Cognitive computing is the next level of software development. The classical appraoches via tabulators, procedural or object based programming will be replaced with a new paradigm called cognitive computing. It can be seen as a new discipline, opening new possibilities for business and IT, connected to the currently joung generation of people. They are used to mobile devices, support chats and phone bots. Old channels like classical support is not always expected anymore as it is too slow in some scenarios. Within this article I would like to print out IBMs Watson technology and how it can be used for cognitive computing technology. Other platforms are based on Microsoft Azure or Amazon Alexa skills. Google itself has no product available at the market at the moment, even if some offers remind to cognitive features.
IBM Watson
IBM Watson is provided via IBMs Cloud offerment Bluemix and can be accessed via its cloud interfaces. As topic on most IT companies CIO Roadmap, it is also a growth place for IBM and solutions not that spread accross the market. The opportunities this technologie enables are therefore not yet understood in most business departments; sometimes not even known. The technology can be used for mutliple purposes:
- Structured data
e.g. based on given big data repositories. - Unstructured data
e.g. within Jeopardy, where the question to an given answer must be genereated:
Nature of Data
IBM provides several Watson offers to address those specific topics, as they require a different approaches. Structured data is normaly available in a known format, such as Standard Operating Procedures, Working Procedures or Frequently Asked Questions. This kind of data is addressed via IBM Conversation. They consists of common questions, which are easy to answer and therefore represent a short tail in complexity. Unstructured data is used when addressing complex questions, which are not provided in single data sources. Sometimes even an understanding of several documents and context is required. IBM Discovery addresses this kind of data sets; it is represented as long tail in complexity. A common scenario is the question for the required visum for a sailor, which flies from Germany to the USA to board a ship for working on international grounds during a trip top South Amerika. They are less often asked and the solution requires several complex topics:
- Visa
- Travel regulations (nationalities)
- Work permit
- Flight restrictions
IBM Conversation
For structured scenarios IBM Conversation can be used. Commonly it is realized via bots, which interact via voice or text chat with users. There are several additional modules which can be addressed, e.g.:
- Internationalization (Language Translator)
- Replying to the social mood of human counterparts (Tone Analyzer)
- etc.
By using Watson services it is possible to process text or voice messages and reply as text or voice to the counterpart. You can also include personal preferences of the counterpart; more modules will mist likely come soon as technology is a growing one. A brief (and incomplete) overview of Watson services can be found below (as of 07/2017)
Use Case: Chat Bot for Frequently Asked Questions
A common use case is the interaction with a chat bot for application related frequently asked questions:
- How to register for an application?
- How to reset the users password?
- How does a certain screen value is calculated (user guidance on screen level)?
For those questions IBM Conversation is used. You can also train the bot to allign to company internal terms and shortcuts. For addressing certain information areas, Watson has Intents, which can be defined. You can add the FAQ questions at this place to reference those knowledge areas:
You can also allign intents with Watson Entities, which are interaction objects, e.g. certain station offices (Frankfurt, New York) or car types (BMW 1-Series or 3-Series). These can be mapped later on for object specific replies:
By mixing those items up in Watson Dialogues, you can create conversations. Watson itself takes care to map given user context to defined intents and objects and provide a defined reply based on these structured items:
The outcome – as text chat – are dialogues which can be used on support chats, before involving employees:
If you compare the dialogue above with the given intent definition details, you might notice that the actual question “How can I go in?” is not defined; but Watson identified the question as related based on the context:
This context understanding can be trained automatically or manually. From my perspective automatic learning is too dangerous, as any context will be learned; if people train bad things, Watson will take over this kind of information without question and reply accordingly. Therefore manual training should be preferred. An introduction to training is available online from IBM.
Conclusion
From my perspective cognitive computing is a new paradigm about how to interact with technology and how to guide and support users. There are some obstacles in organizational and cultural change to be come open for these kind of topics. On the other side, huge efforts can be solved easily and reduce costs in Service Operations.
Further Watson opportunities can be leveraged by evaluating the following use-cases for Watson Discovery purposes:
- Watson News Discovery
Summarizes and score news about companies. An indication with related topics to a company and a score (good/bad) is gathered. - Watson Image Recognition
Identifies objects in images and provides possibilities to interact with it. - Watson Data Transmission Demonstration
Provides a visualization for message sizes for Watson services. Small messages are essential for certain industries (e.g. for plane services or interstelar missions). - Watson Corpus Explorer
Visualizes knowledge databases and provides access to navigate through the data.
