arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2110.02406cs.CL
收藏

神经语言模型中的词汇习得

Word Acquisition in Neural Language Models

University of California, San Diego(加利福尼亚大学圣地亚哥分校);Department of Cognitive Science(认知科学系);Halıcıoğlu Data Science Institute(Halıcıoğlu数据科学研究所)

Tyler A. Chang, Benjamin K. Bergen

摘要:

我们研究了神经语言模型如何在训练过程中获取单个单词,提取学习曲线和MacArthur-Bates沟通发展量表上超过600个单词的获取年龄(Fenson等人,2007年)。借鉴儿童词汇习得的研究,我们评估了LSTM、BERT和GPT-2中词汇习得年龄的多个预测因素。我们发现,具体性,字长和词汇类的影响是有针对性的不同,在儿童和语言模型,加强互动和感觉运动经验在儿童语言习得的重要性。语言模型比儿童更依赖于词频,但像儿童一样,他们在较长的话语中表现出较慢的单词学习。有趣的是,模型在单向和双向模型以及LSTM和Transformer架构的训练过程中遵循一致的模式。模型在训练早期基于unigram标记频率进行预测,然后松散地过渡到bigram概率,最终收敛到更细微的预测。这些结果揭示了儿童分布式学习机制的作用,同时也为语言模型中更像人类的语言习得提供了见解。

英文摘要:

We investigate how neural language models acquire individual words during training, extracting learning curves and ages of acquisition for over 600 words on the MacArthur-Bates Communicative Development Inventory (Fenson et al., 2007). Drawing on studies of word acquisition in children, we evaluate multiple predictors for words' ages of acquisition in LSTMs, BERT, and GPT-2. We find that the effects of concreteness, word length, and lexical class are pointedly different in children and language models, reinforcing the importance of interaction and sensorimotor experience in child language acquisition. Language models rely far more on word frequency than children, but like children, they exhibit slower learning of words in longer utterances. Interestingly, models follow consistent patterns during training for both unidirectional and bidirectional models, and for both LSTM and Transformer architectures. Models predict based on unigram token frequencies early in training, before transitioning loosely to bigram probabilities, eventually converging on more nuanced predictions. These results shed light on the role of distributional learning mechanisms in children, while also providing insights for more human-like language acquisition in language models.

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。

↑