Project description
1. Build a word base on which the neural network will learn. For example, you can take ready-made bases on the Internet. Or you can take a text (e.g. War and Peace) and break it down into words.
2. Write Python code that translates words into matrices. You can, for example, use such encoding: the alphabet is the column of the matrix, the number of the letter in the word in order by the rows. At the intersection of the desired column and row, a unit is placed, the remaining elements are zeros.
3. To start with, build a simple network of autoencoders with Dense layers and try to compress the matrix and then convert it into a vector.
4. Check the quality of the network: the percentage of errors after decoding words.