Skip to content

Dataset · microsoft

Bing search queries and ranking

microsoft/ms_marco

MS MARCO collection for search: Bing queries, answers, passages, and ranking tasks.

What's inside

MS MARCO is a dataset collection for deep learning in search, released following a 2016 paper. It includes 100k real Bing queries with human answers, as well as versions with 1M queries, data for generation, passage ranking, keyphrase extraction, and dialogue. The language is English, and the volume ranges from 1M to 10M examples.

What you can do with it

  • Training retrieval and ranking models
  • Fine-tuning reranker for search
  • Evaluating QA systems on Bing questions
Languages
English
Size
1M–10M rows
Downloads this month
22.6K
Likes
263

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("microsoft/ms_marco")