Artificial Intelligence (AI)

AI stands for artificial intelligence. AI is a field of computer science in which software and hardware are used to recreate human thinking and human behavior. At FIEBIG, we focus on improving customer service processes through appropriate AI solutions. We see that the combination of several AI techniques like speech recognition, natural language processing and semantic analysis leads to successful and sustainable results.

Narrow AI

Narrow AI or weak AI is a term for AI systems that cover only a defined i.e. restricted range of tasks. Most of today’s AI systems are narrow AIs. Although the currently available machine learning systems make it possible to create AIs for many different areas, each AI covers only the task area on which it has been trained. An intelligent chatbot on an airline’s website is expected to be perfectly familiar with flight bookings and aviation issues, but it will not be able to have a meaningful conversation about sports, cars or fashion.

Artificial General Intelligence (AGI)

An artificial general intelligence (AGI) is an artificial intelligence that can solve all tasks that a human being can solve with his or her intellect. Unlike the currently available narrow AI systems, AGI can be used for any task without being completely re-trained. AGI needs to be able to develop solution approaches independently, for it should resemble human beings in terms of operation and solution competence. Thus, AGI is the original goal of AI research and development. At the moment, no AI systems are known that fulfill these requirements.

The enhancement of AGI would be super AI, which would surpass the intellectual performance of man.

Natural Language Processing (NLP) / Natural Language Understanding (NLU)

NLP includes many related techniques for processing natural language in form of textual data or audio data, which is why NLP is a combination of linguistics and computer science.

Natural language understanding (NLU) is a subdomain of natural language processing (NLP) and hence of computational linguistics. The goal is to recognize, to understand and to analyze natural language with the help of software algorithms. This is why NLU is part of artificial intelligence.

By employing NLU technology, it is possible to analyze complete sentences in real time as well as to capture their meaning and context. Entities and concerns are extracted and passed on to other systems for processing.

For these reasons, NLU is perfect for interactive speech recognition and speech dialogue systems and hence for the use in your customer service.

NLU in Customer Service

In the context of customer service, you want to:

  1. understand what a customer tells you,
  2. understand what the customer means, 
  3. analyze this message and then 
  4. return a defined response, or perform an action.

Actually, this process is what your employees practice day after day. NLU is part of the solution if this work is optimized by AI.

FIEBIG Integration Services

The FIEBIG AI solutions use state-of-the-art NLU technologies which allow you to process natural language across all contact channels and to recognize your customers’ intents. Due to the FIEBIG integration services, FIEBIG can integrate these functions into almost any system and application landscape. ​​​​​​

Learn more about FIEBIG integration services

Intent Recognition

Intent recognition is about extracting the intent out of a text or utterance. In the AI environment, the recognition of intents belongs to the NLP domain. The aim is, for example, to interpret customers’ intents automatically and to recognize their actual intent in the requests. Within the context of an airline, the recognized intent of the sentence “I would like to fly from Hamburg to Berlin” would probably be flight booking. Once the intent has been identified, the next steps can be initiated.

Intent recognition is the central component for the automated processing of customer queries. If the system has recognized what the customer wants, it can initiate the next dialogue steps or start the respective processes. Currently, intent recognition is able to identify even multiple intents out of a document.

Intent recognition is often part of systems that classify emails or documents. Moreover, it is employed in chat or dialogue Systems. Especially in speech processing, the possibility of real-time recognition is important in order to avert unwanted breaks in automated dialogues.

Computational Linguistics

Computational linguistics comprises the modeling of natural language for further processing with software as well as the development of IT-based methods for linguistic questions.

Simply put, computational linguistics can be seen as basic linguistic research for automation. The respective results are used in natural language processing.

Sentiment Analysis

Sentiment analysis refers to the use of NLP, text analysis and biometrics for the systematic identification, extraction, quantification and investigation of affective states and subjective information.

Sentiment analysis is the attempt to determine a speaker’s or writer’s attitude. The attitude may be a judgement or evaluation, an affective state (i.e. the author’s or speaker’s emotional state) or the intended emotional communication (i.e. the emotional effect intended by the author or interlocutor).

Machine Learning

Machine learning is related to the topics of data mining and big data. A machine learning system is fed with data based on which it automatically recognizes rules or patterns in the data. It learns from the data and is subsequently able to apply the learned rules and patterns to new data.

Neural Networks

Neural networks or artificial neural networks picture biological neural systems in hardware and software. This is an attempt to copy biological processes such as thinking inside of the human brain. Neural networks are mostly used as basis for AI systems.

Deep Learning

Deep learning or deep structured learning is a subdomain of machine learning. Deep learning contains various optimization methods, for example for neural networks. As the name implies, the depth of a neural network is increased by the addition of new layers.

Unsupervised Learning

Unsupervised learning is a variant of machine learning. Such a system is given a sufficiently large amount of input based on which it recognizes patterns on its own, i.e. without the guideline of results and categories. By reference to the autonomously determined patterns and categories, the system can accordingly allocate new data as well.

Supervised Learning

Supervised learning is also a variant of machine learning. In contrast to an unsupervised learning system, a supervised learning system learns by being given the expected result for each input sentence. From a large amount of data, the system obtains rules for determining the result. These rules are also used to evaluate unknown input correctly.

Difference between Unsupervised and Supervised Learning

Both are variants of machine learning. The difference is that in supervised learning, each input is given an expected result. This allows the system to build rules that lead to the requested result. Unsupervised learning systems lack these presets and thus have to find patterns and categories in the data themselves.

Cognitive Services

Cognitive services are described as a collection of mostly AI-based or machine-learning-based functions. The focus of cognitive services is on recognition, interpretation or generation of images or language. Moreover, the interpretation of emotions is often part of cognitive services. FIEBIG provides cognitive services by Microsoft. Due to the high demands on computational power, these services are usually offered as cloud services. With the FIEBIG AI integration services, FIEBIG presents a toolset as well as long-term experience in order to easily integrate cognitive services into your system landscape.

Turing Test

The Turing test was defined by Alan Turing in 1950. It is used to check whether a machine’s or AI’s ability to "think" corresponds to that of a human being.  For that purpose, a test person has an extensive conversation with two subjects –one being a person, the other being an AI– via keyboard and screen. The test person does neither know nor see which subject is hiding behind which screen. If the test person cannot identify with certainty which subject is the AI and which subject is the human being, the AI has passed the test. The Turing test is repeatedly referred to in the field of computer science and artificial intelligence.


Singularity or technological singularity describes the point of time at which the development of current futuristic technologies like AI or robotics results in a technology that independently improves, evolves and thus decouples itself from human development. Once the singularity has been reached, predictions about further developments are no longer possible because man will no longer be the leading element.

I want to know more!

All fields marked with * are mandatory