speech

Author: Sean Trott

analyzer.sh

@@ -3,5 +3,7 @@ export ECG_FED=FED2
 export JYTHONPATH=build/compling.core.jar:src/main/nluas/language
 jython -J-Xmx2g -m analyzer ../ecg_grammars/research.prefs #&
 
+#jython -J-Xmx2g -i src/main/nluas/language/analyzer.py ../ecg_grammars/research.prefs #&
+
 #jython -i src/main/nluas/language/analyzer.py ../ecg_grammars/research.prefs #&
 #python3 src/main/nluas/language/user_agent.py AgentUI 

ntuples.sh

@@ -1,4 +1,5 @@
 #!/bin/bash
 
 #python3 src/main/nluas/language/ntuple_visualizer.py
-python3 src/main/nluas/language/ntuple_visualizer.py
+python3 src/main/nluas/language/ntuple_visualizer.py
+#python3 src/main/nluas/language/partial_parse.py

src/main/nluas/app/core_solver.py

@@ -181,7 +181,7 @@ def route_action(self, parameters, predicate):
             except AttributeError as e:
                 #traceback.print_exc()
                 message = "I cannot solve the '{}_{}' action".format(predicate,action)
-                pprint.pprint(parameters)
+                #pprint.pprint(parameters)
                 self.history.insert(0, (parameters, False))
                 self.identification_failure(message)
 

src/main/nluas/language/core_specializer.py

@@ -67,6 +67,10 @@ def __init__(self, analyzer):
     def set_spans(self, spans):
         """ Sets the current constructional spans to input spans. These are used for referent resolution. (TODO)."""
         self.spans = spans
+        self.np_spans = []
+        for k in self.spans:
+            if self.analyzer.issubtype("CONSTRUCTION", k[0], "NP"):
+                self.np_spans.append(k)
 
     def specialize_fragment(self, fs):
         """ Specializes a sentence fragment, e.g. 'the red one' or a non-discourse-utterance. """
@@ -249,7 +253,6 @@ def get_process_modifiers(self, eventProcess):
                         returned[prop] = value
                     elif pointer == "AdjunctModification":
                         modifier = mod.modifier
-                        print(modifier.type())
                         if modifier.type() == "Accompaniment":
                             returned["co-participant"] = self.get_objectDescriptor(modifier.co_participant)
                         elif modifier.type() == "Instrument":

src/main/nluas/language/specializer_utils.py

@@ -208,6 +208,7 @@ def merge_descriptors(self, old, new):
     """ Simple reference resolution gadget, meant to unify object pronouns with potential
     antecedents. """
     def resolve_referents(self, item, antecedents = None, actionary=None, pred=None):
+        #self.find_closest_antecedent([7,8])
         if antecedents is None:
             antecedents = self._stacked
         popper = list(antecedents)
@@ -225,6 +226,17 @@ def clean_referent(self, ref):
         ref['objectDescriptor'].pop('property', None)
         return ref
 
+    def find_closest_antecedent(self, target):
+        """ Takes in target span/word, ranks previous spans. """
+        ranks = []
+        for k in self.np_spans:
+            #if self.analyzer.issubtype("CONSTRUCTION", k[0], "Pronoun"):
+            span = k[2]
+            if span[0] < target[0] and span[1] < target[1]:
+                ranks.insert(0, k)
+        #print(ranks)
+
+
     def ordering(self, fs, ref):
         for index, value in fs.rootconstituent.__features__.items():
             if hasattr(value, "m") and value.m and value.m.type() == "RD":

src/main/nluas/language/user_agent.py

@@ -88,13 +88,16 @@ def initialize_specializer(self):
             quit()
 
     def match_spans(self, spans, sentence):
-        sentence = sentence.replace(".", " . ").replace(",", " , ").replace("!", " ! ").split(" ")
-        final = OrderedDict()
+        sentence = sentence.replace(".", " . ").replace(",", " , ").replace("?", " ? ").replace("!", " ! ").split()
+        #final = OrderedDict()
+        #word_spans = OrderedDict()
+        final = []
         for span in spans:
             lr = span['span']
-            final[span['type']] = (sentence[lr[0]:lr[1]], lr, span['id'])
+            final.append([span['type'], sentence[lr[0]:lr[1]], lr, span['id']])
         return final
 
+
     def process_input(self, msg):
         try:
             full_parse = self.analyzer.full_parse(msg)
@@ -111,7 +114,7 @@ def process_input(self, msg):
                     #json_ntuple = self.decoder.convert_to_JSON(ntuple)
                     return ntuple
                 except Exception as e:
-                    self.output_stream(self.name, e)
+                    #self.output_stream(self.name, e)
                     #traceback.print_exc()
                     index += 1
         except Exception as e:
@@ -125,7 +128,9 @@ def speech_callback(self, ntuple):
         """ Processes text from a SpeechAgent. """
         #print(ntuple)
         #ntuple = json.loads(ntuple)
-        new_ntuple = self.process_input(ntuple['text']) 
+        text = ntuple['text'].lower()
+        print("Got {}".format(text))
+        new_ntuple = self.process_input(text) 
         if new_ntuple and new_ntuple != "null" and "predicate_type" in new_ntuple:
             self.transport.send(self.solve_destination, new_ntuple)