Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Journal of Innovation Information Technology and Application (JINITA)
Mardi Utomo
Universitas Stikubank
Computer Science & IT Decision Sciences, Operations Research & Management Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Evaluating Cross-Language Structural Generalization of the Unified Abstract Syntax Tree Mardi Utomo; Ema Utami; Kusrini; Arief Setyanto
Journal of Innovation Information Technology and Application (JINITA) Vol 8 No 1 (2026): JINITA, June 2026
Publisher : Politeknik Negeri Cilacap

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35970/jinita.v8i1.3191

Abstract

Cross-language code analysis requires syntax-aware representations that can reduce language-specific syntactic variation while preserving comparable program structure. Although unified Abstract Syntax Tree (AST) representations have been proposed, empirical evidence on their representation-level structural behavior across datasets and programming languages remains limited. This paper evaluates the structural generalization of a Unified AST representation as a schema-level abstraction, not as a parser-free or full semantic equivalence mechanism. The Unified AST schema is constructed from the CodeXGLUE code-to-text dataset covering Python, PHP, Ruby, Java, JavaScript, and Go. Its generalization is then examined on function-level aligned benchmarks from CodeXGLUE code-to-code translation (Java-C#) and multilingual HumanEval (Java, JavaScript, Go, Python, C++, and Rust). Tree Edit Distance (TED) similarity is used as the primary structural metric, while cosine similarity, BLEU, compression ratio, and identifier precision-recall are treated as auxiliary indicators of lexical similarity, reconstruction fidelity, compactness, and identifier retention. The results show an average TED similarity of 0.77 on CodeXGLUE code-to-code translation and 0.60 on HumanEval. These findings indicate that the Unified AST can preserve cross-language structural patterns under aligned benchmark assumptions, although it does not prove behavioral equivalence and remains dependent on language-specific parsing during AST extraction
Co-Authors Arief Setyanto Ema Utami Kusrini