Smartbot – Sentiment Evaluation of Real-time Speech (SERTS)

The human voice audio stream is richer in information than might initially appear. It provides two distinct sources of data, what is being said and how it is being said. The proposed Sentiment Evaluation of Real-time Speech (SERTS) software solution aims to capture this audio stream and apply two distinct analysis methodologies to it. One methodology focuses on what is being said, i.e. 1) the semantic content, while the other concentrates on how it is being said, that is 2) the sentiment or Emotion being expressed by the client. It’s envisaged that results from this novel dual approach could be combined to give the customer contact personnel or conversational chatbot a more intuitive and holistic overview of the client’s sentiment toward a discussed product or service. Combining the results of the semantic content analysis and the voice structure characteristics captured over the duration of the conversation provides a higher probability of accurately assessing the caller’s response to the product, service or the interaction & how a response should be structured based on not just what is said, i.e. the semantic content, but also how it is said, the voice structure or emotion expressed. The secret sauce in what we aim to achieve is combining  NLP + Context + Emotion + Self Learning in real-time to assist contact centres to move agents to higher value engagements, by enabling a chatbot to better handle routine tasks through a conversation interface such as Alexa, Google Home or Siri. New use cases are arising all the time, such as for example, a smart fridge that interprets how you feel and then suggests food to match those feelings or an angry customer can be detected from the beginning and can be routed to a well-trained agent who can also monitor in real-time how the conversation is going and adjust.