Skip to content

Dataset · d0rj

500k Russian Organization Reviews from Yandex Maps

d0rj/geo-reviews-dataset-2023

Large cleaned corpus of real Russian-language reviews from 2023 for NLP tasks.

What's inside

The dataset from Yandex contains 500,000 unique Russian-language reviews of organizations in Russia published on Yandex Maps from January to July 2023. Extremely short single-word comments were excluded, and all texts were scrubbed of personal data such as phone numbers. The collection was published for academic and research tasks in natural language processing.

What you can do with it

  • Training and fine-tuning sentiment analysis and Russian text classification models
  • Named entity recognition (NER) and service quality aspect extraction from user reviews
  • SFT or embedding training for semantic search over Russian reviews
Tasks
Text classification, Sentence similarity, Token classification
Languages
Russian
Size
100K–1M rows
Downloads this month
759
Likes
4

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("d0rj/geo-reviews-dataset-2023")