README.md

metadata

annotations_creators:
  - derived
language:
  - ja
license: cc-by-sa-4.0
multilinguality: monolingual
task_categories:
  - text-retrieval
task_ids:
  - document-retrieval
tags:
  - mteb
  - text
  - code
  - japanese
  - sample
configs:
  - config_name: corpus
    data_files:
      - split: train
        path: corpus/train-*
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
  - config_name: queries
    data_files:
      - split: train
        path: queries/train-*
dataset_info:
  - config_name: corpus
    features:
      - name: _id
        dtype: string
      - name: title
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: train
        num_bytes: 338
        num_examples: 5
    download_size: 1765
    dataset_size: 338
  - config_name: default
    features:
      - name: query-id
        dtype: string
      - name: corpus-id
        dtype: string
      - name: score
        dtype: int64
    splits:
      - name: test
        num_bytes: 150
        num_examples: 5
    download_size: 1551
    dataset_size: 150
  - config_name: queries
    features:
      - name: _id
        dtype: string
      - name: text
        dtype: string
    splits:
      - name: train
        num_bytes: 453
        num_examples: 5
    download_size: 1667
    dataset_size: 453

JapaneseCode1Retrieval-sample

A sample dataset for Japanese-English code retrieval evaluation. This dataset contains Japanese natural language descriptions paired with Python code snippets.

Task category

Retrieval

Domains

Programming, Code Generation

Dataset Structure

The dataset follows the standard MTEB retrieval format:

• corpus/corpus-00000-of-00001.parquet: 5 Python code documents with fields _id, title, text
• queries/queries-00000-of-00001.parquet: 5 Japanese queries with fields _id, text
• data/test-00000-of-00001.parquet: 5 relevance judgments with fields query-id, corpus-id, score

Usage

You can evaluate an embedding model on this sample dataset using the following code:

import mteb

# Load the sample dataset
task = mteb.get_task("JapaneseCode1Retrieval-sample")
evaluator = mteb.MTEB(tasks=[task])

# Run evaluation with your model
model = mteb.get_model("your-model-name")
results = evaluator.run(model)

Sample Content

This sample dataset contains:

• 5 Japanese natural language queries describing programming tasks
• 5 corresponding Python code snippets
• 5 relevance judgments connecting queries to code

The data has been slightly modified for demonstration purposes.

License

Please refer to the CC BY-SA 4.0 license terms.