OTTER - An information retrieval system for office automation

Design choices and most important features of an information retrieval system for office automation are described. The office environment is shown to be quite different from the traditional bibliographic search environment for which well-established information retrieval techniques exist. Therefore, the system offers facilities to model the semantics of portions of text (for example numeric data), to acquire a certain amount of user knowledge, and to support repetitive interactions. All pertinent documents are presented to the user, with the system ordering them in inverse estimated relevance order. It is shown that no single estimator can accurately estimate document relevance, because intended user access patterns are different. Consequently, knowledge on how the user wants to access data is acquired, and the appropriate estimators are used in each case.