Converted Ekman2VAD to centroids

* Changed the way modules are imported -> we can now use dotted
  notation (e.g. senpy.plugins.conversion.centroids)
* Refactored ekman2vad's plugin -> generic centroids
* Added some basic tests

senpy/plugins/__init__.py

@@ -0,0 +1,103 @@
+from future import standard_library
+standard_library.install_aliases()
+
+import inspect
+import os.path
+import pickle
+import logging
+import tempfile
+import copy
+from .. import models
+
+logger = logging.getLogger(__name__)
+
+
+class SenpyPlugin(models.Plugin):
+    def __init__(self, info=None):
+        """
+        Provides a canonical name for plugins and serves as base for other
+        kinds of plugins.
+        """
+        if not info:
+            raise models.Error(message=("You need to provide configuration"
+                                        "information for the plugin."))
+        logger.debug("Initialising {}".format(info))
+        id = 'plugins/{}_{}'.format(info['name'], info['version'])
+        super(SenpyPlugin, self).__init__(id=id, **info)
+        self.is_activated = False
+
+    def get_folder(self):
+        return os.path.dirname(inspect.getfile(self.__class__))
+
+    def analyse(self, *args, **kwargs):
+        raise NotImplemented(
+            'Your method should implement either analyse or analyse_entry')
+
+    def analyse_entry(self, entry, parameters):
+        """ An implemented plugin should override this method.
+        This base method is here to adapt old style plugins which only
+        implement the *analyse* function.
+        Note that this method may yield an annotated entry or a list of
+        entries (e.g. in a tokenizer)
+        """
+        text = entry['text']
+        params = copy.copy(parameters)
+        params['input'] = text
+        results = self.analyse(**params)
+        for i in results.entries:
+            yield i
+
+    def activate(self):
+        pass
+
+    def deactivate(self):
+        pass
+
+
+class SentimentPlugin(models.SentimentPlugin, SenpyPlugin):
+    def __init__(self, info, *args, **kwargs):
+        super(SentimentPlugin, self).__init__(info, *args, **kwargs)
+        self.minPolarityValue = float(info.get("minPolarityValue", 0))
+        self.maxPolarityValue = float(info.get("maxPolarityValue", 1))
+
+
+class EmotionPlugin(models.EmotionPlugin, SenpyPlugin):
+    def __init__(self, info, *args, **kwargs):
+        super(EmotionPlugin, self).__init__(info, *args, **kwargs)
+        self.minEmotionValue = float(info.get("minEmotionValue", -1))
+        self.maxEmotionValue = float(info.get("maxEmotionValue", 1))
+
+
+class EmotionConversionPlugin(models.EmotionConversionPlugin, SenpyPlugin):
+    def __init__(self, info, *args, **kwargs):
+        super(EmotionConversionPlugin, self).__init__(info, *args, **kwargs)
+
+
+class ShelfMixin(object):
+    @property
+    def sh(self):
+        if not hasattr(self, '_sh') or self._sh is None:
+            self.__dict__['_sh'] = {}
+            if os.path.isfile(self.shelf_file):
+                self.__dict__['_sh'] = pickle.load(open(self.shelf_file, 'rb'))
+        return self._sh
+
+    @sh.deleter
+    def sh(self):
+        if os.path.isfile(self.shelf_file):
+            os.remove(self.shelf_file)
+            del self.__dict__['_sh']
+        self.save()
+
+    @property
+    def shelf_file(self):
+        if 'shelf_file' not in self or not self['shelf_file']:
+            self.shelf_file = os.path.join(tempfile.gettempdir(),
+                                           self.name + '.p')
+        return self['shelf_file']
+
+    def save(self):
+        logger.debug('saving pickle')
+        if hasattr(self, '_sh') and self._sh is not None:
+            with open(self.shelf_file, 'wb') as f:
+                pickle.dump(self._sh, f)

