![]() |
![]() |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() ![]() ![]() ![]() ![]() |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Home | Research | Research Programs | People-Centred Technologies |
People-Centred TechnologiesMetaxtract: Text Data Structuring ProjectThe main aim of the Metaxtract project is to enable automatic or semi-automatic semantic annotation using linguistic techniques over text-rich resources for The Semantic Web. The ongoing Metaxtract project, which began in May 2003, is one activity of the Semantic Web Lab and is part of the Semantic New Brunswick Initiative. One application of the research results will have the goal of making regional businesses more visible. Research ContextDevelopment of Semantic Web applications depends on having a set of richly annotated data. One of the challenges of migrating from current web content to Semantic Web enabled content is the massive amount of annotation required. Put another way, the advent of the Semantic Web will create a demand for structured data and today’s web is a vast repository of unstructured text data available for structuring. We are currently experimenting with (semi-)automatic extraction or validation of data plus metadata using information extraction techniques over text. Research PrototypeIn a first phase, we have developed a prototype that combines structured and unstructured data in a novel way. Our source of facts for the prototype is a provincial government database containing consistent data on over 2000 New Brunswick manufacturing companies. This structured data includes typical business information such as contact details, number of employees, sector and product information. Many of the companies also provide a link to their corporate website, and this subset of the web served as our unstructured data source. The resulting prototype is the NBBizMapper. Our interface makes it possible to run a keyword search over the web content and to return the hits in a structured fashion, organized by location and company name. Location information is displayed graphically for visualizing the distribution of relevant companies. This kind of display could be useful for business analysis. Building from here we hope to develop methods for making regional businesses more visible and for increasing the exchange of goods within this region and to external markets. Possible applications includeThe approach of this project is general enough to apply in many application areas. In addition to the initial NBBizMapper prototype, which applies this technology to benefit regional industry, other application areas are under consideration.
Results could impact on the following sectors
Inquiries about collaborating in this research are most welcome. Anyone interested in collaborating in this project is asked to contact the research contact listed below. Research ContactDr. Irina Kondratova Business ContactMarc-Alain Mallet |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() |
|