Knowledge density in raw text: a criterion for assessing the usefulness of texts for expert systems

<p>Raw (unprocessed) text can serve as a source of rules for a rule-based expert system [1]. Several types of sentences from which rules can be generated have been described [2] [3], but this list is far from exhaustive. A prototype Python package for generating rules from raw text has been developed with great potential for further development [4]. Different texts yield different numbers of generated rules. The more rules that can be generated from a text, the more valuable the text is for an expert system and the greater the likelihood that the user will receive a high-quality answer from the expert system. To evaluate a text for the number of rules it contains, and therefore for its usefulness for an expert system, a criterion called knowledge density is introduced. This paper is devoted to familiarization with the knowledge density criterion. This paper describes: the knowledge density of the whole text, point or local knowledge density, text integrity or the emergence criterion of raw text, as well as various properties of the knowledge density of raw text.</p>