[EL] Greek initial support

[dsimop]

No description provided.

[dsimop]

The failing tests occur only because the recognition model cannot handle a specific case for , etc. - which happens always often in Greek for words whose gender is male. Let "word1 word2" be a name and "word3" a domain. In any case the logic is the same and the recognition model has to handle these entity_names as follows:

If there is any word in a name/domain ending with "ς", that character has to be omitted/skipped when reading a sentence/test. For instance: name: "word1 word2", "word1" ends with "ς". But the sentence/test must include that name without the aforementioned character "ς".

[glogiotatidis]

nice work! 🎉

Translations are good. The failing tests are surfacing a problem in the parsing engine imho. We will need to have this flexibility with Greek (maybe other langs as well).

Here's a related issue:

An entity name is ανεμιστήρα οροφής (fan ceiling)
When we say a full sentence to turn it off we would add the word "the" between "fan" and "ceiling" and remove the trailing ς from the word before. I would become

κλείσε τον ανεμιστήρα **της** οροφής

Another example for φωτιστικό υπνοδωματίου (lamp bedroom)

κλείσε το φωτιστικό **του** υπνοδωματίου

here there's no trailing ς and thus none removed.

Thoughts on how to handle those scenarios?

[dsimop]

The problems with the male genders mentioned above, have been resolved by adapting the name definitions. Even the case presented by @glogiotatidis is covered. That is a workaround and not the optimal solution, which is to update the recognition model itself.