Merge pull request #25 from philip-schrodt/master

added null_verbs/actors options to config.ini and updated documentati…

Petrarch2.tex

@@ -423,8 +423,9 @@ \subsection{\texttt{--nullverbs} and \texttt{--nullactors}: Identifying verb and
 These are complementary command line options intended for use in dictionary development:
 
 \begin{itemize}
-\item \texttt{--nullverbs} (or \texttt{-nv}) produces a file of verb phrases which have a valid source and target but which are not in the dictionary 
-\item  \texttt{--nullactors} (or \texttt{-na}) produces a file of noun phrases which are associated with a codable verb phrase but which are not in the dictionary 
+\item \texttt{--nullverbs} (or \texttt{-nv}) produces a file of verb phrases which have a valid source and target but which are not in the dictionary. This can also be invoked using the option \texttt{null\_verbs  = True} in the  \textit{config.ini} file.
+\item  \texttt{--nullactors} (or \texttt{-na}) produces a file of noun phrases which are associated with a codable verb phrase but which are not in the dictionary.  This can also be invoked using the option \texttt{null\_actors  = True} and setting \texttt{new\_actor\_length} \textgreater 0 in the  \textit{config.ini} file.
+
 \end{itemize}
 
 In other words,  \texttt{--nullverbs} shows unrecorded actions that are occurring between known actors, and  \texttt{--nullactors} shows unrecorded actors that are engaging in known behaviors. With both options, no events are generated and instead the output file is a series of JSON records which can be read as Python dictionaries: examples are given below.  
@@ -497,7 +498,7 @@ \subsubsection{\texttt{--nullactors}}
 \noindent \textbf{Implementation notes:}
 	\begin{itemize}
 		\setlength{\itemsep}{4pt}
-		\item The required integer following the command sets the \texttt{new\_actor\_length} \\parameter---the maximum number of words that a new actor can have---and overrides any settings in the \texttt{config.ini} file. Large values of this will give add a lot of extended noun phrases that probably aren't named-entities into the output; small values run the risk that a named entity preceded by several adjectives---e.g. \textit{``The beleaguered Japanese Prime Minister Shinzo Abe''}---will be missed. So, adjust this value based on your source texts, experience, and post-filtering programs.
+		\item If set in the command line, the required integer following the command sets the \texttt{new\_actor\_length} parameter---the maximum number of words that a new actor can have---and overrides any settings in the \texttt{config.ini} file (if set in the \texttt{config.ini} file, that value will be used). Large values of this will give add a lot of extended noun phrases that probably aren't named-entities into the output; small values run the risk that a named entity preceded by several adjectives---e.g. \textit{``The beleaguered Japanese Prime Minister Shinzo Abe''}---will be missed. So, adjust this value based on your source texts, experience, and post-filtering programs: values of 4 to 8 will usually work for typical English-language names.
 		\item If a source or target is in the dictionary, the phrase will be followed by the code in brackets \texttt{[\ldots]}
 		\item  \texttt{"evtcode"} is the CAMEO code for the verb phrase; \texttt{"evttext"} is the text which generated this code.
 		\item  The actors not in the dictionary are assigned temporary codes of the form \texttt{"*i*"} where $i$ is an integer. These will show up in agent code such as  \texttt{"*4*GOV"} (instead of \texttt{"- - -GOV"}) and in the source and target markers in the verb text, e.g. \texttt{"<*2*> ... contradicted by <*3*>"}. See Section \ref{sssec:writeeventtext} on the construction of event texts.

petrarch2/PETRglobals.py

@@ -54,10 +54,10 @@
 # NULL CODING OPTIONS
 NullVerbs  = False  # Only get verb phrases that are not in the dictionary but are associated with coded noun phrases
 NullActors = False  # Only get actor phrases that are not in the dictionary but associated with coded verb phrases
+NewActorLength = 0  # Maximum length for new actors extracted from noun phrases
 
 # CODING OPTIONS
 # Defaults are more or less equivalent to TABARI
-NewActorLength = 0  # Maximum length for new actors extracted from noun phrases
 RequireDyad = True  # Events require a non-null source and target
 StoponError = False  # Raise stop exception on errors rather than recovering
 

petrarch2/PETRreader.py

@@ -179,10 +179,12 @@ def get_config_boolean(optname):
                 raise
         print("new_actor_length =", PETRglobals.NewActorLength)
 
-        PETRglobals.StoponError = get_config_boolean('stop_on_error')
+        PETRglobals.StoponError    = get_config_boolean('stop_on_error')
         PETRglobals.WriteActorRoot = get_config_boolean('write_actor_root')
         PETRglobals.WriteActorText = get_config_boolean('write_actor_text')
         PETRglobals.WriteEventText = get_config_boolean('write_event_text')
+        PETRglobals.NullVerbs  = get_config_boolean('null_verbs')
+        PETRglobals.NullActors = get_config_boolean('null_actors')
 
         if parser.has_option(
                 'Options', 'require_dyad'):  # this one defaults to True

petrarch2/data/config/PETR_config.ini

@@ -59,15 +59,6 @@ code_by_sentence = True
 # CODING OPTIONS: 
 # Defaults are more or less equivalent to TABARI
 
-# new_actor_length: Maximum length for new actors extracted from noun phrases if no 
-#                   actor or agent generating a code is found. To disable and just 
-#                   use null codes "---", set to zero; this is the default. 
-#                   Setting this to a large number will extract anything found in a (NP
-#                   noun phrase, though usually true actors contain a small number of words 
-#                   This must be an integer.                       
-new_actor_length = 0
-
-
 # write_actor_root: If True, the event record will include the text of the actor root: 
 #                   The root is the text at the head of the actor synonym set in the 
 #                   dictionary. Default is False
@@ -81,6 +72,24 @@ write_actor_text = True
 #                   the verb phrase that was used to identify the event.  Default is False
 write_event_text = True
 
+# NULL CODING OPTIONS
+# null_verbs: If True, only get verb phrases that are not in the dictionary but are associated 
+#             with coded noun phrases
+null_verbs  = False
+
+# null_actors: If True, only get actor phrases that are not in the dictionary but associated with 
+#             coded verb phrases. This also requires new_actor_length to be set to a value > 0:
+#             typically a value of 4 to 8 will give good results.
+null_actors = False  
+
+# new_actor_length: Maximum length for new actors extracted from noun phrases if no 
+#                   actor or agent generating a code is found. To disable and just 
+#                   use null codes "---", set to zero; this is the default. 
+#                   Setting this to a large number will extract anything found in a (NP
+#                   noun phrase, though usually true actors contain a small number of words 
+#                   This must be an integer.                       
+new_actor_length = 0
+
 # require_dyad: Events require a non-null source and target: setting this false is likely
 #               to result in a very large number of nonsense events. As happened with the 
 #               infamous GDELT data set of 2013-2014. And certainly no one wants to see