arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2007.06761cs.CL
收藏

神经网络能从原始语言数据中获得结构偏差吗?

Can neural networks acquire a structural bias from raw linguistic data?

New York University(纽约大学)

Alex Warstadt, Samuel R. Bowman

摘要:

我们评估是否BERT,一个广泛使用的神经网络的句子处理,通过对原始数据进行预训练,获得对形成结构概括的归纳偏见。我们进行了四个实验测试其偏好的结构与线性概括在不同的结构依赖的现象。我们发现,BERT使3的4个经验领域的结构概括-主辅倒置,自反约束,并在嵌入式子句的动词时态检测-但使线性概括时,测试NPI许可。我们认为,这些结果是迄今为止最有力的证据,从人工学习支持的命题,结构性偏差可以从原始数据中获得。如果这一结论是正确的,这是初步的证据,一些语言共性可以习得的学习者没有先天的偏见。然而,人类语言习得的确切含义尚不清楚,因为人类学习语言的数据明显少于BERT。

英文摘要:

We evaluate whether BERT, a widely used neural network for sentence processing, acquires an inductive bias towards forming structural generalizations through pretraining on raw data. We conduct four experiments testing its preference for structural vs. linear generalizations in different structure-dependent phenomena. We find that BERT makes a structural generalization in 3 out of 4 empirical domains---subject-auxiliary inversion, reflexive binding, and verb tense detection in embedded clauses---but makes a linear generalization when tested on NPI licensing. We argue that these results are the strongest evidence so far from artificial learners supporting the proposition that a structural bias can be acquired from raw data. If this conclusion is correct, it is tentative evidence that some linguistic universals can be acquired by learners without innate biases. However, the precise implications for human language acquisition are unclear, as humans learn language from significantly less data than BERT.

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。

↑