Shadows: a new means of representing documents

Document production tools are present everywhere, resulting in an exponential growth of increasingly complex, distributed and heterogeneous documents. This hampers document exchange, as well as their annotation and retrieval. While information retrieval mechanisms concentrate on textual features (corpus analysis), annotation approaches either target specific formats or require that a document follows interoperable standards – defined via schemas. This work presents our effort to handle these problems, providing a more flexible solution. Rather than trying to modify or convert the document itself, or to target only textual characteristics, the strategy described in this work is based on an intermediate descriptor – the document shadow. A shadow represents domain-relevant aspects and elements of both structure and content of a given document. Shadows are not restricted to the description of textual features, but also concern other elements, such as multimedia artifacts. Furthermore, shadows can be stored in a database, thereby supporting queries on document structure and content, regardless document formats.