Monolingual alignment with moves for genetic criticism Alignement monolingue avec recherche de déplacements pour la critique génétique

This paper describes the problem of monolingual alignment with search for displaced segments (hereafter called move search). It occurs in genetic criticism, a subfield of literary studies. Existing alignment applications fare poorly on this NP-hard problem. We propose to borrow a family of algorithms from bioinformatics and text algorithmics called fragment chaining alignment. An adaptation of this type of algorithms to NLP is described. Our method enables us to align a pair of texts with move search at the character level. It works for any alphabetic language and achieves scalability. An experimental evaluation presents the good results we obtained compared to other approaches.