Automatic generation of fill-in clues and answers from raw texts for crosswords

This paper presents a method to generate fill-in clues and answers for building automatically a crossword. Answers are capitalised words present in an input sentence and clues are segments of the dependency syntactic structure of that sentence. The pairs (Clue, ANSWER) are extracted from a collection of raw sentences related to the history of Sarawak. This work is at its early stage, and thus the proposed method that generates automatically fill-in clues, was tested on a small set of sentences and the obtained results are promising. Near 53% of the generated fill-in clues are considered correct. The major contribution of this work is the innovative strategy used to read the result of a pre-order depth-first search applied on a dependency graph to generate the clues. The clues and answers generator is implemented in Python.