The explosive growth in recent years – forseeably still accelerating – in quantities and types of digitally encoded, stored and transmitted information throughout many sectors of society has created an urgent need for effective ways of making this information accessible and usable to its intended audiences. Prominent among the problems posed is that of searching for and actually finding the desired information within the vast quantity available, whether this is in the form of texts, pictures, musical or other audible forms, and so on. Research on how people seek and recognize the things they are looking for is therefore of great relevance today both as a subject of study in itself and as an aid to educational, commercial and leisure activities. This paper will present research on this subject using architecture as a specific case, particularly the example of 3D digital models created to visualize, analyze and document buildings. Collaborative design characterizes much of the building industry’s work – including architecture, engineering, construction and facilities management – and the emerging capabilities of computer-aided-design and -documentation techniques offer many potential advantages, not least of which is the ability to reliably share information among a project’s various participants. This potential is seriously undermined, however, by incompatibilities among digital models generated and/or exchanged by these parties due, for example, to differing modeling software platforms as well as to differing conventions for organizing model information even when platforms are identical. Some efforts have been made at standardization to address this. However, recognizing that the reasons for differences in organization or in choice of platform may be quite validly made (such as for the purpose of promoting the parties’ differing specialties and modes of work, which would suffer from excessive standardization) a more promising approach seems to be to find ways of reliably and usefully translating models which are exchanged. The project presented here addresses translation of models on the basis of geometric information which is held in common across platforms and organizational methods, relying only in a secondary role, or not at all, upon naming conventions and other labeling methods. Classification and searching are seen to be improved when based on the essential geometric content rather than on more arbitrary (even if conventional) descriptive schemes. The abilities conferred by such a technique are applicable in education as well as practice, helping users of digital models explore and exchange information. Although the approach and results discussed here are oriented specifically to buildings, it is apparent that similar strategies can also be used in other fields with digital information which is not entirely or primarily text-based.