What's so special about BERT's layers? A closer look at the NLP pipeline in monolingual and multilingual models

Malvina Nissim, Wietse De Vries, Andreas Van Cranenburgh

Code that ran

We lifted 2 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

Repository	Role	Ran
copy not recorded	—	2 of 2

Function	Status	Where it lives
load_data	Ran	this paper's copy was not recorded; identical code first harvested from wietsedv/bertje pointer only · get_code("2b4139171701786a")
read_examples	Ran	this paper's copy was not recorded; identical code first harvested from wietsedv/bertje pointer only · get_code("5328605f9e05de05")

Repositories linked to this paper

wietsedv/bertje — marked official in Papers with Code; named in the paper
wietsedv/bertje — marked official in Papers with Code

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Peeking into the inner workings of BERT has shown that its layers resemble the classical NLP pipeline, with progressively more complex tasks being concentrated in later layers. To investigate to what extent these results also hold for a language other than English, we probe a Dutch BERT-based model and the multilingual BERT model for Dutch NLP tasks. In addition, through a deeper analysis of partof-speech tagging, we show that also within a given task, information is spread over different parts of the network and the pipeline might not be as neat as it seems. Each layer has different specialisations, so that it may be more useful to combine information from different layers, instead of selecting a single one based on the best overall performance.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2004.06499")
get_code_for_paper("2004.06499")
have("2004.06499")

Connect an agent — have() is free.