Natural languages, multilinguisme and ontologies

Natural languages, multilinguisme and ontologies international experts in a specific domain (not in logic) wants to editan ontology automatic information extraction from texts by users that don’t want to be bothered with logic specific requirements in the project OMNIA automatic image indexation, w.r.t. an ontology (instantiate theA-box) , using companion texts generate short image descriptions w.r.t. the indexation the methods have to be robust for ontology T-box modifications SC and SR are the lexicons for concepts and relations resp.
Refx : x −→ Sx is the lexical reference assignment 1Maedche et al. 2003 : “Bootstrapping an ontology-based information −→ extend the ontology lexicon with synonym sets the ontology lexicon is very restricted (often one head-word for eachconcept) there are a lot of ways to say the same thing in NL utilisation of controlled natural languages (CNLs) CNL input engines to query/edit ontologies : 2http://attempto.ifi.uzh.ch/site & http://attempto.ifi.uzh.ch/ape3Dolbear et al. 2007 : “The Rabbit Language : description, syntax and conversion to OWL”4Funk et al. 2007 : “ CLOnE : Controlled Language for Ontology Editing”5Cregan et al. 2007 : “ Sydney OWL Syntax : towards a controlled natural language syntax for OWL 1.1”6Bernstein et al. 2005 : “Querying ontologies : a controlled english interface for end users“7Bernstein et al. 2006 : “GINSENG : a guided input natural language search engine for querying ontologies”8Bernstein et al. 2006 : “GINO : a guided input natural language ontology utilisation of controlled natural languages for queries andedition reduces ambiguity and complexity of natural languageallows users to insert or ask for information in the ontology without anylogical backgroundthe systems are portables OK for user interfaces but unusable for information extraction fromspontaneous texts allow unrestricted natural language for queries first approach : use a parsed syntactic tree highly dependent on advanced NLP methodshow reliable the parsed syntactic tree is ?OK for queries but too complex for longer texts 9Wang et al. 2007 : “PANTO: A Portable Natural Language Interface to Ontologies” 10Bernstein et al. 2006 : “Querix: A Natural Language Interface to Query second approach : ontology guided content extraction try to fill in the relations with the rest of the words NLP-reduce 11AquaLog 12 & PowerAqua 13 use few NLP notions (reduces complexity)usable to retrieve relevant information w.r.t. an ontology inspontaneous texts 11Kaufmann et al 2007 : “NLP-reduce : a naive but domain independent natural language interface for querying ontologies”12Lopez et al. 2004 : “Ontology driven question answering in AquaLog“13Lopez et al. 2006 : ”PowerAqua : fishing the Semantic Web“ translation “on the fly” of texts to the ontology lexiconlanguage 14 no extra-work when the ontology is modified 14Chang et al. 2007 : “Approaches of Using a Word-Image Ontology and an Annotated Image Corpus as Intermedia for automatic→ is the translation reliable ? extra-work for each language when the ontology is modified 15Espinoza et al. 2008 : “Enriching an Ontology with Multilingual Info collaborative Wiki environment to develop a huge ontologyexperts propose new concepts for the ontology from Wordnet entries (7languages supported)the new concept is actually inserted in the ontology when a consensusis reached between every supported languages benefits all Wordnet featuresconstructs of a real multilingual and multicultural ontology 16Vossen et al. 2008 : “KYOTO : A system for Mining, Structuring and Distributing Knowledge Across Languages and graft linguistic sub-ontologies to the concepts 17 built a linguistic sub-ontology frame (translation and morphosyntacticdata) for each concept of the source ontology : instantiate the frame in the target language 17Buitelaar et al. 2006 : “A Multilingual/Multimedia Lexicon Model for use UNL as interlingua so the system is a priori open on every language map UNL lexicon and relations with the ontology lexicon and logicconstructors allow a mapping update when the ontology is modified ! not compute complete UNL graphs for sentences ! annotate chunks of the text with UWs or very simple graphs try to find matching elements on the ontology side

Source: http://atoum.imag.fr/geta/User/david.rouquet/partage_documents/perso/NL_ML_Onto_28-08.pdf