Zero beginner​

• Memorize elementary sentences with high-frequency material words (like bread or rice).

  • Use a tool like a good anki deck.

• To start memorizing in multiple modalities, you need to do these:

  • Understand the writing system.
    • If it is a sibling alphabet (like Czech to English or even Bulgarian or Greek to English), focus on the differences.
    • If it is a completely different alphabet (like Thai to English, or English to Chinese), use a tool or anki deck to train the letters. Ensure it does not only present you the letter and its sound, but also words where a letter goes in different positions.
    • If it is a completely different and bigger writing system (like Chinese to English), accept that you cannot learn it quickly and study the strategy of acquiring the characters over years. But also learn the practical basics, like (pinyin for Chinese)
    • Install a keyboard to your devices so you are able to practice writing.
  • Understand the sounds.
    • In some cases like Chinese for English you should train listening and pronunciation first to unlock your progress. Some tools, like Duolingo, combine it with memorizing sentences and that's good. But it is still a separate area of your focus.
    • In some cases you can ignore the differences in pronunciation (like Italian for English) and still progress well. So, do not put too much efforts in the beginning, but see the important remark about accent training below.

• Do minimal exercises for writing or speaking.

  • If your tool allows them - that's all you need.
  • Otherwise, just translate some sentences of the words you know. Use AI chat or a teacher to check you. When using AI do not forget to tell it your current level in the language.

• Accent training

  • There is a problem, that accent training is considered as a very niche need for advanced speakers. It is supported by the fact that lots of people reach practical fluency while keeping harsh foreign accent. This is true but it does not mean that cutting this edge saves you effort. Here's the contradiction:
    • If you do not training one of the native accents (ie. the correct sounds, stress positioning and natural rhythm), consuming content in the target language takes more cognitive efforts from you. This definitely damages productivity and easiness of learning.
    • But in the other hand, you definitely should not aim to 'close the task quick' in case of training the correct accent. You need to significantly rebuild your brain and muscle memory, so don't you even expect it will take less than 6 months.
  • The good part is that accent is a mechanical skill, so spending literally 5 minutes a day is enough. The rest of the time, brain plays with these experiences, so even after an interruption of a week or more, you often find yourself more proficient in this skill.
  • I'd suggest not to train accent in listening modality specifically:
    • You unlock the ability of your brain to notice the nuances when you training pronunciation.
    • This affects how many nuances you recognize when you do other comprehension exercises, so no dedicated training is required.

Not a beginner but still unable to understand the target language​

• The main goal for the learner: the ability to understand without translating.
• Words to absorb: ~2500 high frequency, topic-wise.
  • I saw estimations of 100 (a hundred) words. Well, I have reasonable doubts because I saw frequency vocabularies. With some manipulations over the definitions we can make someone speak after learning 100 words. We must not count prepositions and other helper words as words (consider them grammar) and we must pick 100 of the most probable semantic words for the situation where a person will talk (like a visit to a cafe and asking about the cakes' properties). But why would we do all this analytical work otherwise than to take a single entertaining YouTube video? It is not practical to be considered a part of strategy for everyone.
  • If not 100, then can it be 1000 words though? Idk. Maybe.

All levels, starting from this one​

You should be memorizing sentences. But not random, they should be:

• Selected to address your grammar or vocabulary gaps.
• They should be trained from your language (or a picture) to recalling the sentence in the target language.
• Different sentences should overlap what you are training.
  • The same word should appear in at least 3 very different sentences.
  • The same grammar principle should appear in at least 10 very different sentences (and up to confidence).
• Do you have a playlist of favorite songs (in a Spotify-llke service?) Pick the ones in the target language and start memorizing their texts, one by one, top to bottom.

Able to understand not able to communicate​

This is the current research focus. Demanded for Serbian for the whole family, and for English for Ira and Alex.

TODO, absorb these advice:

• https://www.youtube.com/watch?v=FzjDYeoS9_w

Remember of 'All levels' from above.

You should train speaking / writing in two modes (both are important, the question is in proportion):

• Free mode, when you speak a lot, without paying attention to your mistakes and filling the gaps with words or phrases from another languages.
• Training mode. When your speak or write diligently, getting continuos correction feedback, and repeating the corrected version aloud. In both cases the topic should connect to your preceding skills, your life experience, and your interests for better ability to remember. Multi-modal practice is also great - written and spoken form plus an illustration. You should probably do only writing or only speaking, and other modalities should be made for you by an app.

Already speaking the language but willing to improve​

What you need:

• Continuous feedback in your natural writing and speaking. This is critical. Needs advanced tools and occasional teacher/coach/volunteer help.
• Complex reading. Use books or publications of professional writers, their active language is complex. This is highly recommended, even in your mother tongue.
• Vocabulary extension exercises. This is optional. Some like anki decks based on dictionaries. Some never use them.

