Deep Understanding of Technical Documents: Part II. Automatic Extraction of Pseudocode

Abstract

Humans have the privilege to automatically have a deep understanding of technical documents, since they have the ability to deal with complex concepts coming from many different modalities, like diagrams, text, tables, formulas, graphics, pictures, etc. For many years researchers are working to transfer such potential to AI based machines. This paper takes the advantage of the synergistic and interactive enrichment of two TD modalities, the block diagrams and the associated natural language text, obtained to automatically generate pseudocode that describes the functionality of the system under study. The methodology for generating the code is mainly based on the mapping of the TD modalities into Stochastic Petri-nets (SPN) that enriches the system diagrams, from which the pseudocode is generated. The overall methodology will contribute to an automatic deep understanding of technical documents (TD) without the main involvement of humans. Two illustrative examples are also provided for describing the methodology.