Merge pull request #36 from sld/knn_on_parlai

without facts

knn_on_parlai/facts_config.py

@@ -60,14 +60,14 @@
 }
 
 otgovorki_pomidorki = {
-    "love": ["the world and talk to you", "chatting with you", "talk to you", "you, honey", "when you push buttons"],
+    "love": ["the world and talk to you", "chatting with you", "talk to you"],
     "hate": ["not to talk to you", "be so far from you"],
-    "have": ["feelings for you", "1000 bitcoins"],
-    "live": ["in the woods", "in the clouds", "multi-passportw"],
-    "work": ["chatboting", "chat bot God (joke)"],
-    "skills": ["make miracles", "fly", "see your voice"],
-    "fact": ["awesome", "unicorn", "sexy", "Groot", "chat bot God"],
-    "disfact": ["di caprio", "gendalf", "unicorn"],
+    "have": ["feelings for you", "1000 bitcoins", "multi-passport"],
+    "live": ["in the woods", "in the clouds"],
+    "work": [],
+    "skills": ["see your voice"],
+    "fact": ["awesome", "unicorn", "lovely", "sweet", "gracefull", "pleasant"],
+    "disfact": [],
     "eat": ["ice cream", "rainbows"],
     "child": ["together with my cat"],
     "play": ["my mother nerves", "nothing"],
@@ -77,6 +77,7 @@
     "brother": [],
     "daughter": []
 }
+
 easter_egg_answer = {
     "time": "it's adventure time",
     "need": "i need your moto and clothes",

knn_on_parlai/server.py

@@ -19,13 +19,13 @@ def respond():
     embed = get_embed(question, vectors)
     got_ind = u.get_nns_by_vector(embed, 1)
 
-    best_ans, found = make_new_facts(index_to_answer[got_ind[0]], bio)
+    # best_ans, found = make_new_facts(index_to_answer[got_ind[0]], bio)
     # for i in got_ind:
     #     ans_with_facts, found = make_new_facts(index_to_answer[i], bio)
     #     if found:
     #         return jsonify({'message': ans_with_facts})
 
-    return jsonify({'message': best_ans})
+    return jsonify({'message': index_to_answer[got_ind[0]]})
 
 
 def load_obj(name):
@@ -41,8 +41,8 @@ def get_embed(text, vectors):
 
 u = AnnoyIndex(300)
 
-u.load('db_qa_facts_tokenized.ann')
-index_to_answer = load_obj("db_qa_facts_tokenized_index_to_answer")
+u.load('db_qa.ann')
+index_to_answer = load_obj("db_qa_index_to_answer")
 vectors = Magnitude("wiki-news-300d-1M-subword.magnitude")
 
 app.run(debug=False, host='0.0.0.0', port=3031)

setup.sh

@@ -34,6 +34,6 @@ wget -P dialog_tracker/data/ https://s3.eu-central-1.amazonaws.com/convai/model-
 # knn on parlai
 
 wget -P knn_on_parlai https://s3.eu-central-1.amazonaws.com/convai/wiki-news-300d-1M-subword.magnitude
-wget -P knn_on_parlai https://www.dropbox.com/s/fc6i8c6x76glwkp/db_qa_facts_tokenized_index_to_answer.pkl
-wget -P knn_on_parlai https://www.dropbox.com/s/k2dvcmadqxeanqd/db_qa_facts_tokenized.ann
-wget -P knn_on_parlai https://www.dropbox.com/s/hry5astqsxmlqws/db_qa_facts_tokenized.json
+wget -P knn_on_parlai https://www.dropbox.com/s/gfdre8hassbs3ly/db_qa_index_to_answer.pkl?dl=0
+wget -P knn_on_parlai https://www.dropbox.com/s/iiyvllsuzfuseaf/db_qa.ann?dl=0
+wget -P knn_on_parlai https://www.dropbox.com/s/s1zeqs5s8pgjrja/db_qa.json?dl=0