Management of semistructured data

A huge amount of data is available today on the Internet, or on the private Intranets of many companies. This data is structured in a multitude of ways. At an extreme we find data coming from traditional relational or object-oriented databases, with a completely known structure. At another extreme we have data which is fully unstructured, such as images, sounds, and raw text. But most of the data falls somewhere in between these two extremes, for a variety of reasons: the data may be structured, but the structure is not know to the user; the user may know the structure, but chooses to ignore it, for browsing purposes; the structure may be implicit, such as in formatted text, and is not as rigid and regular as in traditional databases; the data may be in non-traditional formats, such as the ASN.1 exchange format; the schema of the data is huge and changes often, so that we may prefer to ignore it. Several researchers have worked recently on problems related to data fitting this description, and have coined the term semistructured data for it. Two recent tutorials [Abi97, Bun97] contain an excellent introduction to semistructured data and a comprehensive bibliography on this new research topic.