--- a
+++ b/development/paraphrase/nlu_augment.py
@@ -0,0 +1,93 @@
+import re
+from tqdm import tqdm
+from ruamel import yaml
+from paraphrase import wordtune
+from decouple import config
+
+
+class DataAugment():
+    def __init__(self, nlu_file, token, draft_id):
+        super(DataAugment, self).__init__()
+        self.token = token
+        self.paraphrase = wordtune(token, draft_id)
+
+        with open(nlu_file, 'r') as f:
+            self.nlu_dict = yaml.load(f, Loader=yaml.RoundTripLoader)
+
+    # Function to remove entity tags
+    def __tag_remove(self, inp):
+        entity_tag = ""
+        entity = ""
+
+        # Check if entity tags in input exist or not
+        if inp.find("[") != -1:
+            # entity and its tag
+            entity_tag = inp[inp.find("["):inp.find(")")+1]
+            # entity
+            entity = inp[inp.find("[")+1:inp.find("]")]
+            # remove tag from input. only entity exists in input
+            inp = inp.replace(entity_tag, entity)
+
+        return inp, entity_tag, entity
+
+    # augment nlu.yml datas with wordtune suggestions
+    def data_augment(self):
+        for intent in tqdm(self.nlu_dict['nlu']):
+            inps = intent['examples'].split('\n')[:-1]
+
+            # give each intents examples to wordtune
+            for inp in inps:
+                # strip dash and space from example
+                inp = inp[2:]
+                # remove tags from words
+                inp, entity_tag, entity = self.__tag_remove(inp)
+                # paraphrase examples
+                payload = self.paraphrase.payload_generator(inp)
+                headers = self.paraphrase.headers_generator()
+                response = self.paraphrase.requests(payload, headers)
+                # list of wordtune suggestions
+                suggestion_list = self.paraphrase.get_suggestion(response, inp)
+
+                for suggestion in suggestion_list:
+                    # add tags to each entity case insensitivly
+                    prog = re.compile(re.escape(entity), re.IGNORECASE)
+                    intent['examples'] += f'- {prog.sub(entity_tag, suggestion)}\n'
+
+    # remove all lines without entity tags
+    def data_clean(self):
+        for intent in self.nlu_dict['nlu']:
+            # exclude request_drug and request_lab intents. there is not entity tags in these intents examples
+            if intent['intent'] not in ['request_drug', 'request_lab']:
+                example_list = intent['examples'].split('\n')
+                intent['examples'] = ''.join(
+                    example+'\n' for example in example_list if "[" in example)
+
+    # write nlu datas to the disk
+    def write_data(self, out_name):
+        with open(out_name, 'w') as f:
+            yaml.dump(self.nlu_dict, f, Dumper=yaml.RoundTripDumper)
+
+
+def main():
+    # environmental variables must be in .env file in the same dir as this file
+    token = config('TOKEN')
+    draftId = config('DRAFTID')
+
+    nlu_file = 'nlu.yml'
+    output_file = 'nlu_augment.yml'
+    cleaned_output_file = 'nlu_cleaned.yml'
+
+    augment = DataAugment(nlu_file, token, draftId)
+    # rephrase input examples with wordtune
+    augment.data_augment()
+    # write augmented data to the disk
+    augment.write_data(output_file)
+    # remove examples without tags.
+    # manual remove preferred
+    augment.data_clean()
+    # write cleaned data to the disk
+    augment.write_data(cleaned_output_file)
+
+
+if __name__ == '__main__':
+    main()