This paper is one of a series of papers describing RESEDA--an “intelligent” information retrieval system. RESEDA stores historical data on the Middle Ages, and is able to answer quite complex queries, such as: “Give information concerning posts that Robert de Bonnay may have had during 1414,” or “Give plausible explanation why a given administration gave a particular post to a particular person.”

The paper concentrates on the inference techniques of the system. At the lowest level, there is a pattern matching method which can deal with temporal data (time dependent). At the higher level, there is a top-down inference mechanism similar to that of PROLOG. This mechanism is enriched by two special purpose functions. The “transformation” function enables transforming a search-pattern to a semantically equivalent pattern and continues the search with the new pattern. This is a generalization of a common technique which is used in dictionary-based information retrieval systems [1]. The second function, the “hypotheses,” enables verifying a plausible explanation by using condition-based rules. The “hypotheses” function is used, for example, in answering the second query above. To execute its search heuristics, RESEDA provides a back-tracking mechanism similar to PROLOG’s, but does not use an equivalent of the “CUT” operator.

The paper is interesting and contains original ideas. However, it is not very clear. The presentation of the algorithms is too informal, and the examples are written in a meta-language which is not defined. Perhaps this paper should be read in the context of the other referenced papers on RESEDA.

One interesting point about the system implementation is that the system was implemented in APL. The author claims that APL is a very effective language to implement text-based expert system. This is certainly an interesting addition to PROLOG or LISP.