Discord

Checklist NEOAI 2026 Homework (take-home practice problems) · Task 3

Find Comments

Predict the five most likely authors of each chat comment from a social network.

  • NLP
  • Multi-class text classification (top-5 authorship attribution)

The task

The training set contains chat comments from a social network with their authors; the test set contains only the texts. Data are in chronological order, and the test set contains the last 50 messages (filtered by length) of each user. For each test comment the contestant returns the five most probable authors.

Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.

At a glance

You get
train_texts.csv (index, text, target = commentator id), test_texts.csv (index, text), sample_submission.csv.
You submit
CSV with columns index, target, where target lists 5 predicted commentator ids.
Scoring
Recall@5: the share of test comments whose true author is among the five predicted authors.
Rules
  • External data are prohibited; downloading NLTK corpora is allowed.
  • Libraries outside the standard Kaggle environment are prohibited.
  • The only pretrained weights allowed are deepvk/USER2-small and deepvk/USER2-base.
Format
Take-home practice problem for NEOAI 2026 participants (Kaggle, 12 Apr – 1 May 2026). Individual.

Details

Year
2026, Online
Round
Homework (take-home practice problems) · Task 3
Language
English
License
Per-competition Kaggle data licence: Subject to Competition Rules (Code Stars, Find Comments); MIT (Terminal Animals), as stated by the source