BERTRAM: Improved Word Embeddings Have Big Impact on Contextualized Model Performance

Schick, Timo; Schütze, Hinrich

doi:10.18653/v1/2020.acl-main.368

Cited by 34 publications

(31 citation statements)

References 26 publications

(61 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…E-BERT-concat. E-BERT-concat combines entity IDs and wordpieces by string concatenation, with the slash symbol as separator (Schick and Schütze, 2019). For example, the wordpiecetokenized input…”

Section: Using Aligned Entity Vectorsmentioning

confidence: 99%

E-BERT: Efficient-Yet-Effective Entity Embeddings for BERT

Poerner

Waltinger

Schütze

2020

Findings of the Association for Computational Linguistics: EMNLP 2020

Self Cite

108

View full text Add to dashboard Cite

We present a novel way of injecting factual knowledge about entities into the pretrained BERT model (Devlin et al., 2019): We align Wikipedia2Vec entity vectors (Yamada et al., 2016) with BERT's native wordpiece vector space and use the aligned entity vectors as if they were wordpiece vectors. The resulting entity-enhanced version of BERT (called E-BERT) is similar in spirit to ERNIE (Zhang et al., 2019) and KnowBert (Peters et al., 2019), but it requires no expensive further pretraining of the BERT encoder. We evaluate E-BERT on unsupervised question answering (QA), supervised relation classification (RC) and entity linking (EL). On all three tasks, E-BERT outperforms BERT and other baselines. We also show quantitatively that the original BERT model is overly reliant on the surface form of entity names (e.g., guessing that someone with an Italian-sounding name speaks Italian), and that E-BERT mitigates this problem.

show abstract

Section: Using Aligned Entity Vectorsmentioning

confidence: 99%

E-BERT: Efficient-Yet-Effective Entity Embeddings for BERT

Poerner

Waltinger

Schütze

2020

Findings of the Association for Computational Linguistics: EMNLP 2020

Self Cite

108

View full text Add to dashboard Cite

show abstract

“…The only exception is RoBERTa-large, which had the lowest initial isotropy. Interestingly, Schick and Schütze (2020a) show that RoBERTa-large outperforms BERT models on tasks designed explicitly for rare words. Moreover, according to common leaderboards (Wang et al, 2019b,a), RoBERTa performs best on downstream tasks among the models we analyzed.…”

Section: Resultsmentioning

confidence: 99%

“…Performance of pretrained language models is inconsistent and tends to decrease when input contains rare words (Schick and Schütze, 2020b,a). Schick and Schütze (2020a) observe that replacing a portion of words in the MNLI (Williams et al, 2018) entailment data set with less frequent synonyms leads to decrease in performance of BERT-base and RoBERTa-large by 30% and 21.8% respectively. 2 After enriching rare words with surface-form features and additional context, Schick and Schütze (2020a) decrease the performance gap to 20.7% for BERT and 17% for RoBERTa, but the gap remains large nonetheless.…”

Section: Introductionmentioning

confidence: 95%

Too Much in Common: Shifting of Embeddings in Transformer Language Models and its Implications

Biś¹,

Podkorytov²,

Liu³

2021

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

View full text Add to dashboard Cite

The success of language models based on the Transformer architecture appears to be inconsistent with observed anisotropic properties of representations learned by such models. We resolve this by showing, contrary to previous studies, that the representations do not occupy a narrow cone, but rather drift in common directions. At any training step, all of the embeddings except for the ground-truth target embedding are updated with gradient in the same direction. Compounded over the training set, the embeddings drift and share common components, manifested in their shape in all the models we have empirically tested. Our experiments show that isotropy can be restored using a simple transformation. 1

show abstract

“…Handling Rare Words. These remain challenging even for large transformer models (Schick and Schütze, 2020). Recent work has explored copying mechanisms and character based generation (Kawakami et al, 2017), with some success.…”

Section: Related Workmentioning

confidence: 99%

Improving Low Compute Language Modeling with In-Domain Embedding Initialisation

Welch

Mihalcea

Kummerfeld

2020

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)

View full text Add to dashboard Cite

Many NLP applications, such as biomedical data and technical support, have 10-100 million tokens of in-domain data and limited computational resources for learning from it. How should we train a language model in this scenario? Most language modeling research considers either a small dataset with a closed vocabulary (like the standard 1 million token Penn Treebank), or the whole web with bytepair encoding. We show that for our target setting in English, initialising and freezing input embeddings using in-domain data can improve language model performance by providing a useful representation of rare words, and this pattern holds across several different domains. In the process, we show that the standard convention of tying input and output embeddings does not improve perplexity when initializing with embeddings trained on in-domain data.

show abstract

BERTRAM: Improved Word Embeddings Have Big Impact on Contextualized Model Performance

Cited by 34 publications

References 26 publications

E-BERT: Efficient-Yet-Effective Entity Embeddings for BERT

E-BERT: Efficient-Yet-Effective Entity Embeddings for BERT

Too Much in Common: Shifting of Embeddings in Transformer Language Models and its Implications

Improving Low Compute Language Modeling with In-Domain Embedding Initialisation

Contact Info

Product

Resources

About