Replication package for "Word Segmentation of Separator-Free Labels in UML Design Models for Semantic Proximity Estimation"

This package contains the data, code and reference results needed to reproduce every table and figure of the article "Word Segmentation of Separator-Free Labels in UML Design Models for Semantic Proximity Estimation" (B. Priyambadha, A. Arwan, N. Santoso), submitted to the Journal of Computing Theories and Applications. Data built for this study:- 342 class and attribute labels from the public GenMyModel collection, with reference word segmentations, and their assignment to development and test splits- both annotation records of the 300 candidate labels and the agreement analysis- a 150-label stratified sample drawn from the GenMyModel and Lindholmen collections, with its strata, the two annotators' sheets and the resulting references- the lexicons, abbreviation map and source-code frequency tables used by the segmenter Code and results:- the proposed segmenter and the evaluation harness (Java)- wrappers for the baseline splitters (Samurai, Ronin, INTT, WordNinja, WordSegment)- evaluation, statistical-testing and figure scripts (Python)- per-identifier predictions of every segmenter and the reference outputs A single script (reproduce.sh) regenerates every result and compares it with the reference outputs. The public UML collections (text2uml), the Ludiso oracle and WordNet are not redistributed; setup_env.sh retrieves them at pinned versions. See README.md for details.