idoazzz / memorizer

Help people learn new english words.

Memorizer

Help people that struggle to remember new english words, based on associative memory method.
There are a lot of methods, tricks and ways to learn new english words.
Memorizer takes your new word and suggest you similar soundlike words that are known to you.
Now, you have only one job. Bound the words that Memorizer suggested you, and the new associative soundlike words.

Example

• Search your target word.
  • bargain
• Pick the most associative words to each part of your word.
  • bar game.
• Bound your association to the new word! We know by the displayed definitions that bargain is negotiation of terms -
  • Every time you will negotiate you have to start with bar game. Without it you will lose!

Now, we will remember what is bargain! (hopefully)

memorizer's Issues

Improve algorithm using gradient descent and AI tools.

Take all the samples and prepare dataset - think about solution.

Fix Git broken repo.

Get rid of console errors/warnings.

Fix sass colors names and imports.

Clean and document frontend code.

Optimize requests with Asyncio.

Limit bug - fetching from 100 only with defs (e.g 120/100 associations fails)

Add search settings (limit and more...)

Add tests to backend.

Make the front compatible to all screens resolutions.

Add README.

Lint + CI

Split by syllables checkbox.

Add configurations file

Add images and gifs to README.

Add definitions to words and splits (modal).

Fix the difference between the camelcase methods.

Apply new algorithm.

Search Datamuse replacement.

The algorithm should be based on recommendation system algorithms.

• Get input. (Correlation)
• Transform it to verb\noun - use NodeBox. (Correlate)
• Search sound-like words in Datamuse for each word split combination (Co rrelate, Cor relate, Corre late...).
  ** Filter returned words with NodeBox (connecting words, 1 letter words, words without defs, remove stopwords with tfidfvectorizer paremter).
  ** Sort by Regressive Imagery Dictionary (NodeBox).
  ** Score the associations together not individualy - Use weighted sum (Look at recommendation systems score functions for compatible function). Consider use avg of all the popularities instead of popularity of each individual.
• Score all the splits.
• Return max-scored split.

Add adjectives to suggested associations

In datamuse:
adjectives that are often used to describe ocean

Add save option and my dictionary feature with firebase.

Make big dataset and measure performances with RMSE or MAE (score deltas)

Limits in non-split don't work

Handle special symbols in words. (*#@!()@!)

Add tests to frontend.

Algorithm optimization: Only dictionary defined words (associations) will get into the word lists.

Algorithm optimization: Only dictionary defined words (associations) will get into the word lists.
For example the word knapsack generate alot of words that don't have any meaning.
Filter them!

Move to sass.

Optimize the frontend with Profiler.

Insert hebrew association feature (Detailed).

Make transliteration table from English to Hebrew.
Translate naively words and after search them in the stopword dictionary. Consider use search that matches words with hamming distance of 1/2 "אהיו" letters.

1. Search the splitted/whole word in here: https://stevemorse.org/hebrew/eng2heb.html.
2. Get the hebrew result.
3. Search it on hebrew dictionary.
4. If the word exists => Add it to the associations list.

Manual splitting feature.

Changing word during search (alert) bug

Start over the new backend (NodeJS)

Think about new design.

Add word correction feature, with typing timeout (optimization).

Fix changing word with highlighted words bug