SEMANTIC ANNOTATION TO SUPPORT AUTOMATIC TAXONOMY CLASSIFICATION
S. Kim, Saeema Ahmed‐Kristensen · 2006
Electronic documents are used throughout a product lifecycle and are created and accessed by users working in substantially different business operations and with very different areas of expertise. One way to organise corporate documents into easily accessible formats is to use a corporate taxonomy. A corporate taxonomy is the hierarchical organisation of domain concepts. It provides a holistic view of the organisation’s information architecture leading to better information flow between business operations and improved accessibility to unstructured texts. Unlike classic taxonomies (e.g. medical, biology, etc) in which objects map onto just one entry, a corporate taxonomy needs to be flexible in order to encompass diverse business operations. It must also facilitate searching and navigating through the multiple facets of enterprises. A corporate taxonomy should allow texts to be encoded with multiple metadata reflecting each operation’s domain model. However, organisations still need to maintain an integrated view of their documents. One way to maintain a balance between ensuring common access to the taxonomy and dealing with diverse business interests of the users is to focus on the main messages that the authors of the documents intended to communicate. These messages contain the information that people are most likely to reuse or refer to in the future. If these main messages can be mapped onto the taxonomy, the goals of consistent access to the taxonomy and maximum information sharing might be achieved. In organisations, people create documents to distribute information rather than to convey the views of individuals or groups. A document is encoded with various semantics and accessed by users who have very different interests. For example in the case of product reviews by customers, negative and positive customer opinions are the main messages for market researchers who are trying to assess market demands. On the other hand, designers are more interested in design-related issues, comments and problems. It would therefore be beneficial to include those semantics that facilitate searching for information in a way that reflects the interests of the users. For example, for a designer whose task is to reduce fan noise, guidance on how to minimise aerodynamic noise should be retrieved. On the other hand, if that designer is more interested in using a specific method for noise reduction, then documents describing the methods along with their advantages or disadvantages are more useful. It is therefore important to investigate how to extract and use the most appropriate semantics for indexing and searching documents using a corporate taxonomy. In general, taxonomy development proceeds by defining initial taxonomy entities, arranging them into a category tree and assigning information to each category. For the first two of those stages, manual intervention is necessary and professional indexers are often used in the process. However, when a taxonomy is to be used for indexing and retrieving documents, it is not cost effective nor practical to rely solely on the professionals for cataloguing corporate documents. Domain expertise is essential for