MultiNet

from Wikipedia, the free encyclopedia

MultiNet is a formalism for the representation of knowledge and for the representation of the meaning of natural language expressions. The abbreviation stands for multilayered extended semantic networks (multilayered extended semantic networks ).

The MultiNet paradigm was developed by Hermann Helbig as an extension of earlier semantic networks. It is a formalism that, in contrast to other methods of knowledge representation , aims particularly at the representation of knowledge formulated in natural language . For this purpose, a repertoire of around 140 relations and functions is provided, which capture typical semantic relationships between concepts. Furthermore, node attributes for aspects such as degree of generalization, quantification and facticity are introduced, which allow a precise description of the meaning of the term. As a further specialty, subnetworks can be encapsulated in MultiNet and represented by a higher-order node, which can itself appear as an argument for relations and functions.

The approach has been used in numerous practical applications of Natural Language Processing (NLP). It was used, for example, for the implementation of natural language interfaces to the Internet , for question-and-answer systems via semantically annotated corpora with millions of sentences, in deep semantic search engines, for the semantically based legibility check of texts as well as for the semantically based detection of duplicates and Plagiarism . In the (also commercially available) meaning-oriented, cognitive search engine SEMPRIA-Search, the MultiNet formalism serves, for example, as the basis for the description of the computer lexicon or the background knowledge base, for the syntactic-semantic analysis, for the logical answer finding and for the generation of natural language Reply.

The language processing technology based on the MultiNet paradigm was rated in 2019 by the jury of the # KI50 project in the automatic language processing category as "an important contribution to cognitively oriented language understanding by the computer". The MultiNet formalism is supported by various software tools. Among other things, these are used as the basis for building a large semantics-based computer dictionary. The range of tools includes a parser for German that converts natural language expressions (phrases, sentences or texts) into formal MultiNet expressions; a workbench for the knowledge engineer that supports knowledge acquisition and automatic knowledge processing; and a workbench for the computer lexicographer that simplifies the creation and management of large computer lexicons.

literature

  • Hermann Helbig, The Semantic Structure of Natural Language - Knowledge Representation with MultiNet. Springer, Heidelberg, 2001.
  • Hermann Helbig. Knowledge Representation and the Semantics of Natural Language, (2006) Springer, Berlin

Individual evidence

  1. ↑ Sven Hartrumpf, Hermann Helbig, John Leveling, Rainer Osswald: An Architecture for Controlling Simple Language in Web Pages . In: eMinds: International Journal on Human-Computer Interaction . tape 1 , no. 2 , 2006, p. 93-112 .
  2. ↑ Sven Hartrumpf, Hermann Helbig: Semantic search and keyword: a comparison of approaches and results on a German-language document archive . In: info 7: Media, Archives, Information . tape 29 , no. 2 , 2014, p. 62-66 .
  3. ↑ [1] GI website for the # KI50 campaign. Retrieved September 10, 2019.

Web links