Lachmannian Archetype Reconstruction for Ancient Manuscript Corpora

Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL, pages 1209–1214, Denver, Colorado, May 31 – June 5, 2015. c⃝2015 Association for Computational Linguistics Lachmannian Archetype Reconstruction for Ancient Manuscript Corpora Armin Hoenen Goethe University Robert- Mayer Strasse 10 60486 Frankfurt, Germany hoenen@em.uni-frankfurt.de Abstract Two goals are targeted by computer philol- ogy for ancient manuscript corpora: firstly, making an edition, that is roughly speaking one text version representing the whole cor- pus, which contains variety induced through copy errors and other processes and secondly, producing a stemma. A stemma is a graph- based visualization of the copy history with manuscripts as nodes and copy events as edges. Its root, the so-called archetype, is the supposed original text or urtext from which all subsequent copies are made. Our main con- tribution is to present one of the first com- putational approaches to automatic archetype reconstruction and to introduce the first text- based evaluation for automatically produced archetypes. We compare a philologically gen- erated archetype with one generated by bio- informatic software. 1 Introduction In philology, oftentimes more than one single manuscript of the same tradition (that is, the same text) has survived. These manuscripts often differ in their wording in various places since copy errors, corrections, and other processes have led to devia- tion from the original text. This causes two prob- lems: uncertainty about the original wording and uncertainty about which manuscript has been copied from which other. The reconstruction of the copy history of manuscript texts is largely similar to that of DNA, which is why phylogenetic approaches have been adopted (Robinson and O’Hara, 1996; Robinson et al., 1998; van Reenen et al., 1996; van Reenen et al., 2004; Spencer et al., 2004; Roos and Heikkil¨a, 2009; Roelli and Bachmann, 2010; Andrews and Mac´e, 2013). However, the main goal of the philo- logical work on ancient manuscripts is not the re- construction of the copy history but compiling an edition of a historical text. This entails reducing the variation encountered so that one single main text as the prototypical representation of the manuscript corpus emerges. Ideally, this text is believed to be the author’s original or closest possible to it. Two model approaches to making an edition are most widespread. The earlier one by Lachmann (see for instance Lachmann (1853)) opts for reconstructing an urtext actively, that is if needed by means of emendation, which refers to inferring the original (authorial) wording from the extant variants even if the so-inferred form is not itself extant and thus not attested in any of the manuscripts. The later ap- proach after B´edier (see for instance B´edier (1928)) bases the edition directly on the text of the best avail- able manuscript. In this paper, we will present a first automatic im- plementation of the earlier approach. Additionally, an algorithm for evaluation of a so-reconstructed text will be presented and applied to artificial bench- mark data sets (gold standard). First, we will in- troduce the data sets. …

Content truncated to 500 words