The week's pick
Reseach Article
Adaptive Crowdsourcing Task Generation and Workflow Control for Human Feedback Data Collection
| International Journal of Computer Applications |
| Foundation of Computer Science (FCS), NY, USA |
| Volume 186 - Number 70 |
| Year of Publication: 2025 |
| Authors: Liqing Wang, Wanjin Chen, Yongyue Xu, Juan Wang |
10.5120/ijca2025924552
|
Liqing Wang, Wanjin Chen, Yongyue Xu, Juan Wang . Adaptive Crowdsourcing Task Generation and Workflow Control for Human Feedback Data Collection. International Journal of Computer Applications. 186, 70 ( Mar 2025), 22-28. DOI=10.5120/ijca2025924552
Abstract
Due to the dire scarcity of corpora, the quality of low-resource language translation falls short of the public expectation. However, adopting Reinforcement Learning from Human Feedback (RLHF) can remarkably improve model quality. Nonetheless, obtaining human feedback data is typically time-consuming, costly, or plagued by severe inconsistency. Thus, this paper develops and implements a self-generating crowdsourcing workflow tailored for low-resource translation to address the above-mentioned issues. Under the consideration of quality and cost, this workflow will operate automatically and continually until it obtains the final results, according to a process in which the generated options are filled in the blanks for evaluating and selecting crowdsourcing tasks in various formats, then managing their iterative execution. This method enables the acquisition of various feedback data with varied requirements in diverse forms—including ranking, scoring, comparative judgments, and error correction—at low cost and with high efficiency. These data can then be used to train reward models with the consequence of enhancing reinforcement learning performance. This paper’s experimental results testify to the effectiveness of this approach.RLHF; low-resource translation; adaptive; iterative workflow; crowdsourcing
References
- Yazar, B. K., Şahın, D. Ö., & Kiliç, E. 2023. Low-resource neural machine translation: A systematic literature review. IEEE Access, 11, 131775-131813.
- Chen, H., Yuan, K., Huang, Y., Guo, L., Wang, Y., & Chen, J. 2023. Feedback is all you need: from ChatGPT to autonomous driving. Science China. Information Sciences, 66(6), 166201.
- Wang, L., & Xiao, Y. 2024. Improving Low-Resource Machine Translation Using Reinforcement Learning from Human Feedback. Intelligent Automation & Soft Computing, 39(4).
- Ri, J. 2016. Utilizing crowdsourced translation in translator training—focusing on the practices in Japan—. The Journal of Translation Studies, 17(2), 91-117.
- Vaughan, J. W. 2018. Making better use of the crowd: How crowdsourcing can advance machine learning research. Journal of Machine Learning Research, 18(193), 1-46.
- Howe, J. 2006. The rise of crowdsourcing. Wired magazine, 14(6), 176-183.
- Skalse, J., Howe, N., Krasheninnikov, D., & Krueger, D. 2022. Defining and characterizing reward gaming. Advances in neural information processing systems, 35, 9460-9471.
- Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., . . . Hüllermeier, E. 2023. ChatGPT for good? On opportunities and challenges of large language models for education. Learning and individual differences, 103, 102274.
- OpenAI, T. 2022. Chatgpt: Optimizing language models for dialogue. OpenAI. In.
- Christiano, P. F., Leike, J., Brown, T., Martic, M., Legg, S., & Amodei, D. 2017. Deep reinforcement learning from human preferences. Advances in neural information processing systems, 30.
- Hejna III, D. J., & Sadigh, D. 2023. Few-shot preference learning for human-in-the-loop rl. Paper presented at the Conference on Robot Learning.
- Krippendorff, K. 2004. Reliability in content analysis: Some common misconceptions and recommendations. Human communication research, 30(3), 411-433.
- Gilardi, F., Alizadeh, M., & Kubli, M. 2023. ChatGPT outperforms crowd workers for text-annotation tasks. Proceedings of the National Academy of Sciences, 120(30), e2305016120.
- Georgescu, M., Pham, D. D., Firan, C. S., Gadiraju, U., & Nejdl, W. 2014. When in doubt ask the crowd: Employing crowdsourcing for active learning. Paper presented at the Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14).
- Simpson, E. D., Venanzi, M., Reece, S., Kohli, P., Guiver, J., Roberts, S. J., & Jennings, N. R. 2015. Language understanding in the wild: Combining crowdsourcing and machine learning. Paper presented at the Proceedings of the 24th international conference on world wide web.
- Fuge, M., Tee, K., Agogino, A., & Maton, N. 2014. Analysis of collaborative design networks: A case study of openideo. Journal of Computing and Information Science in Engineering, 14(2), 021009.
- Feng, J. H., & Li, G. L. 2015. A survey on crowdsourcing. Jisuanji Xuebao/Chinese Journal of Computers, 38, 1713-1726. doi:10.11897/SP.J.1016.2015.01713
- Bhatti, S. S., Gao, X., & Chen, G. 2020. General framework, opportunities and challenges for crowdsourcing techniques: A comprehensive survey. Journal of Systems and Software, 167, 110611.
- Kang, Q., & Tay, W. P. 2020. Task recommendation in crowdsourcing based on learning preferences and reliabilities. IEEE Transactions on Services Computing, 15(4), 1785-1798.
- Y Bo, Wang L. 2020. Task assignment method based on worker multi-attribute characteristic model. Computer Engineering and Design.;41(5):1398-403.
- Chen B, Wang L, Jiang X, Yao H. 2021. Survey of task assignment for crowd-based cooperative computing. Computer Engineering and Applications.;57(20):1-2.
- Carvalho, A., Dimitrov, S., & Larson, K. 2016. How many crowdsourced workers should a requester hire? Annals of Mathematics and Artificial Intelligence, 78, 45-72.
- Papineni, K., Roukos, S., Ward, T., & Zhu, W.-J. 2002. Bleu: a method for automatic evaluation of machine translation. Paper presented at the Proceedings of the 40th annual meeting of the Association for Computational Linguistics.
Index Terms
Keywords