My solution performs better on eng language - cause im using pretrained en models. but language detection works good.

I spend 10 hours for this true MVP fast and stable solution. There are some problems:
1) for threads grouping i used Minhash algo, but it performs bad cause requires some tuning - now it finds word with minimum hash in title and group but hash value. That causes threads with many articles like first thread where each article has word "police"
2) works bad on russian language
3) news detection works on domain "white list" - so it has low accuracy

In our preliminary tests, this submission received the following scores (out of 100):

Languages: 100
News EN: 72
News RU: 85
Categories EN: 44
Categories RU: 9
Threads EN: 51
Threads RU: 27
Top EN: 53
Top RU: 30

This is not the final result, please stay tuned for updates. We apologize for the delay.

Final score for this submission (out of 100):

Languages: 90.12
News EN: 38.34
News RU: 45.8
Categories EN: 13.71
Categories RU: 13.29
Threads EN: 13.33
Threads RU: 13.34
Top news EN: 10.82
Top news RU: 10.83

These data reflect the relative accuracy, precision and speed of the algorithm as compared to the other submissions.

В ходе предварительного тестирования алгоритма были выявлены следующие недостатки в ранжировании:

– Отсутствуют многие главные сюжеты в разделе 'Main' и внутри категорий. В топе отражены нерелевантные сюжеты.

– Нарушена сортировка статей в сюжетах: релевантные статьи смешаны с нерелевантными.

We were unable to launch this submission due to the following reasons:
- unable to run MacOS binary under Debian GNU/Linux 10.1 (buster), x86-64

#comment9998
We had to rebuild your binary and will apply relevant penalties during the final scoring.

Low categorization accuracy in Russian. Almost all articles are classified as Other or Entertainment.

Threads containing unrelated articles