Abstract:
Abstract—Conceptual Graph is a representation commonly
used to express semantic relationship of natural language. This
work presents a method to translate Thai natural language
text to conceptual graphs regarding semantic relations based on
semantic roles between predicate and its arguments. Shallowing
parsing of Thai text and verb patterns as case frames are utilised
in identifying core entities in a context and their semantic roles.
Then, the argument with annotated roles are translated into
conceptual graphs that are able to logically and visually represent
relations of core terms. As a result, conceptual graphs of Thai
natural texts from Facebook posts in a marketing group were
generated. In the study, found issues regarding Thai specific
natural style are encountered and discussed.
Index Terms—Semantic Relation, Conceptual Graph, Information Extraction, Semantic Role