Developing more robust NLI models that generalise better to other unseen datasets
Developing more robust NLI models that generalise better to other unseen datasets
批准号:
2613081
负责人:
金额:
$0.0万
依托单位:
依托单位国家:
英国
项目类别:
Studentship
财政年份:
2020
资助国家:
英国
项目状态:
已结题
起止时间:
2020 至 --
中文摘要
博士学位将涉及制定策略,以帮助模型减少对训练数据中的偏差或伪影的依赖,使模型能够更好地概括看不见的数据,而不会有相同的偏差或伪影。本论文将专注于自然语言推理的任务,以更健壮的方式在单个数据集上训练模型,从而提高其他NLI数据集上的零射击性能。
英文摘要
The PhD will involve developing strategies to help models rely less on biases or artefacts within their training data, allowing the models to generalise better to unseen data without the same biases or artefacts. The thesis will focus on the task of Natural Language Inference, training models on a single dataset in a more robust way that will improve zero-shot performance on other NLI datasets.General Topic: data science.
期刊论文(0)
专著(0)
科研奖励(0)
会议论文
海外基金