Deduction of metadata for tabular structures in plain text

Tables are a very effective way of presenting information. Because of their flexibility and ease of reading, tables have long been used as a method of presenting information in a clear and concise manner. With the introduction of the Internet, large bases of unstructured text have become available, many of which contain information in a tabular form. There is a growing awareness of this information and the need to link it with more conventionally held information in other applications so that its full potential can be realised. The problem with tabular information is that although the nature of its presentation makes it easy for the reader to understand, it is not possible to access and utilise this information automatically in a computer application and so link it with other information. Our aim is to develop a set of tools that will allow us to locate, extract and process tables from plain text documents in order to reuse the potentially valuable information they contain.