Add cache and NLP for weather

nl_processors/weather.py

@@ -0,0 +1,48 @@
+import re
+
+from nl_processor import as_processor
+
+_keywords = ('天气', '气温', '空气(质量)?', '温度', '多少度', '(风|雨|雪|冰雹|霜|雾|霾)')
+
+
+def _match_keywords(word):
+    for regex in _keywords:
+        if re.match(regex, word):
+            return True
+    return False
+
+
+@as_processor(keywords=_keywords)
+def _processor(sentence, segmentation):
+    possibility = 100
+    location_segs = list(filter(lambda x: x.flag == 'ns', segmentation))
+    if not location_segs:
+        return None
+
+    if len(location_segs) == 1:
+        # Just city name
+        city = location_segs[0].word.rstrip('市县区')
+    elif len(location_segs) == 2:
+        # Maybe has both province and city name
+        city = location_segs[0].word.rstrip('省') + location_segs[1].word.rstrip('市县区')
+    else:
+        # More than 3 location name, use the last one
+        city = location_segs[-1].word.rstrip('市县区')
+
+    for seg in location_segs:
+        segmentation.remove(seg)
+
+    for seg in segmentation:
+        # Scan over all segments and decrease possibility
+        if _match_keywords(seg.word):
+            continue
+
+        flag = seg.flag
+        score_dict = {'v': -10, 'l': -8, 'n': -5, 'p': -3, 't': +3, 'other': -1}
+        for k, v in score_dict.items():
+            if flag.startswith(k):
+                possibility += v
+                continue
+        possibility += score_dict['other']
+
+    return possibility, 'weather.weather', city, None