Mining defining contexts to help structuring differential ontologies

Abstract

In this paper, we present an experiment dealing with corpus-based construction of “differential ontologies”, which are organised according to semantic similarity and differential features. We argue that knowledge-rich defining contexts can be useful to help an ontology modeller in his task. We present a method, based on lexico-syntactic patterns, to spot such contexts in a corpus, then identify the terms they relate (definiendum and genus or “characteristics”) and the semantic relation that links them. We also show how potential co-hyponyms can be detected on the basis of shared words in their definiens. We evaluate the extracted defining sentences, semantic relations and co-hyponyms on a test corpus focusing on childhood and on an evaluation corpus about dietetics (both corpora are French). Definition extraction obtains 50% precision and recall of approximately 40%. Semantic relation identification reaches an average of 48% precision, and co-hyponyms 23.5%. We discuss the results of these experiments and conclude on perspectives for future work.

Definitional verbal patterns for semantic relation extraction

“…The closest previous work to compare our results with is Malaisé et al (2005), who report an average of up to 55% Precision which is fairly similar to the 60% we obtained.…”

“…In accordance with authors such as Morin (1998) or Malaisé et al (2005), we believe that definitions contained in specialised texts could be considered the departing point for extracting semantic relations. Therefore, the automatic extraction of definitions from DCs is an important step for obtaining semantic relations within the specialised knowledge of any scientific or technical field.…”

Thesaurus enrichment for query expansion in audiovisual archives

“…The approach was later followed by [14]. For the comparison of definitions, the similarity measure can range from the percentage of words that occur in both definitions, which was used in [16], to cosine similarity between vectors of words in the definitions. Much to our surprise, in the present case the one-sentence descriptions of the online dictionary are the WordNet glosses for 99% of the words, giving us the luxury to avoid the choice of a similarity measure.…”

“…[25]) and has been applied to a wide range of use cases. For example, in [16], we have explored applying lexico-syntactic patterns to term definitions to discover semantic relations. Also other types of relations than hyponyms or subclasses have been discovered using Hearst-like patterns.…”

Pattern-based approaches to semantic relation extraction: A state-of-the-art

“…The work of Alfonseca et al (2006) explores a multitude of relations using the same general approach, such as employee-organization, painter-painting, filmdirector, etc. As shown in Malaisé et al (2005), in terminology, the main relations of interest are those revealing definitional properties of terms. Some relations have been studied much more than others.…”