Conducting Sentiment Analysis in real-time with your Chatbot visitors interface is a challenging proposition, yet can yield successful results as it allows you to better understand your visitor's needs and pivot the conversation accordingly. Sentiment Analysis has become more popular thanks to significant advances in deep learning, which has brought about cutting-edge algorithms to better extract the intent of a human's text communications. 

Chatbot Sentiment Analysis analyzes every message your visitors send during the conversation with your Chatbot and provides you with a sentiment to understand how your Chatbot performs and where it can be improved. Sentiment scores range from -10 to 10 and can be manually adjusted. The sentiment types are as follows: "Very positive", "Positive", "Neutral", "Negative", and "Very negative". 

Comm100’s Sentiment Analysis makes your Chatbots more powerful in understanding the emotion in visitor messages, helping you mitigate risks and delight customers by identifying and interacting with those who need your attention most.  

For example, if a visitor sentiment reaches Very negative, you can set a trigger to stop an Intent flow and transfer the chat to an agent. 

Benefits of the Chatbot Sentiment Analysis 

The following are the benefits of the Chatbot Sentiment Analysis: 

• Use the Chatbot Sentiment Analysis to make the Chatbot engage better with your visitors and improve their experience with your brand. 
• View the sentiment level for each visitor utterance and let the Chatbot drive the flow of conversation in tune with the visitor’s sentiment. 
• Improve your Chatbot’s Intent library by checking chat transcripts if you notice a Negative or Very negative sentiment. 

Sentiment types and recommended threshold adjustment for each sentiment type 

Sentiment type defines the visitor’s emotions among positive, neutral, and negative. 

The Chatbot can identify sentiment types and understand the sentiments in conversations. Sentiment analysis can help detect a range of emotions such as happiness, anger, sadness, positivity, etc.  

Comm100’s Chatbot Sentiment Analysis provides you with the following five sentiment types: 

• Very positive: The Very positive sentiment is achieved when the visitor is delighted with the Chatbot’s response. We recommend setting the threshold score between 7 and 10.
  
• Positive: The Positive sentiment score can be achieved when the visitor finds the Chatbot’s response to be helpful. You can set the threshold score between 2 and 7. 
  
• Neutral: At times, the conversation with the visitor might not contain any emotion or sentiment. The neural score can be linked to such conversations. You can consider setting the threshold score between -2 and 2 for these conversations.
   
• Negative: Negative sentiments happen when the visitor is unhappy or dissatisfied with the Chatbot’s service. You can set the threshold score between -7 and -2. 
  
• Very Negative: When the visitor is extremely unhappy and/or angry with the Chatbot, the Very negative sentiment occurs. We recommend you set the threshold score between -10 and -7.
   

Chatbot Sentiment score  

The Chatbot Sentiment Analysis returns a score for each visitor message and indicates the sentiment situation.  

Each sentiment score ranges from -10 to 10 and can be attached to a sentiment type. You can manually adjust the score from the Advanced drop-down option on the Edit Chatbot page.

Check the visitor’s sentiments in History 

When the visitor’s sentiment is Negative and Very Negative, it will be attached to 'sad face' emoticon and reflected in the History section. When the Chatbot's Sentiment Analysis is turned off, you will see an Off icon. 

You can navigate to Live Chat > History > Chats in the Control Panel to check the Sentiment Analysis for any Chatbot. You can also click the Details option found under Operations to review the transcripts for visitor messages with red and orange 'sad face' emoticons.