Automatic Construction Of Water Environment Ontology From Unstructured Texts

Water environment is the most complex part of the natural environment system,which contains vast amounts of data and information. In order to ensure that users caneasily access, share, and reuse the data and information in the field of water environment,a suitable technical method is needed for effective organization and integration. As a clearand specific repository, ontology has received high attention and application. This paperintroduces a method of ontology to manage a lot of information effectively in the field ofwater environment. At present, most of existing ontologies mainly are built by hand orsemi-automatic methods. This period involves a number of experts who must have clearand comprehensive understanding about the field. Building ontology manually is a timeand man power-consuming job, but quality can not be guaranteed, and the existingontologies have the problem of poor universal application to lead them not be used in thefield of water environment. So how to extract concept and relation from waterenvironment quickly and effectively and to express domain knowledge is becoming anurgent demand.In order to automatically build the ontology of water environment, firstly a massiveset of knowledge texts in the field of water environment has been collected as unstructureddata source, and the technology of natural language processing is used to convert texts towords. Secondly get the "word-text matrix" based on statistical method. Use the method ofsingular value decomposition to project the matrix onto a low-dimensional space andeliminate the semantic ambiguity between words and texts to make the concept ofinformation come out. Now, the concept extraction has been completed. Finally, extractthe hyponymy relations between concepts based on the hierarchical agglomerativeclustering algorithm. Compute pairwise similarity between the concepts, and then mergetwo concepts whose distance is the most minimum until merging into the largest superiorconcept. In this way, we complete the construction of water environment from unstructured texts automatically.In this paper, the system of water environment ontology from unstructured textsshortens the cycle of ontology construction, saves the cost and avoids the differencescaused by inconsistent understanding between the domain experts. It improves the qualityof the ontology construction partly and provides a valuable reference for the automaticconstruction of water environment ontology.