Structural Alignment as the Basis to Improve Significant Change Detection in Versioned Sentences

Structural Alignment as the Basis to Improve Significant Change Detection in Versioned Sentences Tan Ping Ping, Karin Verspoor and Timothy Miller Department of Computing and Information Systems University of Melbourne Victoria 3010, Australia. {pingt@student., karin.verspoor@, tmiller@}unimelb.edu.au Abstract Some revisions of documents can change the meaning of passages, while others merely re-phrase or improve style. In a multi-author workflow, assisting readers to assess whether a revision changes mean- ing or not can be useful in prioritising re- vision. One challenge in this is how to de- tect and represent the revision changes in a meaningful way to assist users in assess- ing the impact of revision changes. This paper explores a segmentation approach which utilises the syntactic context of re- visions to support assessment of signifi- cant changes. We observe that length of normalised edit distance or Word Error Rate (WER) correlates better to the signif- icance of the revision changes at sentence level compared to general sentence simi- larity approaches. We show that our pro- posed method, SAVeS, supports improved analysis of change significance through alignment of segments rather than words. SAVeS can be used as the basis for a com- putational approach to identify significant revision changes. 1 Introduction Revision of documents is a common component of the writing process. In this work, we introduce an approach to analysing revisions that will support the identification of significant changes, such that attention can be focused on revisions that impact meaning. We define a versioned text as a text document that has been revised and saved to another ver- sion, where the original version is directly avail- able for comparison. An edit is defined as change that involves operations such as insertion, dele- tion or substitution of characters or words within a revised text. We define a significant change between versioned texts as a meaning altering change, which goes beyond string edit operations. Faigley and Witte (1981) proposed a taxonomy to assist in evaluating the effect of revisions on meaning (Figure 1). They identify a range of revi- sion types. On a general scale, they define surface changes as edits that improve readability without actually changing the meaning of the text, and text-base changes as edits that alter the original meaning of the text. These categories are sub- divided. The subcategories for surface changes: formal changes includes copy editing operations such as correction in spelling, tense, format, etc., while meaning preserving changes includes re- phrasing. For text-base changes, microstructure changes is meaning altering changes which do not affect the original summary of the text and macrostructure changes are major changes which alter the original summary of the text. Although they provided some examples, the definitions are insufficient for computational implementation. Framed by this taxonomy, we consider sig- nificant change to be a macro-structure revision change while a minor meaning change is a micro- structure revision. We adopt surface revision change to be no meaning change. Based on one original sentence, we provide examples of how we distinguish between meaning-preserving, micro- structure …

Content truncated to 500 words