This paper presents a semantic data model called AIS", in which we can directly embody entities", and associations" between entity sets. For the formal foundation of the concepts for modelling information, we should choose as few as possible basic concepts. From this point of view, three kinds of association, existential association", Cartesian aggregation" and referential association" are proposed. Especially, in existential association, we can represent order" of entities when the entities are developed from one original entity. We also present the operators which can extract information from the given database and construct complex objects" temporarily. As the complex objects have hierarchical structures, that is, an object may consist of sub-objects and the subobject may consist of sub-subobjects, we should provide some mean to watch and scan a level of subobjects in an object. For this purpose, we introduce a new notion rank descriptor", which enables users to specify a set of subobjects of same depth, and to apply them operators simultaneously. Finally we discuss about a logical data structure and its manipulation system ASD, to show the capability of efficient implementation of AIS.