Easy to Read in Accessibility of Machine Learning and Generative AI

[LuisMalhadas]

The document speaks of plain language but not easy to read text.
Easy to read text follows stricter normative, here is an example: https://www.inclusion-europe.eu/wp-content/uploads/2017/06/EN_Information_for_all.pdf

Here I share a csv of a set of documents over the subject (including some papers about generative AI applied to easy2read):

Accessibles-Easy2Read.csv

The statement:

For example, a translation of the popular poem, “Mary Had a Little Lamb”, into a non-English language, and then back to English observes the error of “it’s fleece was white as snow” being converted to “it snowed sheep hair”. 

Does not mention the models used nor the source.
While it may be true, for the mentioned instance, it is stated as a linear deterministic event, which is not.
Generative AI is non-deterministic, and therefore one can't state one single example as if it always follows that path. Although it may struggle with translating meanings, it is only due to lack of context, and lack of proper curated training data.
You can have better results given better data and better context.

The document also lacks, in general, a consideration towards the cognitive accessibility.
Another example of cognitive accessibility that generative AI will enhance, is the generation of pictograms, by allowing a non-deterministic generation of pictograms that can iteratively improve the process of communicating meaning and semantics.