A Cooperative Comprehension Assistant for Intranet-Based Information Environments

A cooperative comprehension-assistant is described which represents text documents at three different levels of abstraction (surface, propositional and situational levels of representation). At the surface level texts are represented as bags of words without any linguistic structure. For the propositional level a definite but evolutionary growing set of predicates and concepts is used for selectively representing the most interesting parts of each text. By a latent semantic analysis the situational level is calculated as the third level, which is a parsimonious representation of the complete contents of a document with respect to some previously established frame of reference (metric vector space). Together the three levels provide more flexibility as well as precision for assisting human comprehension, document storage, retrieval and distribution. Furthermore, it is shown how this comprehension assistant is applied for knowledge dissemination in a particular kind of intranet-based information environment (oligo-agent system). The individual and social components of knowledge sharing within such an environment are described by knowledge construction and knowledge integration processes. To take the problem of long-term maintenance into account, the comprehension assistant and the oligo-agent system is embedded in a life-cycle model of cooperative knowledge evolution, consisting of a seeding phase, the system's evolutionary growth and periodical reseedings.