Fixed issues with reading in and setting translations for groups.

Groups can appear in multiple places in group files (in which case translations get appended).
Reversed translations and explicit target->source translations no longer conflict when they overlap.

Author: megasser

gyez_logo.odg graphics/gyez_logo.odg

@@ -46,10 +46,10 @@ def load(reverse=False, bidir=False):
     s, t = iwqet.Language.load_trans(src, targ, bidir=bidir)
     return s, t
 
-def tra(oracion, reverse=False, html=False, user=None, choose=False, verbosity=0):
-    return sent(oracion, user=user, max_sols=2, translate=True,
-               connect=True, generate=True, html=html, choose=choose,
-               verbosity=verbosity)
+def tra(sentence, reverse=False, html=False, user=None, choose=False, verbosity=0):
+    return sent(sentence, user=user, max_sols=2, translate=True, reverse=reverse,
+                connect=True, generate=True, html=html, choose=choose,
+                verbosity=verbosity)
 
 # test sentences, including named entities
 T1 = "በቀለም ደረሰ።"
@@ -74,8 +74,9 @@ def document(text, process=True):
 def sent(text, reverse=False, user=None, max_sols=3, translate=True,
         connect=True, generate=False, html=False, choose=False, verbosity=0):
     return iwqet.አረፍተነገር(text, user=user, max_sols=max_sols, translate=translate,
-                        connect=connect, generate=generate, html=html, choose=choose,
-                        verbosity=verbosity)
+                         reverse=reverse,
+                         connect=connect, generate=generate, html=html, choose=choose,
+                         verbosity=verbosity)
 
 def sentence(sentence, ambig=False, solve=True, user=None, segment=True,
              max_sols=1, verbosity=0):

gyez_logo.png graphics/gyez_logo.png

@@ -173,18 +173,22 @@ def gui_trans(gui, session=None, choose=False, return_string=False,
                    verbosity=verbosity, terse=terse)
 
 def አረፍተነገር(text='', src=None, targ=None, user=None, session=None,
-            sentence=None,
-            max_sols=2, translate=True, connect=True, generate=True,
-            html=False, choose=False,
+            sentence=None, reverse=False,
+            max_sols=3, translate=True, connect=True, generate=True,
+            html=False, choose=False, finalize=False,
             return_string=False, verbosity=0, terse=False):
     """
     Analyze and possibly also translate a sentence from Amharic to Chaha.
     """
     if not src and not targ:
+        if reverse:
+            s, t = 'sgw','amh'
+        else:
+            s, t = 'amh', 'sgw'
 #        src = iwqet.Language.languages.get('amh')
 #        targ = iwqet.Language.languages.get('sgw')
 #        if not src:
-        src, targ = Language.load_trans('amh', 'sgw', bidir=False)
+        src, targ = Language.load_trans(s, t, bidir=False)
     if not session:
         session = make_session(src, targ, user, create_memory=True)
     s = Sentence.solve_sentence(src, targ, text=text, session=session,
@@ -195,6 +199,7 @@ def አረፍተነገር(text='', src=None, targ=None, user=None, session=None,
     segmentations = s.get_all_segmentations(translate=translate,
                                             generate=generate,
                                             agree_dflt=False, choose=choose,
+                                            finalize=finalize,
                                             connect=connect, html=html,
                                             terse=terse)
     print("SEGMENTATIONS: {}".format(segmentations))

logo.odg graphics/logo.odg

@@ -570,6 +570,11 @@ def match_segments(self, segments, startindex, seglimit=8, verbosity=1):
         return Match(self, matches)
 
     def reverse_trans(self, sgroup, sfeats):
+        if not self.trans:
+            self.trans = []
+        if any([sgroup == sg for sg, ft in self.trans]):
+#            print("*** {} already has trans {}".format(self, sgroup))
+            return
         rev_feats = sgroup.reverse_feats(self, sfeats)
         self.trans.append((sgroup, rev_feats))
 
@@ -1144,8 +1149,17 @@ def from_string(string, language, trans_strings=None, target=None,
                   trans_strings=tstrings, cat=cat, comment=comment,
                   intervening=intervening)
         if target and not trans:
-            # Add translation to source group
-            g.trans = tgroups or []
+            if not g.trans:
+                g.trans = []
+            gt = g.trans
+            if tgroups:
+                if gt:
+                    existing_tg = [t[0] for t in gt]
+                    tgroups = [tg for tg in tgroups if tg[0] not in existing_tg]
+                gt.extend(tgroups)
+#        if target and not trans:
+#            # Add translation to source group
+#            g.trans = tgroups or []
         if not existing_group:
             # Add group to its language in the appropriate POS groups
             language.add_group(g, posindex=posindex, cat=cat)

mit'mit'a_logo.ai graphics/mit'mit'a_logo.ai

@@ -1492,7 +1492,7 @@ def load_morpho(self, generate=False, analyze=True, segment=False,
         Load words and FSTs for morphological analysis and/or generation.
         """
         if verbose:
-            print('Loading morphological data for {} {}'.format(self.name, "(gen)" if generate else "(anal)"))
+            print('Loading morphological data for {} (anal:{}, gen:{})'.format(self.name, analyze, generate))
         # Load pre-analyzed words
         self.set_analyzed()
         if analyze:
@@ -2557,7 +2557,7 @@ def load_trans(source, target, bidir=False):
         targlang = Language.languages.get(target)
         src_loaded = False
         targ_loaded = False
-        srcuse = SOURCE
+        srcuse = BIDIR if bidir else SOURCE
         targuse = BIDIR if bidir else TARGET
         if srclang:
 #            print("Srclang use: {}".format(srclang.use))

