Automatic Ontology-Based Knowledge Extraction from Web Documents vs. Automating the Extraction of Data from HTML Tables with Unknown Structure

In this report we compare the papers [AKM + 03] and [ETL03]. We show that the two proposed systems realize different goals with the same or similar underlying technics. • Source data of interest [ETL03] takes web pages containing HTML tables of interest for a given application domain as the input whereas [AKM + 03] considers unstructured text from webpages for the knowledge extraction process. • Resulting data format The difference between the output data is similar to the input. While [ETL03] returns structured data fit into a target schema, [AKM + 03] generates text in a narrative form specified by the user. • Interface specification [AKM + 03] is designed as a whole process from knowledge extraction over information management to narrative generation and thus provides an interface for human beings. [ETL03] however, only provides an interface to an information extraction procedure and is not designed as an complete application. • Internal data source While [AKM + 03] uses a knowledge base to store the aquired knowledge, [ETL03] doesn't outline a specific internal data storage approach. • Project structure The approach of [ETL03] is a stand alone project only built up previous research on extraction ontologies. [AKM + 03] uses several projects for the diverse tasks needed and combines them to a whole process. • Extraction ontology For both papers the central key aspect of the information extraction process is the extraction ontology. Without it, for every new page a wrapper has to be created. But with the ontology the wrapper creation can be automated.