Are Neural Networks Extracting Linguistic Properties or Memorizing Training Data? An Observation with a Multilingual Probe for Predicting Tense - HAL-SHS - Sciences de l'Homme et de la Société Access content directly
Conference Papers Year : 2021

Are Neural Networks Extracting Linguistic Properties or Memorizing Training Data? An Observation with a Multilingual Probe for Predicting Tense

Bingzhi Li
  • Function : Author
  • PersonId : 1095862

Abstract

We evaluate the ability of Bert embeddings to represent tense information, taking French and Chinese as a case study. In French, the tense information is expressed by verb morphology and can be captured by simple surface information. On the contrary, tense interpretation in Chinese is driven by abstract, lexical, syntactic and even pragmatic information. We show that while French tenses can easily be predicted from sentence representations, results drop sharply for Chinese, which suggests that Bert is more likely to memorize shallow patterns from the training data rather than uncover abstract properties.
Fichier principal
Vignette du fichier
chinese_tense.pdf (143.2 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

halshs-03197072 , version 1 (13-04-2021)

Identifiers

  • HAL Id : halshs-03197072 , version 1

Cite

Bingzhi Li, Guillaume Wisniewski. Are Neural Networks Extracting Linguistic Properties or Memorizing Training Data? An Observation with a Multilingual Probe for Predicting Tense. EACL 2021, Apr 2021, Kiev (on line), Ukraine. ⟨halshs-03197072⟩
140 View
121 Download

Share

Gmail Mastodon Facebook X LinkedIn More