Artificial intelligence — the next level

AI is in countless applications and tools that make life easier for us in private and professional environments. A large part of the support systems that we use every day fall into the area of so-called weak AI. Such systems usually work based on rules and are limited to a clearly defined area of application. They are unable to “think outside the box.” The latter is only possible when artificial intelligence is used, which is based on a modern understanding of machine learning.

So-called deep learning algorithms are based on artificial neural networks whose structure and function are based on the neurons of the human brain. Thanks to a large information base and the structure of neural networks, deep learning systems can independently link learned content with new content. They are also able to make analogies without being confronted with certain key terms in the context of a question. You know comparable situations from everyday interpersonal communication. For example, when it comes to the statement “This is the sports car among telephones,” it should be clear to everyone that the word “sports car” is to be understood as a synonym for “top model” in this context. Accordingly, do you ask a chat or voice bot based on deep learning the question “What is the London of France called? “, he correctly answers with “Paris.” In this example, the bot presents the appropriate solution without mentioning the key term “capital”.

The more meaningful data such systems process, the more reliable they become. In this way, the “learning effect” ideally leads to continuous self-optimization.

But how does deep learning work in practice?

Let's say an insurance company wants their Automate customer dialogue and set up a modern language portal to adequately answer customer inquiries. Ideally, there is already a sufficiently large number of documented customer inquiries. This database is used to Voicebot to train him in his industry-specific field of application and to familiarise him with the logic and range of formulations used by insurance customers. In addition, data can be imported from industry-specific or general knowledge databases, such as online encyclopedias such as Wikipedia. In principle, it is possible to tailor the body of the text exactly to the requirements of a company.

In order to enable artificial intelligence to link and analyze existing data and draw precise conclusions from it, the immensely large database must be made manageable. Words that are not meaningful in the context of application are deleted, whereas relevant terms are reduced to the root of the word. In this way, instead of various word forms (e.g. “asked,” “ask,” “question,” etc.), only the basic form of each word (in this example: “ask”) is retained, so that terms in the existing body of text can be weighted more precisely. The text or input information is then converted into vectors. In this way, distances can be calculated in vector space and words can be weighted according to their context-dependent significance.

As soon as the available database, i.e. the entire body of text, has been processed, artificial intelligence has an easy time. In our example, the language portal would now have the task of searching incoming requests for patterns based on training data. In the casually worded human statement “my car is junk,” the deep learning system recognizes, for example, the relevant terms “car” and “scrap” and consistently assigns the request to the claims department responsible for motor insurance. In addition, artificial intelligence can fully identify the issue by asking the customer to specify the damage through specific questions.

With the help of these deep learning systems, even terms that do not appear in the underlying data set can be processed. For example, if a customer says “It's about my dental crown,” the request can be processed even if the word “dental crown” wasn't part of the body of the text. Traditional bots are unable to do such tricks. Modern AI systems, on the other hand, can assign unknown terms using the distance measures mentioned, provided that they are similar to words that are contained in their data set. In our example, it would be sufficient to recognize the term “dental crown” if the body of the text contained a word such as “dental treatment.”

One bot, many benefits

So it is different from conventional chatbots or voice bots do not need to think through all potential dialogue paths in advance and take into account the entire range of customer formulations. A system based on modern AI methods only requires a sufficiently representative, appropriately prepared database to be able to work completely independently. For example, by opening up terms unknown to him without outside help.

The benefits are obvious: On the one hand, the quality and flexibility of dialogue management are improved, and on the other hand, the scalability of the solution is significantly simplified. If the insurance company, from our example, wants to expand its language portal to other areas of the company or cover a larger number of use cases, it is therefore not necessary to develop new dialogue paths that are sufficient to provide and prepare a suitable database. The nice side effect? It not only saves time, but also a lot of money. In addition, the maintenance costs for modern AI systems are comparatively low. To adjust the chat or voice bot to changing conditions or new requirements, all you have to do is replace or adapt the data set. The system does everything else by itself and more: The longer it plays in a specific area of application, the more reliably it works.

IP Dynamics has the necessary know-how and the database to make deep learning algorithms even more efficient through industry-specific adjustments, such as vocabulary. In addition, the portfolio includes ready-made modules, for example for identifying concerns (especially in the insurance sector), as well as solutions for identification and authentication (e.g. for information on the processing status of transactions or recording missing data) or recall management. Another decisive advantage: In Contact Center IP Dynamics brings together all relevant customer service channels — telephone, email, chat, messenger, social media. Linking a modern AI system with our omnichannel platform offers considerable potential: In the complete solution, all incoming data is easily collected, linked and evaluated. And the AI is getting smarter and smarter and...

‍