LinguaFlow - A Natural Command Interpreter with Grammar + LLM Synergy Report for WIF3010 Programming Language Paradigm, Software Engineering Department, Semester 1, 2025/2026

LinguaFlow is an educational, grammar-first natural command interpreter designed to improve transparency and understanding of programming language execution for beginner learners. Developed as part of the WIF3010 Programming Language Paradigm course (Semester 1, 2025/2026) in the Software Engineering Department, the project addresses the common “black box” perception of interpreters by exposing how natural language input is transformed into structured computation through formal language mechanisms. The system supports multiple syntactic forms for expressing arithmetic computations, including symbolic notation, infix worded operators, functional list-based expressions, and conversational natural phrasing. To manage the ambiguity inherent in human language, LinguaFlow integrates a constrained Large Language Model (LLM) as a semantic resolution layer. The LLM is strictly limited to mapping unfamiliar or synonymous operation words (e.g., add, accumulate, aggregate) to one of four arithmetic operators, ensuring deterministic execution while preserving natural language flexibility. LinguaFlow follows a classical interpreter architecture consisting of lexical analysis, parsing, and interpretation stages, all governed by a formally defined EBNF grammar. An optional verbose mode exposes intermediate artifacts such as token streams, abstract syntax trees, and evaluation traces, making the internal execution process explicit and suitable for educational use. This repository contains the complete project report, including design rationale, grammar specification, implementation details, testing methodology, and reflections on limitations and future improvements.