Dazzling Tech
Conversational Bots: Lower Costs, and Better Service

As products like intelligent assistants, chatbots, and smart speakers grow in popularity, the field of natural language processing (NLP) has become a hot topic. In recent years, there have been a number of breakthroughs in this field. New training models are constantly being developed, and techniques like transfer leaning and multimodal sentiment analysis are becoming more and more widely adopted. In the enterprise market, there is a wide range of applications. The ability to truly understand human language will be one of the crowning achievements of artificial intelligence, but just what are some of the most typical applications of this amazing new technology. Just what does the future have in store for NLP?   

The History of Semantic Technology

There are four stages involved in human communication: hearing, understanding, response formulation, and the expression of that response. As such, NLP involves four different technologies: automatic speech recognition (ASR), natural language understanding (NLU), natural language generation (NLG), and text-to-speech (TTS).

Speech recognition and text-to-speech technologies, ASR and TTS, are relatively mature. Computers can recognize and produce human speech fairly well, and there are already a range of applications in use, but understanding and generating human speech has proved more challenging. NLU and NLG technologies still have a ways to go. Of course, in recent years, we have also seen NLU and NLG make significant progress in academia and in industry.

In the consumer field, more and more virtual personal assistants are showing up on smartphones; smart speakers are more common than ever, and the user experience is much better than three years ago. In the enterprise field, intelligent customer service and NLP text processing applications have helped many enterprises cut costs even while improving customer experience. And we expect more and more commercial applications will inevitably lead to more breakthroughs in NLP and NLG technologies.

Applications of Intelligent Semantics Technology

The HUAWEI CLOUD EI voice semantic team provides three types of services: NLP services, voice interaction, and man-machine interaction. NLP services include basic capabilities like dealing with word segmentation and text similarity, but also include sentiment analysis, text classification, keyword extraction, text summaries, language generation, knowledge mapping, machine translation, and more. Voice interaction includes voice recognition and synthesis, real-time streaming voice recognition, and voice extension. And finally, there is man-machine interaction, which covers Question Answering Bot, Intelligent Phonebot, and the like.

Intelligent Chatbots

At the 2018 I/O conference, Google demonstrated their intelligent chatbot, "Duplex", a chatbot that can make restaurant reservations and interact with humans at the other end of a phone call. This chatbot was so scarily human, many people found it unsettling and the demonstration quickly went viral. Some people have imagined that once this technology really gets rolling, we may find ourselves saying things like, "What a great idea. I’ll have my robot call your robot a little later to work out the details."

But in reality, for everyday consumers, this sort of science fiction may still be a ways off. For enterprises, however, the Conversational Bot service, developed by HUAWEI CLOUD in 2018, has already made multiple rounds of automated customer service calls in a variety of enterprise scenarios. Typical scenarios include automatic outbound call verification, customer satisfaction surveys, pre-screening interviews, checking up on an order's status, evaluating prospective clients, making reservations, or scheduling appointments. The HUAWEI CLOUD Conversational Bot Service uses a Huawei-developed intelligent chatbot engine to handle a range of custom designed man-machine conversation scenarios based on specific outbound calling scenarios and linguistic requirements.

This Conversational Bot Service can make 800 or more calls a day, and it never gets worn down by the repetition. It is significantly more efficient than a real person. And as the enterprise grows, it is not necessary to train new employees to handle the additional calls. Additional customer service chatbots can be added with a click of the mouse.

Question Answering Bot

The intelligent question answering bot is already a part of a comprehensive customer service solution.

In a pre-sales consultation or after-sales service, these chatbots can quickly respond to a range of customer inquiries coming in from a web page, a chat app, a public account, or some other small programs. An intelligent algorithm uses technologies such as semantic matching and sorting models to find appropriate answers to customer inquiries.

In cases where the problem is beyond the chatbot's capabilities, the customer query can be seamlessly transferred to a live customer service representative. The question answering bot can help improve efficiency  without losing any valuable sales leads or missing any possible customer feedback.

The HUAWEI CLOUD Question Answering Bot can even learn, automatically classifying similar customer problems, exploring hot issues and trends, and helping enterprises more deeply analyze customer feedback.

Natural Language Processing

NLP can also enable automated customer service, and it can provide multi-dimensional analysis of customer feedback. For many enterprises, there is a tremendous amount of user generated contents. However, collecting the scattered comments into one place for analysis is time consuming and expensive, and it is not possible to give them the level of attention and depth of analysis they deserve.

For example, when a new smartphone comes out, manufacturers need to find online complaints  very fast because they wish to respond to market demands in a timely manner. Hotel chains need to collect and aggregate user comments from multiple channels and automatically organize the data based on numerous variables because they wish to guarantee their quality of service and their brand reputation.

HUAWEI CLOUD Natural Language Processing can automatically analyze a vast sea of user comments, provide public opinion analysis, attribute-based sentiment analysis, and tag comments by type, so that an enterprises can respond to user feedback in a timely manner. HUAWEI CLOUD NLP provides the kind of fine-grained analysis you need for precise planning.

The Future of Semantic Technology

The next ten years will be a golden age for NLP, and HUAWEI CLOUD is dedicated to applying the most cutting-edge technologies of academia to business scenarios to solve the very real, very specific challenges our customers are facing today. We boldly envisage:

Voice interaction will be one of the next big trends for human-computer interaction. Intelligent chatbots and phonebots will be everywhere. In different vertical industries, there will be chatbots that help people with various tasks. Whether it's pre-sales consultations, order placement, or after-sales service; intelligent chatbots will be there to guide you through the entire process.

In the next five to ten years, these bots will develop the ability to analyze vast quantities of text. From reading and analyzing, to understanding and generating a summary of the content, all of this will be done by bots.

Personally customized services will become very popular. Virtual sales assistants or enterprise-level customer service bots will come with their own unique sets of capabilities and their own distinct personalities. They will be custom designed for the particular requirements of the scenario they will work in, or to be better able to put a particular enterprise’s competitive advantages front and center.

For consumers, interacting with robots is going to become a regular part of the consumer experience over the next few years. For enterprises, the use of AI technology can significantly reduce costs even while improving customer service. The potential is limitless. So what are you waiting for? The future is now.  

NLU Public Opinion Analysis

NLU Sentiment Analysis

TaskBot_Voice Interface