mit'mit'a_logo.png graphics/mit'mit'a_logo.png

@@ -2117,28 +2117,29 @@ def create_segmentation(self, dstore=None, verbosity=0, terse=False):
                 gnode = self.gnodes[gn]
                 gn_group = gnode.ginst.index
                 if gn_group not in tree_attribs:
-                    tree_attribs[gn_group] = [[], []]
-                tree_attribs[gn_group][0].append(snindex)
-            if len(sg) == 2:
-                # Record group merger when an snode is associated with two gnodes
-                gn0, gn1 = self.gnodes[sg[0]], self.gnodes[sg[1]]
-                group0, group1 = gn0.ginst.index, gn1.ginst.index
-                if gn0.cat:
-                    # Group for gnode0 is merged with group for gnode1
-                    tree_attribs[group0][1].append(group1)
-                else:
-                    tree_attribs[group1][1].append(group0)
-        for gindex, sn in tree_attribs.items():
-            # First store the group's own tree as a set of sn indices and
-            # the third element of sn
-            sn.append(set(sn[0]))
-            # Next check for mergers
-            Sentence.update_tree(tree_attribs, gindex, sn[2])
-        # Convert the dict to a list and sort by group indices
-        trees = list(tree_attribs.items())
-        trees.sort(key=lambda x: x[0])
-        # Just keep the snode indices in each tree
-        trees = [x[1][2] for x in trees]
+                    tree_attribs[gn_group] = [] # [[]] # [[], []]
+                tree_attribs[gn_group].append(snindex)
+            # if len(sg) == 2:
+            #     # Record group merger when an snode is associated with two gnodes
+            #     gn0, gn1 = self.gnodes[sg[0]], self.gnodes[sg[1]]
+            #     group0, group1 = gn0.ginst.index, gn1.ginst.index
+            #     if gn0.cat:
+            #         # Group for gnode0 is merged with group for gnode1
+            #         tree_attribs[group0][1].append(group1)
+            #     else:
+            #         tree_attribs[group1][1].append(group0)
+        # for gindex, sn in tree_attribs.items():
+        #     # First store the group's own tree as a set of sn indices and
+        #     # the third element of sn
+        #     sn.append(set(sn[0]))
+        #     # Next check for mergers
+        #     Sentence.update_tree(tree_attribs, gindex, sn[2])
+        # # Convert the dict to a list and sort by group indices
+        # trees = list(tree_attribs.items())
+        # trees.sort(key=lambda x: x[0])
+        # # Just keep the snode indices in each tree
+        # trees = [x[1][2] for x in trees]
+        trees = [set(x) for x in tree_attribs.values()]
         # Get the indices of the GNodes for each SNode
         segmentation = Segmentation(self, ginsts, s2gnodes, len(self.segmentations),
                                     trees=trees, dstore=dstore, session=self.session,
@@ -2148,7 +2149,8 @@ def create_segmentation(self, dstore=None, verbosity=0, terse=False):
         return segmentation
 
     def get_all_segmentations(self, translate=True, generate=True,
-                              connect=False, html=False, agree_dflt=True, choose=False,
+                              connect=False, html=False, agree_dflt=True,
+                              choose=False, finalize=False,
                               verbosity=0, terse=False):
         """After a sentence has been translated and segmented, collect all the
         segmentations, including those resulting from altsyn sentences."""
@@ -2178,7 +2180,8 @@ def get_all_segmentations(self, translate=True, generate=True,
                         # Realize target morphology
                         segmentation.generate()
                         # Generate the final translation strings and HTML for the GUI
-                        segmentation.finalize_segments(html=html, agree_dflt=agree_dflt, choose=choose)
+                        segmentation.finalize_segments(html=html, agree_dflt=agree_dflt,
+                                                       choose=choose, finalize=finalize)
                 if generate and choose:
                     # Set the final output sentence string.
                     final = ' '.join([seg.final for seg in segmentation.segments])
@@ -2695,7 +2698,7 @@ def get_segs(self, terse=False):
     #######
 
     def finalize_segments(self, html=True, user_input=None, agree_dflt=True,
-                          choose=False, verbosity=0):
+                          choose=False, finalize=False, verbosity=0):
         """Set the final strings and morphology for each segment in this
         segmentation and the HTML too if html is True."""
         for i, segment in enumerate(self.segments):
@@ -2711,7 +2714,7 @@ def finalize_segments(self, html=True, user_input=None, agree_dflt=True,
                     first = False
         self.do_seg_feat_agreement(user_input=user_input, agree_dflt=agree_dflt,
                                    verbosity=verbosity)
-        if choose:
+        if choose or finalize:
             self.choose_final(html=html, verbosity=verbosity)
 
     def choose_final(self, html=True, verbosity=0):