arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.02810cs.CLcs.AIcs.LGcs.NE
收藏

数据文本生成中的词级幻觉控制

Controlling Hallucinations at Word Level in Data-to-Text Generation

Sorbonne Université(索邦大学);CNRS(法国国家科学研究中心);BNP Paribas(法国巴黎银行);University of Turin(都灵大学);Criteo AI Lab(Criteo人工智能实验室)

Clément Rebuffel, Marco Roberti, Laure Soulier, Geoffrey Scoutheeten, Rossella Cancelliere, Patrick Gallinari

摘要:

数据文本生成(DTG)是自然语言生成的一个分支,旨在将结构化数据转换为自然语言描述。该领域最近通过使用基于神经元的生成器而得到了推动,这些生成器一方面表现出很好的语法技能,而不需要手工制作的管道;另一方面,生成的文本的质量反映了训练数据的质量,在现实环境中只提供不完全对齐的结构文本对。因此,最先进的神经模型在其输出中包含误导性陈述-通常称为幻觉。这种现象的控制是当今DTG的一个主要挑战,也是本文要解决的问题。以前的工作是在实例级别处理这个问题的:为每个表引用对使用对齐得分。相比之下,我们提出了一种更细粒度的方法,认为幻觉应该在单词水平上进行治疗。具体来说,我们提出了一个多分支解码器,它能够利用单词级标签来学习每个训练实例的相关部分。这些标签是根据一个简单而有效的评分过程的基础上共现分析和依存句法分析。广泛的评估,通过自动化的指标和人类的判断标准WikiBio基准,显示我们的对齐标签的准确性和建议的多分支解码器的有效性。我们的模型能够减少和控制幻觉,同时保持生成文本的流畅性和连贯性。对降级版本的ToTTo的进一步实验表明,我们的模型可以成功地用于非常嘈杂的设置。

英文摘要:

Data-to-Text Generation (DTG) is a subfield of Natural Language Generation aiming at transcribing structured data in natural language descriptions. The field has been recently boosted by the use of neural-based generators which exhibit on one side great syntactic skills without the need of hand-crafted pipelines; on the other side, the quality of the generated text reflects the quality of the training data, which in realistic settings only offer imperfectly aligned structure-text pairs. Consequently, state-of-art neural models include misleading statements - usually called hallucinations - in their outputs. The control of this phenomenon is today a major challenge for DTG, and is the problem addressed in the paper. Previous work deal with this issue at the instance level: using an alignment score for each table-reference pair. In contrast, we propose a finer-grained approach, arguing that hallucinations should rather be treated at the word level. Specifically, we propose a Multi-Branch Decoder which is able to leverage word-level labels to learn the relevant parts of each training instance. These labels are obtained following a simple and efficient scoring procedure based on co-occurrence analysis and dependency parsing. Extensive evaluations, via automated metrics and human judgment on the standard WikiBio benchmark, show the accuracy of our alignment labels and the effectiveness of the proposed Multi-Branch Decoder. Our model is able to reduce and control hallucinations, while keeping fluency and coherence in generated texts. Further experiments on a degraded version of ToTTo show that our model could be successfully used on very noisy settings.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