Dataset card for branch

Dataset summary

This dataset contains threads parsed from /b/ board of arhivach, which in turn pulls data from 2ch. The dataset is regularly updated.

The dataset consists of two parts:

• index.tsv - a registry, which lists all pulled threads;
• threads - files with threads content, threads are grouped into batches, each batch contains 10000 elements (maximum number of files in a directory allowed by git). Since huggingface doesn't allow to store more than 1000000 files in a single repo, there is the other repo, which contains threads violating this requirement.

Dataset structure

Data instance

An example of an entry from the dataset is given below:

{
  "thread": 47,
  "title": "- Сап, Антоны, на связи мудило ОП треда нищебродских наушников. Купил всё-таки эти Sennheiser HD 201 (вместе с ними бесплатно по акции был китайский портативный усилок FiiO e3 не ожидал от него ничего необычного, просто интересно было. Сегодня привезли, сперва пробую подключить без усилка к китайской балалайке Texet T-479. Послушал, вполне годно. Подключил усилок — вообще заебца! Громкости хватает с головой, звук отличный, басы присутствуют, верха прозрачные. Тест"
  "date": "04-12-2012",
  "board": "/b/",
  "stars": 315,
  "folder": 00000000-00009999
}

Data fields

Each dataset entry therefore consists of the following fields:

• thread - thread id, which can be used to access thread on arhivach by navigating to /thread/{thread}/;
• title - the beginning of oppost in the thread, which is displayed in board catalog;
• date - oppost publication date, in format DD-MM-YYY;
• board - board identifier, on which the thread was created;
• stars - number of stars assigned to thread by viewers;
• folder - folder in the threads directory, in which the thread is stored.

Contribution

You can contribute by pulling latest threads from arhivach to update the dataset. For doing this, there is another project. After installation, perform the following actions from the dataset folder, which should have symbolic link to the much python module:

1. Identify id of the last downloaded thread:

tail -n 1 index.tsv | awk '{ print $1 }'

2. Open arhivach and manually find offset to the thread with this id. Increasing offset involves moving farther to old threads, and decreasing the thread ids which appear on the page.

3. Update index.tsv by listing threads from this offset to the end:

python -m much filter -t 20 -i index.tsv -s $OFFSET

4. Pull threads content:

python -m much grab -n 20

5. Update folder column:

python -m much update-folder-column