SentiTurca - A Sentiment Analysis Benchmark for Turkish

SentiTurca is a sentiment analysis benchmarking dataset for Turkish including three tasks, e-commerce reviews, hate speech and movie reviews classification. For more information about the dataset, the tasks and data curation process please visit the HF repo.

Benchmarking

Benchmarking code can be find under scripts/. To run a single task run run_single.sh:

#!/bin/bash

python3 run_sentiturca.py \
  --model_name_or_path dbmdz/bert-base-turkish-cased \
  --task_name hate \
  --max_seq_length 128 \
  --output_dir berturk \
  --num_train_epochs 5 \
  --learning_rate 2e-5 \
  --per_device_train_batch_size 128 \
  --per_device_eval_batch_size 128 \
  --do_train --do_eval --do_predict

Available task names are:

• e-commerce
• hate
• movies

To run all the tasks in order, please run run_all.sh. Benchmarking for BERTurk model and a handful LLMs can be found under the HF repo and the research paper.

Research paper and citation

Coming soon!