senpy/plugins/conversion/centroids.py

@@ -0,0 +1,52 @@
+from senpy.plugins import EmotionConversionPlugin
+from senpy.models import EmotionSet, Emotion, Error
+
+import logging
+logger = logging.getLogger(__name__)
+
+
+class CentroidConversion(EmotionConversionPlugin):
+
+    def _forward_conversion(self, original):
+        """Sum the VAD value of all categories found."""
+        res = Emotion()
+        for e in original.onyx__hasEmotion:
+            category = e.onyx__hasEmotionCategory
+            if category in self.centroids:
+                for dim, value in self.centroids[category].iteritems():
+                    try:
+                        res[dim] += value
+                    except Exception:
+                        res[dim] = value
+        return res
+
+    def _backwards_conversion(self, original):
+        """Find the closest category"""
+        dimensions = list(self.centroids.values())[0]
+
+        def distance(e1, e2):
+            return sum((e1[k] - e2.get(self.aliases[k], 0)) for k in dimensions)
+
+        emotion = ''
+        mindistance = 10000000000000000000000.0
+        for state in self.centroids:
+            d = distance(self.centroids[state], original)
+            if d < mindistance:
+                mindistance = d
+                emotion = state
+        result = Emotion(onyx__hasEmotionCategory=emotion)
+        return result
+
+    def convert(self, emotionSet, fromModel, toModel, params):
+
+        cf, ct = self.centroids_direction
+        logger.debug('{}\n{}\n{}\n{}'.format(emotionSet, fromModel, toModel, params))
+        e = EmotionSet()
+        if fromModel == cf:
+            e.onyx__hasEmotion.append(self._forward_conversion(emotionSet))
+        elif fromModel == ct:
+            for i in emotionSet.onyx__hasEmotion:
+                e.onyx__hasEmotion.append(self._backwards_conversion(i))
+        else:
+            raise Error('EMOTION MODEL NOT KNOWN')
+        yield e

senpy/plugins/conversion/emotion/ekman2vad.py

@@ -1,58 +0,0 @@
-from senpy.plugins import EmotionConversionPlugin
-from senpy.models import EmotionSet, Emotion, Error
-
-import logging
-logger = logging.getLogger(__name__)
-
-import math
-
-
-class WNA2VAD(EmotionConversionPlugin):
-
-    def _ekman_to_vad(self, ekmanSet):
-        """Sum the VAD value of all categories found."""
-        valence = 0
-        arousal = 0
-        dominance = 0
-        for e in ekmanSet.onyx__hasEmotion:
-            category = e.onyx__hasEmotionCategory
-            centroid = self.centroids[category]
-            valence += centroid['V']
-            arousal += centroid['A']
-            dominance += centroid['D']
-        e = Emotion({'emoml:valence': valence,
-                     'emoml:arousal': arousal,
-                     'emoml:potency': dominance})
-        return e
-
-    def _vad_to_ekman(self, VADEmotion):
-        """Find the closest category"""
-        V = VADEmotion['emoml:valence']
-        A = VADEmotion['emoml:arousal']
-        D = VADEmotion['emoml:potency']
-        emotion = ''
-        value = 10000000000000000000000.0
-        for state in self.centroids:
-            valence = V - self.centroids[state]['V']
-            arousal = A - self.centroids[state]['A']
-            dominance = D - self.centroids[state]['D']
-            new_value = math.sqrt((valence**2) +
-                                  (arousal**2) +
-                                  (dominance**2))
-            if new_value < value:
-                value = new_value
-                emotion = state
-        result = Emotion(onyx__hasEmotionCategory=emotion)
-        return result
-
-    def convert(self, emotionSet, fromModel, toModel, params):
-        logger.debug('{}\n{}\n{}\n{}'.format(emotionSet, fromModel, toModel, params))
-        e = EmotionSet()
-        if fromModel == 'emoml:big6':
-            e.onyx__hasEmotion.append(self._ekman_to_vad(emotionSet))
-        elif fromModel == 'emoml:fsre-dimensions':
-            for i in emotionSet.onyx__hasEmotion:
-                e.onyx__hasEmotion.append(self._vad_to_ekman(i))
-        else:
-            raise Error('EMOTION MODEL NOT KNOWN')
-        yield e

senpy/plugins/conversion/emotion/ekman2vad.senpy

@@ -1,6 +1,6 @@
 ---
 name: Ekman2VAD
-module: ekman2vad
+module: senpy.plugins.conversion.centroids
 description: Plugin to convert emotion sets from Ekman to VAD
 version: 0.1
 onyx:doesConversion:
@@ -29,6 +29,9 @@ centroids:
     A: 5.21
     D: 2.82
     V: 2.21
+centroids_direction:
+  - emoml:big6
+  - emoml:fsre-dimensions
 aliases:
   A: emoml:arousal
   V: emoml:valence