对联数据集。

Also available on HuggingFace.

This is a project to fetch couplets from 冯重朴_梨味斋散叶_的博客

This dataset contains more than 700,000 couplets.

scrapy runspider sina_spider.py

It will store the data into ./output/.

There is an already fetched and cleaned dataset that can be used directly with the seq2seq model. You can download it at here.

The downloaded data contains 5 files:

1. train/in.txt: The input of the couplets. Each line is an input. Each word is split by space.
2. train/out.txt: The output of the couplets. Each line is the output for the same line in the in.txt. Each word is split by space.
3. test/in.txt: Same as train/in.txt but with less data.
4. test/out.txt: Same as train/out.txt but with less data.
5. vocabs: Vocabs file. Add <s> and <\s> as the first vocabs, which will be used to train in the seq2seq mode.