Frequently Asked Questions

TEP stands for translation, editing and proofreading. At El Turco a native linguist translates the text, a second native linguist edits the translation against the source and a final proofreading pass checks the text as readers will see it. TEP is the standard El Turco workflow for contracts, manuals, product content and other material where accuracy matters. The workflow follows ISO 17100:2015 for translation services.

Translation converts text from one language into another. Localization adapts a whole product for a market: text, date and number formats, currencies, units, layout direction, images and cultural references. El Turco treats software, websites, games and online stores as localization projects. Linguists check strings in context and flag anything that would read as foreign to local users before the product ships.

Transcreation is the creative rewriting of marketing copy so it has the same effect in another market as the original has at home. El Turco uses transcreation for slogans, taglines, ad copy and campaigns, where a literal translation would lose the brand voice. Native linguists with copywriting experience adapt the message, a second linguist reviews it and a back-translation shows the client what the new copy says.

MTPE, or machine translation post-editing, means a native linguist edits machine translation output until it meets an agreed quality level. El Turco runs MTPE under ISO 18587:2017. MTPE suits high-volume, repetitive content such as product catalogs, support articles and documentation. Marketing copy, legal text and content with heavy cultural nuance are better served by full human translation with a second linguist review.

Linguistic quality assessment (LQA) is a structured review of finished translations against an agreed error typology and severity scale. El Turco offers LQA on its own work and on translations from other vendors, with a documented methodology and a report that logs each issue by category. Companies use LQA to evaluate vendors, audit existing content and test localized software or games in context.

Continuous localization translates new and changed content as it is released instead of in large batches. El Turco connects to a client repository, TMS or CMS so new strings are picked up, translated, reviewed and delivered back on a recurring rhythm. Hub, our in-house middleware, links the client TMS with Dispatch, our business management system. Projects then move without manual handoffs.

Yes. El Turco creates and maintains a glossary and a style guide for each client and target language, usually at the start of a program. The glossary fixes product names, key terms and words that stay in English. The style guide sets tone, formality, punctuation and locale conventions. Linguists reuse both on every project, so the voice stays consistent across releases.

El Turco translates common document, web and software formats, including string files such as JSON, XLIFF, PO and RESX, plus content exported from CMS and e-commerce platforms. Files are processed in CAT tools such as RWS Trados Studio, memoQ and Phrase, which keep tags, variables and placeholders protected. For an unusual format, send a sample with the quote request so the project manager can confirm the workflow.

Every El Turco translation is reviewed by a second native linguist under ISO 17100 and then runs through What's Wrong?, our own translation QA tool. Deterministic rules catch the certain issues instantly. An AI judge re-reads only the risky segments, and a human always signs off. Xbench and Verifika checks are added when a client requires them, alongside LQA reports on request.

Yes. El Turco offers localized SEO as part of website and e-commerce localization: local keyword research, localized meta titles and descriptions, URL structure localization and hreflang implementation support. Native linguists research the terms people actually search for in each market, because a direct translation of an English keyword often misses local search behavior on regional search engines and marketplaces.

El Turco provides speech data collection, text annotation, RLHF and human feedback, content moderation datasets, image labeling and custom MT and LLM training data. Native speakers collect or annotate all data in Turkish, Arabic, Persian, Balkan, Eastern European and Turkic languages, including dialects where off-the-shelf datasets are scarce. AI and ML teams use the data for speech recognition, LLM alignment, moderation and machine translation.

El Turco delivers text data as JSONL, CSV/TSV, XLIFF or a custom schema. Speech recordings ship as WAV or FLAC files sampled between 16 and 48 kHz, with metadata on each speaker and recording environment. Time-aligned transcripts come as TextGrid, SRT or VTT. RLHF ranking and rating exports match the client pipeline format. Delivery runs through secure transfer, API or the client's own platform.

El Turco controls AI data quality with annotation guidelines co-developed with the client and versioned per batch, calibration rounds before production and gold tasks seeded throughout the work. Each annotator is tracked against the gold tasks. Inter-annotator agreement is measured and reported per delivery, and transcription gets word error rate (WER) spot-checks against acceptance criteria agreed up front.

Yes. El Turco delivers a free pilot batch of up to 100 annotated samples or 30 minutes of collected audio, with a QA report included. The AI team sends its specification, evaluates the pilot against its own acceptance criteria and then decides whether to continue. Pilots are requested through the contact form with the AI data option selected.

