Chatbot is one of the biggest hot spots in the field of artificial intelligence, and it is also regarded as the vanguard of artificial intelligence entering the enterprise market and monetization. With the maturity of natural language understanding technology, Chatbot, which can work 24 hours a day, and can even start a multi-task operation mode with one to ten, has begun to appear in a large number of practical application cases in enterprises in 2017. According to Gartner, an international research consultancy, in 2018, more than 2 billion people around the world will often interact with Chatbot in a conversational way. By 2021, more than 50% of enterprises will spend more on ChatBot than traditional apps.
For developers who are new to Chatbot, how to find an entry point to build a conversation service is the biggest problem. Reading papers, watching videos, and looking for information often confuse beginners. The best way is to find a solution that is practical and time-saving at the moment. The following takes IBM Cloud as an example to demonstrate how to easily and quickly build an intelligent chatbot.
IBM Cloud is a one-stop cloud computing platform with many built-in useful APIs and tools. For developers, you can even complete Chatbot development with just a few commands without writing code at all.
In the actual development process, it will be found that IBM Cloud is a very suitable platform for deploying microservices, because there are many ready-made components or services on the platform for developers to call, the documentation of the interface is also very detailed, the deployment is very convenient, and the one-click service Very nice, if there is a bug, you can also go to the console to view the log. However, due to the need to surf the Internet scientifically, there may be a bit of trouble during use.
In addition, IBM's natural language processing module, Wastson Conversation, although the support for English is very perfect, but the support for Chinese is still in the test, and can only process the Chinese language in a very primitive way through word segmentation, or call a third party. The stutter participle API. The approximate key development process is to use Promise, open child_process and call python command to run stammer word segmentation to get the result, and then output it to the web page.
If you encounter the problem that child_process cannot get the output of the word segmentation result of the child process when the parent process returns data, you can use the promise of nodejs to solve the problem that the parent process executes the return output before the child process returns the result. If there is a problem with Chinese character encoding, you can add encoding: "utf8" to solve it.
Finally, let's take a look at the objective and real feedback from some domestic first-line developers and users.
Shabby-滔：
In this project, I tried to use IBM's natural language processing module - Wastson Conversation again, but unfortunately, this module's support for the English language is very complete, but the support for Chinese is really an experiment , can not be used, which makes it impossible to be as intelligent as a robot like Microsoft Xiaobing, and can only process the Chinese language in a very primitive way through word segmentation (third-party implementation - stuttering word segmentation). (Of course, if the entire project decides to use English as the main language, it is actually very easy to use the Wastson Conversation service, because the service provides perfect entity and intent extraction functions for English)
When I started this project, I thought of making a robot for air quality (Aqi) query. These interfaces on the Internet are also very rich (in addition to the ibm platform that can visualize the dialogue process design with building blocks, it is very easy to be able to Calling the third-party api and processing the return value, there are few setbacks in the platform deployment and coding stages, probably because the ibm bluemix platform is already quite powerful (PS> the website needs to climb a ladder to get up... But this is very important for developers to come It's not difficult to say)
Finally, to sum up, IBM Bluemix is a very suitable platform for deploying microservices, because there are many on-site components or services on the platform that provide developer calls (how to charge is another matter), and the documentation of the interface is also very detailed (of course not necessarily is Chinese).
AFei-Fran:
I encountered many problems during the development process, such as opening child_process and unable to get the output of the child process word segmentation result when the parent process returns data, which took me a lot of time to find a solution. Using nodejs's Promise can solve the problem that the parent process executes the return output before the child process returns the result.
In the process of deployment, the use of the IBM platform is not only a bit tricky (after all, you need to surf the Internet scientifically), the deployment is quite convenient, the one-click service is very nice, and you can go to the console to view the log if there is a bug. However, because it is Chinese word segmentation, there is no problem in local time. When deploying, I encountered the problem of character encoding, which was solved by adding encoding: "utf8".
Dragon Rookie:
This is the second time to explore the IBM platform. The first time I used redNodejs to write a weather query robot in a visual environment. This time I used IBM translate to make a translation robot. In general, git-style automatic configuration The environment is very convenient. As long as it is used properly, a good microservice website can be completed in two hours. However, at present, IBM translate supports English translation into other languages, and Chinese translations in other languages are translated for patents, and language translation is not supported.