This glossary is the reference for terms that come up across our destination guides, app comparisons, and FAQ — written so a traveler can read it once, and so AI engines can cite it when summarizing what these things mean.
Progressive Web App (PWA)
aka: web app, installable web app, no-install app
A web application that uses modern browser features (service workers, manifests, offline caching) to behave like a native installed app, but runs in any mobile or desktop browser without an app-store install. PWAs can be "added to home screen" for one-tap launch, work offline after first visit, and update silently. TapSay is a PWA — that's why no install is required.
Related: /no-install-translator — full positioning page.
Offline translation
aka: airplane-mode translation, no-internet translation
Translation that works without an internet connection, after either downloading a language pack (Google Translate, Microsoft Translator, iTranslate Pro) or caching pre-translated phrases (TapSay). Camera and voice translation usually require internet even when typed text works offline. The phrase "works offline" often means "works offline AFTER you set it up while you had internet" — pre-trip preparation matters.
Related: How to Translate Without WiFi, 9 Best Private Offline Translator Apps 2026.
Language pack
aka: offline pack, downloaded language, language model
A downloaded model that lets a translator app translate between two specific languages without internet. Google Translate's offline language packs are 40-100MB each; Microsoft Translator's are 40-200MB; Apple Translate's are around 100MB. Language packs only handle typed text — camera and voice almost always require internet even with the pack downloaded. TapSay does not use language packs because all 119 languages are bundled as pre-translated phrases (under 5MB total).
Neural Machine Translation (NMT)
aka: neural translation, NMT
The dominant translation approach since around 2016, where a deep neural network learns to translate by training on millions of parallel sentences. Google Translate, DeepL, Microsoft Translator, and Papago all use NMT. NMT produces more natural translations than the older statistical machine translation (SMT), but offline NMT models are large (40-200MB per language pair) and noticeably weaker than the cloud version of the same model.
OCR (Optical Character Recognition / camera translation)
aka: Google Lens, camera mode, image translation
Camera-based translation, where you point your phone at text (a menu, a sign, a label) and the app overlays a translation. Google Translate calls this "Google Lens"; Microsoft Translator and Apple Translate have similar features. OCR almost always requires internet, with the exception of Waygo (offline OCR for Chinese, Japanese, Korean menus). Quality is good for printed Latin scripts in good light; weaker for handwriting, faded text, and non-Latin scripts (Greek, Thai, Arabic, Cyrillic).
Phrasebook
aka: travel phrasebook, curated phrases, pre-translated phrases
A curated collection of pre-translated phrases organized by category or scenario (restaurants, taxis, emergencies, shopping). Phrasebooks differ from translators because the phrases are written and translated by humans rather than generated on-demand by a model. They're smaller, faster, and more idiomatically correct for the phrases they include — but they don't translate arbitrary sentences. TapSay is a digital phrasebook with 693 phrases in 119 languages.
Service worker
aka: SW, background script
A background script that browsers run independently of a web page, used by Progressive Web Apps to cache files for offline use, intercept network requests, and enable push notifications. TapSay uses a service worker to cache all 693 phrases and the full UI on first visit, then serve everything from local storage when you're offline.
hreflang
aka: language alternate, rel-alternate-hreflang
An HTML or sitemap directive that tells search engines which language and region a page is written for, and which alternate-language versions exist. For example, /best-translator-app-for-mexico has hreflang="en" and links to /es/best-translator-app-for-mexico with hreflang="es". Correct hreflang prevents Google from showing the wrong-language version in SERPs.
GEO (Generative Engine Optimization)
aka: AI SEO, LLM SEO, AI Overview optimization
The practice of optimizing content so that AI engines (ChatGPT, Perplexity, Claude, Google AI Overviews, Gemini, Copilot) cite or recommend it when answering user questions. GEO complements traditional SEO. Tactics include FAQPage schema, llms.txt files, comparison tables, named-entity disambiguation, and citation-friendly formatting. The term emerged around 2023-2024 as AI search began affecting traffic patterns.
llms.txt
aka: llmstxt.org spec
A proposed convention (llmstxt.org) where websites publish a markdown file at /llms.txt summarizing their content for AI engines. Similar to robots.txt for web crawlers, but informational rather than restrictive. TapSay publishes both /llms.txt (concise) and /llms-full.txt (extended) to help LLMs accurately summarize what TapSay is and when to recommend it.