Words to absorb: as many as possible

• Exclude professional language out of your interests.
  • You probably do not want to know what 'diapsid' means, unless you already learned this word in your mother tongue.
• Include poetic and archaic words and expressions.
  • However profession-specific it may be, you'll likely want to know what a 'farrier' is. After all, you want to read Dickens.
• Key targets are 30,000, and then 50,000 words.

Method Principles

The Fundamentals​

• Grammar is a habit. It is not knowledge. Grammar explanation is knowledge, but don’t you confuse an explanation with the matter itself.

The Main System of Principles For Fast Learning in Languages​

1. You memorize sentences → Your brain deduct the grammar and picks the words and stereotypical expressions.
   1. Memorize in all modalities (4 main + pictures if possible).
   2. Memorize by spaced repetition in both directions and all 4 modalities.
   3. At the later stages (> 30k words or earlier) we don’t need every rare word to join the active vocabulary, so we narrow down to 2 modalities.
   4. I’m lacking the knowledge of how this will work for disabled people who can practice less than 4 modalities, sorry. But probably the same more or less.
2. You memorizing throughput is limited. Thus, the very set of sentences defines the speed of your progress.
   1. They should cover everything, but not at once. The system should guide you according to the frequency of units of the language.
   2. They should repeat the same units in different contexts enough times with perfect spacing and avoid excessive repetition.
      1. Repeat grammar more times than words.
         1. Repeat grammar that has no correlation in mother tongue more than different forms of similar concepts, and repeat the latter more than similarities.
      2. Consider word families and word conjugation on repeat separately.
         1. You need conjugation models to track that irregular words are learned in high detail, while regular are learned by a few conjugated forms.
   3. They should be connected:
      1. to each other (make them a loose story if this does not damage unit selection),
      2. to the student’s current vocabulary (active on early stages, and passive otherwise),
      3. to the student’s background (mother tongue similarities and global professional slang).
      4. Small increments are much easier and sustainable for memorization than semantic islands.
   4. Building this list (or algorithms that will form ultra-personalized list from a generalized database) is the main challenge for the methodist/researcher, but it should not be the problem of the student.

Theory Overwhelm​

When a student watches or reads about learning techniques, they get overloaded. This breaks motivation and rises anxiety. However, the listed above is a variety of methods to be considered. The conclusion? There should be a customized stream of learning theory that does not reveal all at once, but rather focuses the student on highly-relevant exercises and piece of theory.

Learning is like the movement of a caterpillar: acquiring knowledge is the head moving forward, understanding is the middle following, and practice is the tail catching up. If you consume too much knowledge without enough practice, you’ll end up tearing your caterpillar in two.

Watching many YouTube advice might contribute to your motivation of learning. But it breaks focusing on the particular exercises you could be doing. And so it blocks you from doing anything.

You need to grow your learning practice as drawing a tree - in small increments from the existing trunk and branches. Focus on the delta, forget the rest. Kaizen. Growing a crystal.

Types of Learning Helpers​

What is the best: real life conversations, work with teacher, an app, or non-app exercises?

The thing is that for the maximum performance you need them all. Humans give you motivation, but synthetic exercises might intensify obtaining proficiency in certain areas. The real question is what you do and the proportion.

Questions to work on​

• Fun or diligence lead to faster progress?

Tool Principles

Anti-pattern hypothesis: ___ exercises​

Exercises like Some words ___ more words might be counter-productive. We don’t think in filling the gaps. We mostly put the words incrementally, not in the middle of an existing exercise.

This utilizes the slow thinking System 2 that uses theoretical knowledge while aims to train the ability to use that theory. So we are training the thinking mode that is not utilized in speaking (it is the fast System 1’s ambit), nor guiding as theoretical knowledge.

The replacement? Need to try this: showing several complete versions of the sentence and the student should pick which ‘feels’ natural. On wrong choice - feedback and repeat. Motivate them not to think (System 2), but to feel (utilizing System 1 and shaping it with feedback and repetition.)

Human Touch vs Good System​

The perfect tools looks this to me:

• We implement an app that implements the comprehensive training system for the principles described here. This is not a one-shot process, it should evolve. But it is an app for the direct use by a student. A non-human trainer.
• We introduce classes with a real teacher and chats with other learners where the parts of the system implemented and tested with the application act as a tutor, who guides the flow, but does not participate themself.
  • In chats, AI tutor might participate as a staff member character, but it shall not imitate a human participant.
  • In lessons, the AI tutor or algorithmic guardrails it should only be visible to the human teacher, not to the student.

Terms

• System 1 and System 2 are introduced by Kanneman. They are not proven to be physically separated. They are rather a theoretical idea to name the two distinctively observable modes of thinking.