This table contains data directly pertaining to References, which are always parsed out of the "CR" ISI field. Each subsequent reference string contains a series of tokens separated by the string ", " (a comma, then a single space character). Further down below, this article describes the limited set of reference string formats supported, which are defined by what tokens can be parsed out of them and in what positions they can be parsed.

There are various ways to determine what data a token is supposed to specify, depending on what position the token occurred in. For example, if a token that starts with the letter 'V', followed by numeric digits is in the third of four positions, those numeric digits in that token is treated as the reference volume (REFERENCE_VOLUME).

This table includes all fields that can possibly be parsed out of any support reference string format, which are:

The following reference string formats are supported, organized by number of tokens present:

In the various built-in extractions, when References are matched to Documents, they are considered as part of the set of inner documents. When they are not matched to Documents, they are considered as part of the set of outer documents.

See CITED_REFERENCES, DOCUMENT, PERSON, SOURCE, and How Abbreviated Names are Parsed.