Acobot LLC, the startup that invented self-learning AI capable of building conversational AI autonomously, released a new natural language understanding (NLU) technology that boosts the answer accuracy of it conversational AI for customer services.

With the support of this new NLU engine, the conversational AI, named Aco by the company brand, can identify up to 95% of questions covered by the given knowledge base and give precise answers to customers, according to the company’s evaluation.

The evaluation is focused on the AI’s ability in matching customer questions with those in the knowledge base. Questions not covered by the knowledge base are excluded because enhancing the coverage of a knowledge base is the task of self-learning AI, whereas the purpose of NLU engine is to match the questions with the existing knowledge entries.

“Answer accuracy is critical to customer experience,” pointed out by Vic Duan, founder and CEO of Acobot. “Many chatbots available on market can only answer 20% or even less of business questions correctly while our observation reveals an accuracy below 60% simply irritates customers and drives them away.”

The Acobot’s NLU engine works together with its self-learning AI builder and conversational AI. The self-learning AI builder firstly studies the documentation of product, extracts knowledge, and generates a knowledge base for the conversational AI. The coverage of such AI generated knowledge base typically exceeds one made by human. During the run time, the NLU engine helps the conversational AI identify the meaning of customer questions and match it with a knowledge entry in the knowledge base. Together they achieve an overall answer accuracy high enough for most customer service use cases.

“Today conversational AI or chatbot is not extensively used in customer service yet,” Duan added. “However the situation is changing rapidly. As we constantly innovate with NLU, the answer accuracy is being improved. The time is around the corner when every business, from enterprise to small business, hires virtual customer assistant.”

The NLU engine is based on a set of unique machine learning technologies that are developed by Acobot and optimized for natural language processing. Acobot confirms the NLU engine does not use any third party conversational AI framework, including IBM Watson, Google Dialogflow, or Microsoft Azure Conversation. In responding to the marketing hypes surrounding the conversational AI, the startup clarifies their artificial intelligence works not by inferring or interpreting the natural language as human does. It does its job by identifying and matching the various texts that convey the same or similar meanings.

An instance of conversational AI powered by the NLU engine is available at https://acobot.ai, where users may ask questions about Acobot and chat on a wide range of topics. After registration, users can get their own AI for customer service automation, with the same NLU engine included, achieving the same level of answer accuracy, without needing to train the AI manually.