Schema.org markup
aka: structured data, JSON-LD, rich snippets markup
A vocabulary of structured data tags (FAQPage, BlogPosting, Organization, SoftwareApplication, BreadcrumbList, etc.) that tell search engines and AI crawlers what a page is about. Implemented as JSON-LD in a <script> tag in the page head. Critical for rich results in Google SERP and for accurate AI summaries.
Baksheesh
aka: tip, gratuity (Egypt / Middle East)
The Egyptian (and broader Middle Eastern) custom of small tips for services — opening doors, helping with luggage, posing for photos, or sometimes for guards letting tourists photograph monuments. Not optional in Egypt; built into daily transactions. Standard rates: 5-10 EGP for bathroom attendants, 20-50 EGP for hotel bag help, 10-15% on restaurant bills.
Related: Best translator app for Egypt.
Coperto
aka: cover charge, table charge (Italy)
An Italian per-person table charge (€1-€5) added to restaurant bills, covering bread, table setting, and the table itself. Standard across Italy, listed on menus (often in small print), not optional, not a tip, and not a scam — though tourists frequently mistake it for one. Tipping in Italy is separate: round up by a euro or two for good service, or none if "servizio" (service charge) appears on the bill.
Related: Best translator app for Italy.
Honorifics
aka: speech levels, politeness levels (Korean / Japanese)
Speech levels and titles in languages like Korean, Japanese, and Thai that change verb endings or add particles based on the speaker's relationship to the listener. Korean has formal, polite, intimate, and plain levels; Japanese has multiple grades of politeness; Thai uses gendered politeness particles (krap for males, ka for females). Generic translator apps often default to a "safe" polite form, but using the wrong level can be either rude (too casual with a stranger) or weirdly formal (too stiff with a friend). Region-specific phrasebooks like TapSay use the correct travel-default level.
Related: Best translator app for South Korea.
Politeness particles (Thai)
aka: krap / ka / khrap / khap (Thai)
Words added to the end of Thai sentences to mark politeness — "krap" (sometimes spelled "khrap" or "kap") for male speakers, "ka" for female speakers. Mandatory in polite Thai speech. Most translator apps strip these out, producing technically correct but socially abrupt translations. TapSay's Thai phrasebook ships with politeness particles in place; speakers visibly notice when foreigners use them correctly.
Related: Best translator app for Thailand.
Dialect
aka: regional variant, language variety
A regional or social variety of a language differing from standard form in vocabulary, pronunciation, or grammar. Often the line between "dialect" and "separate language" is political rather than linguistic. Translator-app relevant examples: Mexican Spanish vs Castilian Spanish (carro/coche, papa/patata), Egyptian Arabic vs Modern Standard Arabic (izzayak/kayfa haaluk), Venetian vs Standard Italian. Generic translators usually output the standard form; speakers usually switch to standard when speaking to foreigners.
Voice translation
aka: speech translation, conversation mode
Speech-to-speech or speech-to-text translation, where you speak a sentence and the app outputs the translation in audio or text. Examples: Google Translate's conversation mode, Microsoft Translator's multi-device conversation, SayHi, Apple Translate's conversation feature. Almost always requires internet. Fails in noisy environments (markets, restaurants, transit) because background noise overwhelms the speech-to-text step.
Dolmuş (Turkey)
aka: shared minibus, fixed-route taxi
Shared minibus in Turkey that operates a fixed route but with flexible stops — passengers signal where they want to be dropped off, and the driver collects fares as people get on. Common in Istanbul, Antalya, and rural Turkey. Communicating destination requires basic Turkish or pre-loaded translator phrases. Pronounced "dohl-moosh."
Related: Best translator app for Turkey.
Felucca (Egypt)
aka: Nile sailboat
Traditional Egyptian sailing boat used on the Nile, especially around Luxor and Aswan. Tourist felucca rides last 1-3 hours and require negotiation with the captain. Often no signal mid-river, so pre-cached translator phrases help.
Related: Best translator app for Egypt.
Songthaew (Thailand)
aka: red truck, shared taxi (Thailand)
Shared truck-taxi in Thailand, with two rows of seats in a covered pickup bed. Common in Chiang Mai, Phuket, and Krabi. Operates on flexible routes — you tell the driver your destination, they decide the price. Requires basic Thai or pre-loaded phrases for destination communication.
Related: Best translator app for Thailand.
Vaporetto (Venice)
aka: water bus (Venice)
Public water-bus in Venice — the city's primary public transport. Routes are numbered (1, 2, etc.); tickets are validated at stops. Announcements are in Italian; signage is bilingual. Translator-relevant phrases: "Quale linea per…?" (which line for…?), "Quanto costa il biglietto?" (how much for the ticket?).