arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1802.04302cs.CLstat.ML
收藏

评价句子嵌入的组合性

Evaluating Compositionality in Sentence Embeddings

Harvard University(哈佛大学);Stanford University(斯坦福大学)

Ishita Dasgupta, Demi Guo, Andreas Stuhlmüller, Samuel J. Gershman, Noah D. Goodman

摘要:

类人AI的一个重要挑战是组合语义。最近的研究试图通过使用深度神经网络来学习句子的向量空间嵌入来解决这个问题,然后将其作为其他任务的输入。我们提出了一个新的数据集,这样的任务,“自然语言推理”(NLI),不能只用单词级的知识来解决,需要一些组合。我们发现,最先进的句子嵌入的性能(InferSent; Conneau等人,2017)在我们的新数据集上很差。我们分析了Infersent学习的决策规则,发现它们与其训练数据集中生态有效的简单算法是一致的。此外,我们发现使用我们的数据集增强训练可以提高我们数据集的测试性能,而不会损失原始训练数据集的性能。这突出了结构化数据集在更好地理解和改进人工智能系统方面的重要性。

英文摘要:

An important challenge for human-like AI is compositional semantics. Recent research has attempted to address this by using deep neural networks to learn vector space embeddings of sentences, which then serve as input to other tasks. We present a new dataset for one such task, `natural language inference' (NLI), that cannot be solved using only word-level knowledge and requires some compositionality. We find that the performance of state of the art sentence embeddings (InferSent; Conneau et al., 2017) on our new dataset is poor. We analyze the decision rules learned by InferSent and find that they are consistent with simple heuristics that are ecologically valid in its training dataset. Further, we find that augmenting training with our dataset improves test performance on our dataset without loss of performance on the original training dataset. This highlights the importance of structured datasets in better understanding and improving AI systems.

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。

↑