Dataset: Russian Social Media Text Classification

About this Dataset

Russian Social Media Text Classification

VKontakte communities can belong to one of several predefined categories. But even among the sports communities there is a fairly strong division by subject! The same authors can write about only one sport or at once about a large number.
Based on a given set of posts, determine the topic - what kind of sport is being discussed in the selected community?

Here is a list of available categories:

athletics,
autosport,
basketball,
boardgames,
esport,
extreme,
football,
hockey,
martial arts,
motosport,
tennis,
volleyball,
winter_sport

evaluate metric look like:

def score(true, pred, n_samples):
    counter = 0
    if true == pred:
        counter += 1
    else:
        counter -= 1
    return counter / n_samples

Tables

Sample Submission

@kaggle.mikhailma_russian_social_media_text_classification.sample_submission

22.6 KB
2626 rows
2 columns


CREATE TABLE sample_submission (
  "oid" BIGINT,
  "category" VARCHAR
);

Test

@kaggle.mikhailma_russian_social_media_text_classification.test

9.65 MB
26260 rows
2 columns


CREATE TABLE test (
  "oid" BIGINT,
  "text" VARCHAR
);

Train

@kaggle.mikhailma_russian_social_media_text_classification.train

14.1 MB
38740 rows
3 columns


CREATE TABLE train (
  "oid" BIGINT,
  "category" VARCHAR,
  "text" VARCHAR
);