Text Alignment Discussion

We are deeply indebted to the work of Professor Gregory Crane, David Bamman and the Dynamic Lexicon project at Tufts University for the ability to provide the Alpheios functionality which aligns words in the sample Greek and Latin Alpheios Enhanced texts with their closest equivalents in the corresponding English translation. The alignment we are currently using was produced by manual adjustment by students of the automatic alignment produced by Tufts and will certainly be subject to further revision. We expect, in fact, to provide an interface to permit users to produce their own alignment and discuss alignment issues.

In working through the process of aligning the texts, a number of interesting issues came to our attention. We found ourselves needing to to make a choice between a hyper-literal alignment, where words are aligned only to those in the translation which are fairly literal matches of vocabulary, morphology and syntax; or a more broadly contextual alignment, in which words are also linked to the implied meaning, aligning as many of the original words as possible with English words which the translator used to convey them. The latter approach was what we chose for the current release, as the functionality we are trying to demonstrate in this release benefits from having as many words as possible matched with English equivalents.

It also became clear that the ability to see alignments with multiple different translations simultaneously would be extremely advantageous to the student, as it sheds light on both the language itself, and the evolution of the language of the translation.

We hope in a future release to be able to present the student with multiple approaches to alignment, as well as multiple aligned translations simultaneously. In addition, we plan to provide tools which students can use to explore text alignment exercises on their own. We have developed a prototype interface for this which has proved invaluable to our process, and upon which we hope to build.