[similar] improve information extraction

entities.py

@@ -17,16 +17,27 @@ from cubes.similarity import register_similarity, reset_similarity, vsm
 def prefix_slice(value, start, end, *args):
     return u'%s: %s' % (value, u' '.join(args[start:end]))
 
+def rgx_replace(value, *args):
+    for idx, arg in enumerate(args):
+        value = value.replace(u'\\%i' % (idx+1), arg)
+    return value
+
+number2digit_rgx = re.compile('^\d\d$')
+number4digit_rgx = re.compile('^\d\d\d\d$')
+number5digit_rgx = re.compile('^\d\d\d\d\d$')
 number_rgx = re.compile('^\d+$')
 surface_rgx = re.compile(u'^\d+m[²2]$')
+annee_rgx = re.compile(u'^ann[ée]e$')
 
 TRANSLATE_TABLE = [
-    (('vlce',), ft.partial(vsm.const, 'valence')),
+    (('s-sol',), ft.partial(vsm.const, 'sous-sol')),
+    (('dt',), ft.partial(vsm.const, 'dont')),
     (('ch',), ft.partial(vsm.const, 'chambre')),
     (('chbs',), ft.partial(vsm.const, 'chambre')),
     (('chbrs',), ft.partial(vsm.const, 'chambre')),
     (('chambres',), ft.partial(vsm.const, 'chambre')),
     (('grd',), ft.partial(vsm.const, 'grand')),
+    (('gde',), ft.partial(vsm.const, 'grande')),
     (('ttes',), ft.partial(vsm.const, 'toutes')),
     (('niv',), ft.partial(vsm.const, 'niveau')),
     (('st',), ft.partial(vsm.const, 'saint')),
@@ -38,12 +49,16 @@ TRANSLATE_TABLE = [
     (('salle','à','manger'), ft.partial(vsm.const, 'salle à manger')),
     (('salle','bains'), ft.partial(vsm.const, 'salle de bain')),
     (('salle','eau'), ft.partial(vsm.const, 'salle d\'eau')),
-    (('jacuzi'), ft.partial(vsm.const, 'jacuzzi')),
+    (('jacuzi',), ft.partial(vsm.const, 'jacuzzi')),
+    ((number5digit_rgx,), ft.partial(vsm.prefix_join, 'postcode')),
     ((number_rgx, u'm²'), ft.partial(vsm.prefix_join, 'surface')),
     ((number_rgx, 'm2'), lambda x,y: u'surface: %s m²' % x),
     ((surface_rgx,), lambda x: u'surface: %s m²' % x[:-2]),
     ((number_rgx, u'pièces'), ft.partial(prefix_slice, u'rooms', 0, 1)),
-    (('drome',), ft.partial(vsm.const, u'drôme')),
+    ((annee_rgx, number4digit_rgx), ft.partial(rgx_replace, u'year-built: \\2')),
+    ((annee_rgx, number2digit_rgx), ft.partial(rgx_replace, u'year-built: 19\\2')),
+    (('drome',), ft.partial(vsm.const, u'location: http://fr.wikipedia.org/wiki/Dr%C3%B4me_%28d%C3%A9partement%29')),
+    (('drôme',), ft.partial(vsm.const, u'location: http://fr.wikipedia.org/wiki/Dr%C3%B4me_%28d%C3%A9partement%29')),
     (('gare', 'tgv'), ft.partial(vsm.prefix_join, 'location')),
     (('danton',), ft.partial(vsm.prefix_join, 'location')),
     (('grand','charran'), ft.partial(vsm.prefix_join, 'location')),
@@ -51,6 +66,7 @@ TRANSLATE_TABLE = [
     (('appartement',), ft.partial(vsm.prefix_join, 'property-type')),
     (('maison',), ft.partial(vsm.prefix_join, 'property-type')),
     (('villa',), ft.partial(vsm.const, 'property-type: maison')),
+    (('vlce',), ft.partial(vsm.const, 'location: http://dbpedia.org/resource/Valence%2C_Dr%C3%B4me')),
     (('valence',), ft.partial(vsm.const, 'location: http://dbpedia.org/resource/Valence%2C_Dr%C3%B4me')),
     (('26000', 'valence',), ft.partial(vsm.const, 'location: http://dbpedia.org/resource/Valence%2C_Dr%C3%B4me')),
     (('st', 'marcel'), ft.partial(vsm.const, 'location: http://dbpedia.org/resource/Saint-Marcel-l%C3%A8s-Valence')),