Same Sentiment Classification Train/Dev/Test Pair IDs

This "dataset" only includes the compiled pairings of the Yelp Business Review Dataset. To get access to the actual review texts, please follow the instructions on the Yelp Dataset webpage.

The data format is JSONlines.
Python Load Example:

import pandas as pd
traindev_df = pd.read_json("df_traindev.jsonl", lines=True)
test_df = pd.read_json("df_test.jsonl", lines=True)

# example access to single business/review id
s1_bid = test_df.iloc[0]["sent1_business_id"]
s1_rid = test_df.iloc[0]["sent1_review_id"]
s2_bid = test_df.iloc[0]["sent2_business_id"]
s2_rid = test_df.iloc[0]["sent2_review_id"]
label = test_df.iloc[0]["is_same_side"]

See documentation at:

Yelp Dataset Schemata (only business.json and review.json were used)

Yelp Business Category Hierarchy (download the json file as all_category_list.json)

For details on how the data was compiled and used in our experiments, please refer to our code repository. Other derived data splits can be reproduced deterministically by using the same random seed as in our experiments.