diff --git a/src/service.py b/src/service.py
index 920cf54..ff6bdff 100644
--- a/src/service.py
+++ b/src/service.py
@@ -53,7 +53,11 @@ def smart_filter(tweets, smart_project_id, threshold_score):
     """
     # apply smart filter
     request_data = [{'text': t['text']} for t in tweets]  # leave only text field to speed up data transfer
-    predictions = heartex.api.run_predict(token=token, project=smart_project_id, data=request_data).json()['results']
+    predictions = heartex.api.run_predict(token=token, project=smart_project_id, data=request_data).json()
+    if 'results' in predictions:
+        predictions = predictions['results']
+    else:
+        raise Exception('ML backend returns incorrect result: ' + str(predictions))
 
     # take only relevant tweets
     new_tweets = []
@@ -62,7 +66,8 @@ def smart_filter(tweets, smart_project_id, threshold_score):
             new_tweets.append(tweets[i])
 
     log.info('Smart filter by project id ' + str(smart_project_id) + ' processed '
-             'with input len = ' + str(len(tweets)) + ' and output len = ' + str(len(new_tweets)))
+                                                                     'with input len = ' + str(
+        len(tweets)) + ' and output len = ' + str(len(new_tweets)))
     return new_tweets
 
 
@@ -137,7 +142,12 @@ def heartex_build_plot(data, threshold_score=0.5, period='1D'):
             request_data.append({'text': reply})
 
     # heartex predict
-    predictions = heartex.api.run_predict(token=token, project=sentiment_project_id, data=request_data).json()['results']
+    predictions = heartex.api.run_predict(token=token, project=sentiment_project_id, data=request_data).json()
+    if 'results' in predictions:
+        predictions = predictions['results']
+    else:
+        raise Exception('ML backend returns incorrect result: ' + str(predictions))
+
     if not isinstance(predictions, list):
         log.warning('No predictions by ML backend returned: ' + str(predictions))
         raise Exception('No predictions by ML backend returned')
@@ -148,6 +158,7 @@ def heartex_build_plot(data, threshold_score=0.5, period='1D'):
         tweet['predictions'] = []
         for _ in tweet['replies']:
             tweet['predictions'].append(predictions[count])
+            tweet['replies_visible'] = False
             count += 1
 
     # collect score values (positives & negatives)
@@ -231,7 +242,7 @@ def api_build_sentiment():
 
     output = heartex_build_plot(tweets)
     log.info('Heartex prediction completed')
-    
+
     return answer(200, 'ok', output)
 
 
diff --git a/src/templates/index.html b/src/templates/index.html
index c8eb28a..df0eeb5 100644
--- a/src/templates/index.html
+++ b/src/templates/index.html
@@ -75,6 +75,16 @@
             color: #ccc !important;
             background: #28206c !important;
             margin-right: 0.2em;
+            cursor: pointer;
+        }
+
+        .news-replies {
+            display: block;
+            font-size: 70%;
+        }
+
+        .news-replies ul {
+            list-style-type: none;
         }
 
         .legend {
@@ -244,12 +254,24 @@
                 <div class="ui tiny label news-time">
                     {{ (new Date(item['created_at']*1000)).format('yyyy.mm.dd HH:MM') }}
                 </div>
-                <div class="ui tiny label news-sentiment">
+                <div class="ui tiny label news-sentiment" @click="show_replies(item)">
                     {{ item['positives'] }} <i class="ui icon thumbs up outline"></i>
                     {{ item['negatives'] }} <i class="ui icon thumbs down outline"
                                                style="margin-right:0 !important;"></i>
                 </div>
                 {{ item['text'] }}
+
+                <!-- Replies -->
+                <div  class="news-replies" v-if="item['replies_visible']">
+                    <ul>
+                        <li v-for="(p, i) in item['predictions']">
+                            <i v-if="item.predictions[i].result[0].value.choices[0] === 'Positive'" class="ui icon thumbs up outline"></i>
+                            <i v-if="item.predictions[i].result[0].value.choices[0] === 'Negative'" class="ui icon thumbs down outline"></i>
+                            <i v-if="item.predictions[i].result[0].value.choices[0] === 'Neutral'" class="ui icon"></i>
+                            {{ item['replies'][i] }}
+                        </li>
+                    </ul>
+                </div>
             </div>
         </ul>
     </div>
@@ -319,7 +341,7 @@
                 chart_instance: null,
                 news: [],
                 timer: null,
-                query: "",
+                query: "Bitcoin",
                 start_date: "",
                 current_request_id: 0,
                 smart_filters: []
@@ -331,6 +353,16 @@
                 Object.assign(this.$data, this.$options.data.apply(this));
             },
 
+            show_replies(item) {
+                if (item.hasOwnProperty('replies_visible')) {
+                    item['replies_visible'] = !item['replies_visible'];
+                }
+                else {
+                    item['replies_visible'] = true;
+                }
+                console.log('show')
+            },
+
             use_ml_query(model_name) {
                 this.query = 'model:"' + model_name + '"'
             },