Skip to content

Latest commit

 

History

History
18 lines (16 loc) · 1.08 KB

File metadata and controls

18 lines (16 loc) · 1.08 KB

Data Description

Input file must be provided in parquet format. If you have csv data, you can save it as parquet using pandas function pandas.DataFrame.to_parquet.

Columns

  • art_ind - index of the article in format "[date]: [title]"
  • url - link to the original article
  • title - raw newspaper title
  • text - raw newspaper text
  • date - publish date in format DD.MM.YYYY
  • _time - publish date in UNIX format
  • source - source of the newspaper (e.g. "Lenta.ru")
  • kw_ne - list of keywords and named entities from newspaper text, lemmatized
  • clean_text - preprocessed text, tokens inside one sentence separated with one space, sentences separated with "###SENT###" (see example)
  • bert_vec - document embedding

Methods

  • Document embedding can be produced with rubert-tiny.
  • Text preprocessing includes lemmatization, stop words removing, punctuation and digits removing and can be done with util/data_preprocessing.py.