# Building IdiomX: language beyond the literal Canonical HTML: https://dsti.school/techblog/idiomx-multilingual-idiom-understanding This Markdown copy is generated from the same DSTI static-site build as the canonical HTML page. It is intended for machine readability and concise retrieval. [DSTI TechBlog](https://dsti.school/techblog)  /  Students Students Student workbench · public contribution “Spill the tea” is not an instruction involving a drink. DSTI student Ayman Ali Sharara built IdiomX to test whether multilingual AI systems can detect idiomatic usage, retrieve figurative expressions and explain meaning across English, Arabic and French. AS Ayman Ali Sharara DSTI student · MSc in Data Science & AI · Online asynchronous 26 May 2026 13 min read Deep Learning with Python project idiomx multilingual-nlp figurative-language semantic-retrieval english-arabic-french student-project ## One expression. Meaning beyond the words. Observed expression “Spill the tea” Idiomatic interpretation EN Reveal gossip or personal secrets. AR كشف الشائعات أو الأسرار FR Révéler des potins. 190K+ contextual examples 12K+ idioms 3 aligned languages 4 benchmark tasks Idioms expose a gap between recognising words and understanding language. Their meanings are often non-compositional: knowing every individual word is not enough to recover what the expression means in context. A system can parse “break a leg” perfectly and still mistake encouragement for injury. AS A student project built for public scrutiny IdiomX began within the DSTI course Deep Learning with Python , supervised by Pr Hanna Abi Akl. Ayman completed the work as a student in the [MSc in Data Science & AI](https://dsti.school/msc-in-data-science-and-ai), studying through the [Online asynchronous](https://dsti.school/online-studies) mode. The resulting dataset, construction pipeline, models, notebooks, paper and demonstrations are publicly accessible. ## 01 When language stops being literal Humans routinely distinguish literal and figurative readings from context. “She spilled the tea on the table” describes an accident; “she spilled the tea about the meeting” describes disclosure. That distinction remains difficult for natural-language systems because lexical overlap can be almost identical while the meaning changes completely. The original project started from a practical limitation: many idiom resources are modest in scale, narrow in language coverage or designed around one isolated task. IdiomX instead treats idiom understanding as a sequence of related problems—from recognising figurative usage to retrieving and explaining meaning across languages. The design question: can one reproducible resource support detection, semantic retrieval, cross-lingual alignment and interpretable meaning retrieval rather than evaluating each capability in isolation? ![IdiomX project cover illustrating multilingual idiom understanding](https://media.dsti.school/wp-content/uploads/2026/05/25102835/IdiomX_Cover.avif) > **Figure caption:** IdiomX frames figurative-language understanding as a multilingual data, modelling and evaluation problem. ## 02 Building a multilingual dataset at scale The public release contains more than 190,000 contextual examples spanning over 12,000 idioms. English expressions are linked to Arabic and French semantic representations, together with idiomatic, literal and borderline usage labels and supporting linguistic metadata. ≈196K rows in the current full Hugging Face dataset 12K+ unique idioms represented EN · AR · FR English, Arabic and French semantic alignment ≈1.04 reported sentence-reuse factor after cleaning The construction process combines lexical resources, controlled generation and validation. Its modular structure matters as much as its size: the objective is to make every stage inspectable and repeatable rather than publishing an opaque final file. 1 ### Collection Extract candidate idioms from sources including Wiktionary-derived data and WordNet, while extending coverage with modern and generated candidates. 2 ### Cleaning and normalisation Filter noise, standardise expressions, remove duplicates and prepare consistent records for enrichment and evaluation. 3 ### Controlled LLM enrichment Use GPT-4.1-mini to generate meanings, contextual examples and aligned English, Arabic and French semantic fields. 