Fax: An Alternative to SGML

We have argued elsewhere (Church and Mercer, 1993) that text is more available than ever before, and that the availability of massive quantities of data has been responsible for much of the recent interest in text analysis. Ideally, we would hope that this data would be distributed in a convenient format such as SGML (Goldfarb, 1990), but in practice, we usually have to work with the data in whatever format it happens to be in, since we usually aren’t in much of a position to tell the data providers how to do their business. Recently, we have been working with a collection of 15,000 AT&T internal documents (500,000 pages or 100 million words). Unfortunately, this data is stored in a particularly inconvenient format: fax.