Text2MARK: A text mining tool to aid knowledge representation - (MARK2)

With the first version of this work, Text2MARK, it was possible to obtain representations of database models with over 90% of accuracy. In this continuing work, some grammatical rules of the Portuguese language, called lexical predicates, were constructed, alongside other text mining metrics. With the incorporation of these new elements, Text2MARK was then applied to text sections taken from five scientific articles from different areas of knowledge. 84.2% of valid tuples were extracted and then applied in the construction of concept maps. Subsequently, these maps underwent a subjective evaluation by three different groups of users, who then attributed grades to the maps, ranging from 0 to 10, with means of 8.1, 8.1 and 8.3 per group. These results indicate that it is possible to extract and use tuples in the NAME-VERB-NAME format to represent knowledge by means of concept maps.