompt's history to 
 62  INPUT_PROMPT_HISTORY_FILE = os.path.join(LOCAL_DATA_DIR, "input-history") 
 63  # File to save the tagger test input loop's history to 
 64  TAG_PROMPT_HISTORY_FILE = os.path.join(LOCAL_DATA_DIR, "tag-history") 
 65  # Ngram query script history 
 66  NGRAM_QUERY_HISTORY_FILE = os.path.join(LOCAL_DATA_DIR, "ngram-history") 
 67   
 68 -class OPTIONS: 
 69      ## These are defaults that may be overridden by cmd line opts 
 70      # Global flag to decide whether to output times on all semantic objects or just TDs 
 71      OUTPUT_ALL_TIMES = False 
 72      # Global flag to toggle between Latex and text output 
 73      OUTPUT_LATEX = False 
 74      # Output options that each formalism defines, indexed by formalism name 
 75      OUTPUT = {} 
 76   
 77 -class CANDC: 
 78      BASE_PATH = os.path.join(PROJECT_ROOT, "lib", "candc") 
 79      MODELS_PATH = os.path.join(PROJECT_ROOT, "etc", "candc_data") 
 80      DEFAULT_TRAINING_PARAMS = { 
 81          'super-category_cutoff' : '0', 
 82          'super-cutoff_default' : '0', 
 83          'super-cutoff_words' : '0', 
 84          #'super-rare_cutoff' : '20', # This makes no difference 
 85          'super-tagdict_min' : '0', 
 86          'super-tagdict_ratio' : '10000', 
 87          'super-beam_width' : '20', 
 88          'super-beam_ratio' : '0.0', 
 89          #'super-model_sigma' : '0.9', 
 90          'sigma' : '0.85', 
 91      } 
 92      LOG_DIRECTORY = os.path.join(PROJECT_ROOT, "etc", "log", "candc") 
 93       
 94 -class PCFG_PARSER: 
 95      ### Default settings ### 
 96      # These can be overridden by module options 
 97      ######################## 
 98      # Module option: threshold 
 99      # Maximum ratio between the highest probability in a cell and the lowest 
100      #  before the lower end gets chucked out 
101      DEFAULT_THRESHOLD = 0.01 
102      # Module option: maxarc 
103      # Maximum number of signs in a cell before the lower probability ones  
104      #  get chucked out 
105      DEFAULT_MAX_ARC_SIZE = 20 
106      ### Default training options ### 
107      # These too can be overridden by module options 
108      ######################## 
109      # Module option: cat_bins 
110      # Number of possible categories to reserve mass for when smoothing 
111      # This number is the number of theoretically possible categories in halfspan 
112      # TODO: This is an overestimate, it should be 1296 
113      # TODO: Try the model with 1296 instead 
114      DEFAULT_CAT_BINS = 4752 
115   
116 -class TEST: 
117      """ 
118      Settings and common constants for unit tests (see L{jptests}). 
119       
120      """ 
121      SEQUENCE_DATA = os.path.join(PROJECT_ROOT, "input", "fullseqs") 
122