README.md

metadata

license: cc-by-nc-sa-4.0
language:
  - ctn
pretty_name: LingReason Chintang Data
task_categories:
  - translation
tags:
  - low-resource-machine-translation
  - linguistic-reasoning
  - universal-dependencies
  - chintang
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/ctn_train.json
      - split: validation
        path: data/ctn_eval.json
      - split: test
        path: data/ctn_test.json
      - split: train_no_thinking
        path: data/ctn_no_thinking_train.json
      - split: validation_no_thinking
        path: data/ctn_no_thinking_eval.json
      - split: test_icl
        path: data/ctn_test_icl.json

Dataset Description

This dataset contains Chintang data for the LingReason project, which is generated from using the code released in the LingReason GitHub repository.

This dataset accompanies the paper Reasoning over Grammar: Can Synthetic Linguistic Reasoning Traces Enhance Low-Resource Machine Translation?.

Data Splits

Split	File	Examples	Description
test_icl	ctn_test_icl.json	344	Test set with linguistic reasoning guides (with placeholders) in the prompt, used for in-context learning experiment.
train	ctn_train.json	1,831	SFT/RFT train set with completed linguistic reasoning traces in the block.
validation	ctn_eval.json	114	Evaluation/Validation set with completed linguistic reasoning traces in the block.
train_no_thinking	ctn_no_thinking_train.json	1,831	SFT Train set without linguistic reasoning traces.
validation_no_thinking	ctn_no_thinking_eval.json	114	Evaluation/Validation set without linguistic reasoning traces.
test	ctn_test.json	344	Test set with completed linguistic reasoning traces in the block, used for ICL baseline as well as SFT and RFT experiments.

Usage

from datasets import load_dataset

dataset = load_dataset("OLAResearchX/LingReason")

print(dataset)
print(dataset["train"][0])