4 ### Structured validation Combine semantic-similarity scoring, rule-based checks, deduplication and leakage-aware splits to support reliable benchmarking. ![IdiomX data preparation pipeline](https://media.dsti.school/wp-content/uploads/2026/05/25135940/IdiomX_Data_perep_Pipeline_v2.avif) > **Figure caption:** The data-preparation workflow moves from heterogeneous lexical sources to normalised, enriched and validated examples. ## 03 More than a dataset: one progressive benchmark IdiomX is organised as a progression. The first task asks whether a model recognises figurative usage. The later tasks ask whether it can retrieve an appropriate idiom from context, align meaning across languages and return an explanation that a person can inspect. ![Full IdiomX dataset and benchmark pipeline](https://media.dsti.school/wp-content/uploads/2026/05/25140144/IdiomX_full_pipeline_V1.avif) > **Figure caption:** The complete workflow joins dataset construction, model training, retrieval benchmarking, multilingual interpretation and deployment-ready artefacts. ## 04 Four tasks, from recognition to interpretation Task 1 ### Idiom detection Determine whether an expression is being used idiomatically or literally within its sentence. Compared TF-IDF with Logistic Regression, DistilBERT and RoBERTa Reported best RoBERTa Capability Contextual disambiguation Task 2 ### Context-to-idiom retrieval Given a contextual sentence, rank the idioms that best express its underlying figurative meaning. Compared Dense, lexical and hybrid retrieval with reranking Reported best Hybrid retrieval with a fine-tuned reranker Capability Semantic retrieval Task 3 ### Arabic-to-English retrieval Use an Arabic context to retrieve the corresponding English idiom, testing semantic alignment across languages. Compared Multilingual MiniLM, multilingual E5 and fine-tuned E5 Reported best Fine-tuned E5 Capability Cross-lingual alignment Task 4 ### Idiom interpretation Retrieve the canonical idiom and explain its meaning in English, Arabic and French. Compared Dense and hybrid retrieval, with and without reranking Reported best Hybrid retrieval with reranking Capability Explainable semantic grounding ## 05 What the benchmark experiments found The project reports that contextual transformers substantially improve idiom detection, while hybrid lexical–dense retrieval outperforms dense retrieval alone. Fine-tuning is particularly important for the Arabic-to-English task, where surface forms provide little direct lexical help. Task | Reported leading configuration | Main result Detection | RoBERTa | 92.6% accuracy · F1 0.926 Context → idiom | Hybrid retrieval + fine-tuned reranker | Top-1 88.5% Arabic → English idiom | Fine-tuned E5 | Top-1 57.8% Interpretation | Hybrid retrieval + reranker | Top-1 67.4% The figures are not interchangeable: each task tests a different search space and difficulty. Their combined value is the progression from classification toward multilingual retrieval and interpretable output. ### A Task 4 output is designed to be readable Input Spill the tea Canonical meaning Reveal gossip or personal secrets Multilingual output English, Arabic and French explanations ## 06 Why figurative-language understanding matters Idioms are not edge cases confined to dictionaries. They appear in conversation, support requests, social media, subtitles, teaching material and everyday instructions. Systems that interpret them literally can misunderstand intent even when every individual token is familiar. Conversational AI Chatbots and assistants that better recognise what users actually mean. Translation Systems that seek an equivalent meaning rather than translating word by word. Language learning Tools that retrieve explanations, contexts and cross-language equivalents. Semantic search Retrieval based on intended meaning rather than surface-form overlap. Content analysis Improved treatment of slang, figurative language and context-dependent usage. Human–robot interaction Interfaces that cope more reliably with natural, culturally situated speech. ### What remains difficult - Some examples are LLM-generated, so controlled enrichment does not remove the need for critical review. - Idiomatic interpretation can legitimately vary by context, culture and register. - Open-ended inputs may retrieve a related idiom rather than the exact intended expression. - Cross-lingual retrieval remains materially harder than monolingual detection. ## 07 The project is open: data, code, paper and demonstrations IdiomX separates the dataset-construction pipeline from the modelling and benchmark repository. This makes the provenance of the resource easier to inspect while keeping the task notebooks, trained artefacts and demonstrations organised around evaluation. AS ### Ayman Ali Sharara DSTI student in the MSc in Data Science & AI, studying through Online asynchronous. His work spans multilingual NLP, data engineering, retrieval systems and practical AI applications. IdiomX was developed as his Deep Learning with Python project. [MSc in Data Science & AI](https://dsti.school/msc-in-data-science-and-ai)[Online asynchronous](https://dsti.school/online-studies) [LinkedIn](https://www.linkedin.com/in/ayman-sharara/)[GitHub](https://github.com/aymanshar)[Hugging Face](https://huggingface.co/aymansharara) Article adapted for the DSTI TechBlog from Ayman Sharara’s original student-project contribution and the project’s current public documentation. The writing and presentation have been revised while preserving the project’s methods, claims and reported results.