Google is now introducing its Meena chatbot as one that "can conduct conversations that are more sensible and specific than existing state-of-the-art chatbots." One of the most distinguishing characteristics of Meena is the bot's ability to give context to a conversation.
For this project, Google decided it was necessary to created a new human evaluation metric, the Sensibleness and Specificity Average (SSA), to capture important attributes of natural conversations. Data from thousands of conversations were collected with other chatbots like Mitsuku, Cleverbot, XiaoIce and DialoGPT, and were compared to humans. While Meena scored an incredible 79%, humans come in at about 86% on the SSA score, which makes Meena very human-like, helping to close the gap between human and chatbot performance.
Image Credit: Google