The interaction of ontologies and natural language processing and even more so at the fundamental level—Ontology and linguistics—can be good for many conversations and debates about long-standing controversies. In this chapter, we shall focus on the former and take an engineering perspective to it1. In this case we can do that, because the points of disagreement are ‘outside’ the ontology as artefact, in a similar way as an OWL file is indifferent as to whether a human who reads the file assumes a particular OWL class to be a concept or a universal. Regardless whether you are convinced the reality is shaped, or even created, by language, or not, there are concrete issues that have to be resolved nonetheless. For instance, the OWL standard had as one of the design goals “internationalisation” (recall Section 4.1: Standardizing an Ontology Language "Historical Notes"), which presumably means ontologies being able to handle multiple languages. How is that supposed to work? Anyone who speaks more than one language will know there are many words that do not have a simple 1:1 translation of the vocabulary. What can, or should, one do with those non-1:1 cases when translating an ontology in one’s own language or from one’s own language into, say, English? Where exactly is a good place to record natural language information pertaining to the vocabulary elements? Does translating an ontology into another language even make sense? The area of multilingual ontologies aims to find answers to such questions, which will be introduced in Section 9.1: Toward Multilingual Ontologies.
A quite different type of interaction between ontologies and languages is that of ontologies and Controlled Natural Language (CNL), which may avail of multilingual ontologies to a greater or lesser extent. One can verbalise—or: put into (pseudo- )natural language sentences—the knowledge represented in the ontology. This can be useful for, among others: interaction with domain expert during the knowledge acquisition and verification phases, automatically generating documentation about the ontology, and anywhere where the ontology is being used in an ontology-driven information system (e.g., SNOMED CT with an electronic health records system). The general idea will be described in Section 9.2: Ontology Verbalisation.
Besides these two interactions between ontology and natural language, and the use of NLP for ontology learning that we have seen in Section 7.4: Text Processing to Extract Content for Ontologies. There are also areas of research and working technologies where ontologies enhance NLP applications2. This version of the textbook does not include a separate section on such ontology-driven information systems, however, for the scope is ontology engineering. It is duly acknowledged that some of the solutions that came out of the application areas can be useful for the aforementioned tasks, and the Semantic Web as application area in particular. This because the Web is global, so it makes sense to create a Multilingual Semantic Web (see also the recently published handbook [BC14]). It turned out there are some inherent problems of the original vision [BLHL01] to overcome [Hir14], and insights gained there assist with solutions for ontologies in the general case, regardless whether that is within the Semantic Web with OWL or another logic.
1Debates include the topic whether language shapes reality and therewith ontology or is used to approximate describing the world.
2E.g., it has been shown to enhance precision and recall of queries (including enhancing dialogue systems [VF09]), to sort results of an information retrieval query to the digital library [DAA+08], (biomedical) text mining, and annotating textbooks for ease of navigation and automated question generation [CCO+13] as an example of adaptive e-learning.