50 likes | 69 Views
Step by Step Approach for Identifying & Constructing Dimensions of an Ontology. Building a Top Down Ontology From the Bottom Up. draft (v0.8): DeniseBedford / 2006.06.08. Step by Step Approach. Step 1: Identify the boundaries of the ontology
E N D
Step by Step Approach for Identifying & Constructing Dimensions of an Ontology Building a Top Down Ontology From the Bottom Up draft (v0.8): DeniseBedford / 2006.06.08
Step by Step Approach • Step 1: Identify the boundaries of the ontology • What will be ‘ontologized’ (broad definition of content)? • Who will use the ontology? • How they will use the ontology? • Following steps pertain to creating one dimension of an ‘ontology’ to apply to content -- • Step 2: Create a content inventory • Identify the sources of content • Use an inventory tool (COAST) to generate a full inventory • Working group weeds/selects from the inventory to create a core set of content to work with • Step 3. Extract list of concepts from the content • Use the inventory to capture content items as a training set • Identify the types of concepts to be extracted – noun-phrase descriptors, entity identifiers, names, institutions, etc. • Configure and run the concept extraction
Step by Step Approach • Step 4. Review the list of concepts • Quickly scan the concept list to determine concentrations of concepts • Check whether these concentrations make sense in terms of ‘categories’ • If so, begin to build a categorization profile and organize the concepts within • Determine what’s missing from the list of concepts (domain experts help us here..) • Determine what is in the list that is not pertinent to the topic (peripheral or out of bounds for the topic) – (domain experts here us here, too) • Prune the list of concepts – in some cases find new content and repeat the process • Step 5x. Build the categorizer profile • Build a rule-based categorizer around the concept clusters (manual bunching at a very coarse level) • Or,…check clustering of concepts using a clustering engine (here you can feed the refined list of concepts back into a clustering engine and run them against the training set)
Questions for Domain Expert Review • If you were talking about ontology with an expert, are all of the concepts you would use included in the list? If not, what is missing? • Are there a few concepts missing, or is there a larger domain or knowledge area that is missing? • What is in the list that doesn’t pertain to ontologies? • If you were looking for information about ontologies – from an expert point of view – would you use any of these concepts to search? Which ones are missing? What shouldn’t be in the list? • If you were looking for information about ontologies from a novice’s point of view – what is missing from the list of concepts? What shouldn’t be there?
Step by Step Approach • Step 6. Test the Categorization Profile against the content • Define the xml output structure for the metadata • Run the profile against the content • Review the categorization results • Accept/refine the profile • Other steps to creating the full ontology • Determining what kind of functionality you need to support use of the ontologized content • …search & discovery system • …browse of categories of content • …reporting • …recommender engines