El Turco sources speech and text contributors ethically, and contributors give consent before any collection starts. Data is handled in a GDPR-aligned way under an ISO/IEC 27001 certified information security management system. Usage terms for each dataset are set out in the project agreement, so the client knows how the delivered data can be used for model training.

Custom MT and LLM training data is a set of parallel corpora, domain-specific datasets and evaluation sets built on a client's own content and terminology. El Turco linguists prepare the data and benchmark engine output, so a team can fine-tune a machine translation engine or a large language model for its domain and measure the result in each regional language.

RLHF, or reinforcement learning from human feedback, uses human judgments to improve how a language model responds. In El Turco RLHF projects, native linguists rank LLM responses, score factual accuracy, evaluate cultural appropriateness and evaluate prompts. Teams use the data for alignment, instruction tuning, chatbot quality and safety guardrails in Turkish, Arabic, Persian and the other languages El Turco covers.

Yes. El Turco collects and annotates Arabic data in Gulf, Levantine, Egyptian and Maghrebi dialects as well as Modern Standard Arabic. Dialect coverage matters because models trained mainly on MSA miss how people actually speak and write online. Native annotators from each dialect area label the data, so slang and code-switching are tagged correctly.

Yes. El Turco offers image labeling with bounding boxes, polygons, segmentation masks and classification labels for images and video frames. Native-speaking annotators handle regional signage, documents, receipts and products, including Arabic and Persian script and Turkish characters that generic annotation teams misread. Typical uses are retail product recognition, document OCR, street-level imagery and image content moderation.

An AI data project with El Turco starts with requirements analysis: model needs, target languages, dialects, data specifications and volume. The project team then recruits native contributors who match the profile, drawing on 700+ vetted linguists and contributors. Collection and annotation run with layered quality control, followed by a final human review and formatting that fits the client pipeline.

Most written business content uses Modern Standard Arabic, which readers across the Arab world understand. Marketing, games, chat interfaces and speech data often work better in a dialect such as Gulf, Levantine, Egyptian or Maghrebi Arabic, chosen by target market. El Turco recommends the variety at scoping, and the Arabic language page explains the trade-offs for each content type.

Speakers of Serbian, Croatian and Bosnian understand one another, yet each standard language follows its own national norms for vocabulary and spelling. Croatian is written in Latin script. Serbian uses both Cyrillic and Latin. Bosnian is mostly written in Latin. Serbian in Serbia uses ekavian forms where Croatian and Bosnian use ijekavian forms. El Turco assigns native linguists for each language.

Persian and Farsi are two names for the same language: Farsi is the name used in Iran and Persian is the usual English name. Dari is the standard variety of Persian used in Afghanistan, with its own vocabulary and conventions. El Turco translates into Iranian Persian by default and confirms the target variety at scoping when content is meant for Afghan readers.

El Turco translates into both main Kurdish varieties. Kurmanji, used by Kurdish speakers in Türkiye, Syria and Europe, is written in a Latin-based alphabet. Sorani, used mainly in Iraq and Iran, is written in an Arabic-based alphabet that runs right to left. El Turco confirms the variety and script with each client before work starts, because one translation cannot serve both audiences.

El Turco agrees the script for Serbian, Kazakh and Uzbek with each client at scoping, because the right choice depends on market and content. Serbian Cyrillic is the official script in Serbia, while Latin is widely used online. Kazakh content is still mostly written in Cyrillic as Kazakhstan moves toward a Latin alphabet in stages. Uzbek uses Latin officially, although Cyrillic remains common.

Arabic, Persian and Sorani Kurdish are written right to left, so localized layouts must mirror navigation, alignment, icons and progress indicators. Numbers, Latin product names and URLs inside right-to-left text still run left to right, which breaks punctuation and cursor behavior when bidirectional text is mishandled. El Turco linguists review right-to-left strings in context and log layout issues during LQA.

No. Turkish and Azerbaijani are closely related Turkic languages, but they are not interchangeable. The two languages differ in vocabulary, spelling and everyday expressions. Some words look the same but mean different things. Azerbaijani also uses letters such as ə that Turkish does not have. El Turco assigns native Azerbaijani linguists to Azerbaijani content, never Turkish linguists adapting Turkish text.

El Turco covers 19 languages across Türkiye, the Middle East, the Balkans, Eastern Europe and Central Asia, each with its own page under Languages. Requests for other regional languages are reviewed case by case. Send the language, target market and content type through the contact form. The project manager confirms whether vetted native linguists are available before quoting.