Conversational AI

Language Processing (NLP) and Language Understanding (NLU)

An important role in the understanding of language in language-based artificial intelligence is played by Natural Language Processing (NLP) and Natural Language Understanding (NLU)


For language processing (NLP), rules from linguistics are converted into computer algorithms to break down a sentence into sub-sentences and finally into individual words to subsequently enable understanding of the content. In the process, the functions of words - for example, subject, predicate and object - and the type of word - for example, preposition or article - are also recognised.

NLU systems enable computer programms to understand the meaning and intention behind natural language. Unlike NLP, NLU technology can detect semantic connections, the meaning and context of statements. This allows inferences to be drawn and connections between sentences and sentence components to be recognised, even when there are incorrect inputs or spellings. NLU systems are used to support natural language human-computer interaction. Conversational AI

Both the NLP and NLU engines were developed specifically for open bi. There are no third party provders involved.

The language understanding of open bi

Central to the understanding of language-based information is the word. In majaAI's Artificial Neural Networks (ANN), a word does not correspond to just one neuron. Rather, a word is a complex construct that contains further information. Conversational AI

open bi knows to every word:

  • different synonyms: for map there are synonyms like world map, atlas
  • its sentiment: "great", "good", "brilliant" are positive sentiments, there are also negative and neutral sentiments
  • its morphology: "go" has different tenses such as "he went", "she will go" or "I go"
  • the meaning of logical operators: "with", "without", "except", etc. as in "I am looking for a flat with a balcony"
  • fitting entities: In response to the question "Which countries are in Europe", the network not only finds 47 countries, it also knows various properties for each country (i.e. the entity "country"), for example which language is spoken in that country

How open bi learns 23 new languages

majaAI has a direct link to DeepL technology that can be optionally activated. DeepL is a service that uses artificial intelligence to translate texts into 24 different languages. Like us, DeepL relies on the highest level of data security through end-to-end encryption and ISO27001-certified data protection concepts.

majaAI is thus also able to answer questions asked in other languages. This works as follows:

Let's assume that Maja can answer the question "Hello, how are you?". The phrase is trained in the system and assigned to a corresponding answer.

  1. For example, a person asks the question "Hallo, wie geht es dir?" (German) via the chatbot.
  2. Maja searches her neural network for the corresponding answer and initially finds none.
  3. Since DeepL is activated, she sends the posed question via an encrypted channel to DeepL. She also tells DeepL that Maja only speaks English.
  4. DeepL detects the language in the question asked and tells Maja: This question was asked in German, the english equivalent is "Hello, how are you".
  5. Maja now searches the neural network again and finds a suitable answer! She now sends it again to DeepL for translation into German.
  6. DeepL sends the translation, for example "Mir geht es sehr gut, danke!" back to Maja.
  7. Maja returns her reply to the user in German.

With the help of this synergy, majaAI will in future only need to be trained in one language and will also be able to answer questions in 23 other languages. DeepL is constantly expanding and improving the service. An up-to-date list of available languages can be found at

KI without programming: majaAI

open bi can be used as a pure API solution and be combined with custom frontend components. However, bi excellence software GmbH also offers a fully comprehensive self-service AI solution called majaAI. With majaAI, users can independently create their own speech platform without any programming knowledge and then integrate it, for example, as a chatbot on their own website.

majaAI is also the basis for various industry-specific platform solutions in the field of artificial intelligence, such as fragBob or hallomeinebank. You can also find more information about majaAI at