LingoBlend

Inside Blend Engine v2: Why We Took Word Selection Away From the AI

LingoBlend's text blending was rebuilt from the ground up: a deterministic selector now chooses which words to blend, and the AI only translates them in context. The result reads better, handles idioms properly, and finishes about twice as fast.

ScienceNikola Artukov6 min read

Blending is LingoBlend's core idea: paste any text in your language, and a chosen percentage of its words come back in the language you're learning, woven into sentences you already understand — the diglot weave method with an AI doing the weaving. For two years, one AI call did everything: it read your text, decided which words to translate, translated them, and wrote the whole passage back out. That worked — but it bundled three very different jobs into one, and each job deserved a better specialist.

The problem with letting one AI pass do everything#

When a single model output carries the entire blended text, three things quietly compete:

Selection is a policy question, not a language question. Which words should be blended depends on things the AI can't know from the text alone — which words you've already mastered, which are due for review, how evenly the blend should spread across the page. A language model guesses; a deterministic algorithm can simply be told.

Re-emitting the whole text invites whole-text mistakes. If the model writes out every word of your document, every word is an opportunity for error — a dropped sentence, a duplicated phrase, a translation glued to the wrong spot. Our old pipeline carried ten repair stages that existed mostly to fix damage this architecture made possible.

Judgment gets diluted. The model's real linguistic skill — knowing that "gave up" is one unit of meaning, that "I would love" must become me encantaría with the pronoun restructured — was spent alongside bookkeeping work it should never have owned.

The new division of labor#

Blend Engine v2 splits the job along its natural seam:

JobOld engineNew engine
Choosing which words to blendAI, in the same passDeterministic selector (frequency data, spacing rules, your mastery — see the vocabulary graph)
Translating in contextAIAI — with the full text as context, unchanged
Handling idioms & fixed expressionsImplicit, inconsistentExplicit contract: expand to the whole expression or skip
Assembling the final textAI re-writes everythingDeterministic code splices translations into your original text
Catching errorsTen repair stages after the factVerified word-by-word against the source before display

The selector chooses candidate words with rules we can state exactly: prefer genuinely useful mid-frequency vocabulary, spread selections evenly through every part of the text, never pick the same word twice in one passage, and — with personalization — skip what you've already mastered. The AI then receives your full text plus that candidate list, and returns only the translations.

The veto: why a smaller job made the AI smarter#

Here's the counterintuitive part. Giving the AI a narrower job made its output better, because we could finally give it explicit linguistic rights:

  • Expand: if a candidate sits inside an idiom, phrasal verb, or fixed expression, translate the whole expression — "kicked" inside "kicked the bucket" comes back as the full idiom, and tapping it in the reader saves the idiom's dictionary form, which is what's actually worth learning.
  • Absorb: if the natural translation swallows neighboring words, take them too. "Write to me" becomes Escríbeme as one unit — never Escríbeme with a stranded "to me" behind it.
  • Skip: if a word doesn't translate cleanly in context — a name, wordplay, a word whose isolated translation would mislead — leave it alone. An untranslated word costs nothing; a wrong translation teaches you something false.

In the benchmark that mattered most to us — 93 trap sentences containing idioms, phrasal verbs, and separable verbs across all 17 source languages — the new engine expanded 65 of them into correct whole-expression translations, against 21 for the old engine. Cross-language pairs produced some of the best results: French posé un lapin rendered into German as sitzen gelassen (stood up), where the old engine offered Kaninchen — a literal rabbit.

Verified before you see it#

Because the final text is now assembled by code rather than written by a model, we can prove properties the old engine could only hope for: every character of your original text survives, translations land exactly where their source words were, repetition loops and echo "translations" are rejected, and constructions that would double-mark grammar (like an English pronoun stranded next to a Spanish clitic) are caught deterministically. If any check fails for a word, that word simply stays untranslated — the failure mode is invisible instead of embarrassing.

One more benefit fell out for free: speed. Since the AI no longer generates your entire document — only the translated words — blends complete in roughly half the time, with the slowest cases improving the most. The blend you used to wait ~6 seconds for now arrives in ~3.

Everything else you know is unchanged: tap any blended word for its grammar context and save the base form or the exact form you met, re-blend any text at a new percentage, and read in the same paginated reader.

Frequently asked questions

What is Blend Engine v2?

It's the new architecture behind LingoBlend's Smart Blend feature. A deterministic selector chooses which words in your text to blend (using frequency data, spacing rules, and your personal mastery), and the AI translates only those words — with your full text as context — instead of rewriting the entire passage. The final blend is assembled and verified by code, word-by-word, against your original text.

Does the new engine handle idioms and phrasal verbs?

Yes — this was the main quality goal. The AI has an explicit contract to expand any candidate that sits inside an idiom, phrasal verb, or fixed expression to the complete unit ("give up", "kicked the bucket", German separable verbs), or to skip it if no clean translation exists. In our 93-sentence idiom benchmark across all 17 languages, v2 translated 65 idioms as whole expressions versus 21 for the old engine.

Is Blend Engine v2 faster?

Yes, about twice as fast on identical documents — roughly 3 seconds instead of 6 for a typical text, with the slowest cases improving the most. The AI now generates only the translated words rather than your whole document, and the new pipeline almost never needs the retries the old one did.

Did the reading experience or the word-saving flow change?

No. The reader, the tap-a-word grammar card, base-form saving, bookmarks, and re-blending all work exactly as before. What changed is what's underneath: better word choice, correct idiom handling, and verified assembly. One visible improvement: expressions are saved as expressions, so tapping estiró la pata offers you the idiom's dictionary form rather than a single misleading word.

Which languages does the new engine support?

All 17 LingoBlend languages, in every direction — including non-English pairs like Spanish→French or Japanese→Chinese. The idiom benchmark deliberately covered every source language, and the engine's guarantees (verified assembly, idiom expansion, skip rights) are language-independent by design.

Share this articleXLinkedInRedditEmail

Nikola Artukov

Builder of LingoBlend. Writes about reading as a way into a language — the methods, the research behind them, and the practical workflows that make them fit into an ordinary week.

More about the author

Related reading

Your Blends Now Skip Words You've Already Mastered

LingoBlend's vocabulary graph has its first consumer: Smart Blend now spends every translated word on vocabulary you're actually learning — skipping mastered words, resurfacing words due for review, and favoring words connected to what you know.

5 min read

Start learning a new language today

Join LingoBlend and turn any text into a personalized language lesson. Free to start, no credit card required.