What Does BERT Learn about the Structure of Language? (P19-1)

Copied to clipboard

Challenge: BERT is a language representation model that has performed well in diverse language understanding benchmarks.
Approach: They perform experiments to unpack the elements of English language structure learned by BERT.
Outcome: The proposed model outperforms state-of-the-art models in the GLUE benchmark by a significant margin.

Similar Papers

Finding Structural Knowledge in Multimodal-BERT (2022.acl-long)

Copied to clipboard

Challenge: Several multimodal-BERT models learn contextualized embeddings through training on linguistic data and visual data.
Approach: They propose to make the structure of language and visuals explicit by a dependency parse . they also propose to encode the scene tree in the multimodal-BERT models .
Outcome: The proposed models do not encode the scene trees in the language description.
What’s so special about BERT’s layers? A closer look at the NLP pipeline in monolingual and multilingual models (2020.findings-emnlp)

Copied to clipboard

Challenge: In addition, information on part-of-speech tagging is spread over different parts of the network and the pipeline might not be as neat as it seems.
Approach: They propose to probe Dutch BERT-based model and multilingual BERT model for Dutch NLP tasks to see if this holds true for other languages.
Outcome: The proposed model is based on a Dutch model and a multilingual model for Dutch NLP tasks.
Does Chinese BERT Encode Word Structure? (2020.coling-main)

Copied to clipboard

Challenge: Existing work has focused on analyzing the features captured by representative models such as BERT . however, little work has investigated word features for character languages such as Chinese .
Approach: They investigate Chinese BERT using attention weight distribution statistics and probing tasks to understand word features.
Outcome: The proposed model improves syntactic, semantic and word sense knowledge on a wide range of NLP tasks.
Does BERT Rediscover a Classical NLP Pipeline? (2022.coling-1)

Copied to clipboard

Challenge: Existing theories of BERT's structure lack conclusive empirical support . however, there is scepticism about the premises of probing itself .
Approach: They propose a new probe called GridLoc that can take into account token positions, training rounds, and random seeds.
Outcome: The proposed probe detects other, stronger regularities suggesting appeals to layer depth may not be the preferable mode of explanation for BERT’s inner workings.
CxGBERT: BERT meets Construction Grammar (2020.coling-main)

Copied to clipboard

Challenge: lexico-semantic elements capture a large amount of linguistic information, but they do not capture all information contained in text.
Approach: They propose to use BERT to train a model that uses a deep bidirectional transformer to capture a significant amount of lexico-semantic information.
Outcome: The proposed model captures lexico-semantic information, but it is redundantly encoded in lexical information.
Exploring the Role of BERT Token Representations to Explain Sentence Probing Results (2021.emnlp-main)

Copied to clipboard

Challenge: Recent studies have focused on enhancing existing models with the primary objective of improving downstream performance on various NLP tasks.
Approach: They propose to use BERT to encode meaningful knowledge in token representations to explain probing results.
Outcome: The proposed model can detect syntactic and semantic abnormalities and distinguish between grammatical number and tense subspaces.
A Primer in BERTology: What We Know About How BERT Works (2020.tacl-1)

Copied to clipboard

Challenge: a new study examines the current state of knowledge about the BERT model . the model is a stack of transformer encoder layers that are based on multiple self-attention ''heads''
Approach: They present a survey of over 150 studies of the popular Transformer-based model BERT . they discuss the current state of knowledge about how BERT works and how it is represented .
Outcome: The proposed model is based on the Transformer-based model with state-of-the-art results . the proposed model has little cognitive motivation and is too small to perform ablation studies .
Let’s Play Mono-Poly: BERT Can Reveal Words’ Polysemy Level and Partitionability into Senses (2021.tacl-1)

Copied to clipboard

Challenge: Pre-trained language models encode rich information about linguistic structure but their knowledge about lexical polysemy remains unclear.
Approach: They propose a setup for analyzing lexical polysemy knowledge in pre-trained language models and multilingual BERT models by analyzing different sense distributions and controlling for parameters that are highly correlated with polysyntax.
Outcome: The proposed model can be used to analyze lexical polysemy in English, French, Spanish, and Greek and in multilingual BERT.
Probe-Less Probing of BERT’s Layer-Wise Linguistic Knowledge with Masked Word Prediction (2022.naacl-srw)

Copied to clipboard

Challenge: Among studies on localization of linguistic knowledge, it is unclear what information is encoded in each layer.
Approach: They analyze BERT’s layer-wise masked word prediction on an English corpus and find syntactic and semantic information is encoded at different layers for words of different syntaktic categories.
Outcome: The proposed model outperforms state-of-the-art models in many downstream tasks.
Revealing the Dark Secrets of BERT (D19-1)

Copied to clipboard

Challenge: Existing models of BERT-based learning systems are lacking specific mechanisms that contribute to its success.
Approach: They propose to use GLUE tasks to analyze the interpretation of self-attention, which is one of the underlying components of BERT.
Outcome: The proposed model outperforms the regular model on GLUE tasks by disabling attention in certain heads.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations