Automatic metadata extraction and classification of spreadsheet documents based on layout similarity

Effective information search is becoming a key success for business. Metadata is an essential part of modern information system since it helps people to find relevant documents from disparate repositories. Automatic document metadata extraction has received attention in recent years as it is an important task in generating powerful search indices to support effective information search. The objective of this paper is to propose an innovative method that automatically performs metadata extraction and classification on the spreadsheets having layout similar to that of a given sample spreadsheet whose metadata is previously defined. Metadata classification is based on document types (e.g. purchase order, sales report etc) and data context (e.g. customer name, order date etc) so that users can define the meanings of the keywords in the search query. Therefore, search engine of this work returns the search results that match user search intention more than those of conventional keyword search engines.