What is natural language understanding?
Natural language understanding, NLU for short, refers by definition to the understanding of natural language by software. The generic term “natural language understanding” covers a multitude of computer applications, ranging from straightforward requirements such as natural language commands to a computer to complex tasks such as the complete comprehension of newspaper articles or lyrical texts. The majority of practical uses lie between the two extremes. Concrete examples of the application of natural language understanding are, for example:
- Classification of e-mails to filter communication within a company
- Chatbots that recognize a user’s concerns in order to generate targeted output
- Virtual language assistants that understand spoken input at every linguistic level and provide the user with an answer
Confusion repeatedly arises as to where the distinction between natural language understanding and other sub-areas of artificial intelligence lies, especially in contrast to natural language processing (NLP). An NLP system for processing natural language covers all aspects of communication between humans and computers, from input and processing to reaction.
Natural language understanding describes a technology that understands content. An NLU solution can solve a problem separately, but natural language understanding is more commonly part of an NLP system. Natural language processing is generally regarded as a superordinate concept; natural language understanding and natural language generation are subdisciplines.
How natural language understanding works
Computers have always been able to understand language. To correctly interpret human input and implement commands, however, the machines required formal programming languages such as C++, Java or Python for a long time. The real advance of developed applications is that modern NLU can process and interpret natural linguistic expressions.
Regardless of exactly what natural language understanding is supposed to do, most NLU systems have common components. The application requires a “parser”, a computer program responsible for dissecting and converting content into a format suitable for further processing.
In addition, the application must be able to access a lexicon of the language as well as grammatical rules. The wider the vocabulary, the better the system performs in complex tasks. Sophisticated solutions are capable of analysing complicated sentences, interpreting incorrect spelling and abbreviations or recognising dialects.
At the level of meaning, natural language understanding finally determines the intended meaning of an utterance. In the interaction of the individual components, NLU thus recognizes the important elements such as people, places, times and intentions from a piece of content.