Dominican Republic

Current language models are usually trained using a self-supervised scheme, where the main focus is learning representations at the word or sentence level. However, there has been limited progress in generating useful discourse-level representations. In this work, we propose to use ideas from predictive coding theory to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations. As a result, our proposed approach is able to predict future sentences using explicit top-down connections that operate at the intermediate layers of the network. By experimenting with benchmarks designed to evaluate discourse-related knowledge using pre-trained sentence representations, we demonstrate that our approach improves performance in 6 out of 11 tasks by excelling in discourse relationship detection.

EMNLP 2021

Augmenting BERT-style Models with Predictive Coding to Improve Discourse-level Representations

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

poster

Transformer models are permutation equivariant. To supply the order and type information of the input tokens, position and segment embeddings are usually added to the input. Recent works proposed variations of positional encodings with relative position encodings achieving better performance. Our analysis shows that the gain actually comes from moving positional information to attention layer from the input. Motivated by this, we introduce Decoupled posItional attEntion for Transformers (DIET), a simple yet effective mechanism to encode position and segment information into the Transformer models. The proposed method has faster training and inference time, while achieving competitive performance on GLUE, XTREME and WMT benchmarks. We further generalize our method to long-range transformers and show performance gain.

A Simple and Effective Positional Encoding for Transformers

Relative position embedding (RPE) is a successful method to explicitly and efficaciously encode position information into Transformer models. In this paper, we investigate the potential problems in Shaw-RPE and XL-RPE, which are the most representative and prevalent RPEs, and propose two novel RPEs called Low-level Fine-grained High-level Coarse-grained (LFHC) RPE and Gaussian Cumulative Distribution Function (GCDF) RPE. LFHC-RPE is an improvement of Shaw-RPE, which enhances the perception ability at medium and long relative positions. GCDF-RPE utilizes the excellent properties of the Gaussian function to amend the prior encoding mechanism in XL-RPE. Experimental results on nine authoritative datasets demonstrate the effectiveness of our methods empirically. Furthermore, GCDF-RPE achieves the best overall performance among five different RPEs.

Explore Better Relative Position Embeddings from Encoding Perspective for Transformer Models

We propose an Adversarial Mixing Policy(AMP) to relax the Locally Linear Constraints (LLC) in Mixup without any auxiliary networks. It can be seamlessly integrated into other Mixup variants for its simplicity.

Adversarial Mixing Policy for Relaxing Locally Linear Constraints in Mixup

It is difficult to rank and evaluate the performance of grammatical error correction (GEC) systems, as a sentence can be rewritten in numerous correct ways. A number of GEC metrics have been used to evaluate proposed GEC systems; however, each system relies on either a comparison with one or more reference texts—in what is known as the gold standard for reference-based metrics—or a separate annotated dataset to fine-tune the reference-less metric. Reference-based systems have a low correlation with human judgement, cannot capture all the ways in which a sentence can be corrected, and require substantial work to develop a test dataset. We propose a reference-less GEC evaluation system that is strongly correlated with human judgement, solves the issues related to the use of a reference, and does not need another annotated dataset for fine-tuning. The proposed system relies solely on commonly available tools. Additionally, currently available reference-less metrics do not work properly when part of a sentence is repeated as opposed to reference-based metrics. In our proposed system, we look to address issues inherent in reference-less metrics and reference-based metrics.

Is this the end of the gold standard? A straightforward reference-less grammatical error correction metric

Recent pretrained language models extend from millions to billions of parameters. Thus the need to fine-tune an extremely large pretrained model with a limited training corpus arises in various downstream tasks. In this paper, we propose a straightforward yet effective fine-tuning technique, ChildTuning, which updates a subset of parameters (called child network) of large pretrained models via strategically masking out the gradients of the non-child network during the backward process. Experiments on various downstream tasks in GLUE benchmark show that ChildTuning consistently outperforms the vanilla fine-tuning by 1.5~8.6 average score among four different pretrained models, and surpasses the prior fine-tuning techniques by 0.6~1.3 points. Furthermore, empirical results on domain transfer and task transfer show that ChildTuning can obtain better generalization performance by large margins.

Raise a Child in Large Language Model: Towards Effective and Generalizable Fine-tuning

In this paper, we present a new method for training a writing improvement model adapted to the writer’s first language (L1) that goes beyond grammatical error correction (GEC). Without using annotated training data, we rely solely on pre-trained language models finetuned with parallel corpora of reference translation aligned with machine translation. We evaluate our model with corpora of academic papers written in English by L1 Portuguese and L1 Spanish scholars and a reference corpus of expert academic English. We show that our model is able to address specific L1-influenced writing and more complex linguistic phenomena than existing methods, outperforming what a state-of-the-art GEC system can achieve in this regard. Our code and data are open to other researchers.

Beyond Grammatical Error Correction: Improving L1-influenced research writing in English using pre-trained encoder-decoder models

Researchers use figures to communicate rich, complex information in scientific papers. The captions of these figures are critical to conveying effective messages. However, low-quality figure captions commonly occur in scientific articles and may decrease understanding. In this paper, we propose an end-to-end neural framework to automatically generate informative, high-quality captions for scientific figures. To this end, we introduce SCICAP, a large-scale figure-caption dataset based on computer science arXiv papers published between 2010 and 2020. After pre-processing – including figure-type classification, sub-figure identification, text normalization, and caption text selection – SCICAP contained more than two million figures extracted from over 290,000 papers. We then established baseline models that caption graph plots, the dominant (19.2%) figure type. The experimental results showed both opportunities and steep challenges of generating captions for scientific figures.

SciCap: Generating Captions for Scientific Figures

Byte-pair encoding (BPE) is a ubiquitous algorithm in the subword tokenization process of language models as it provides multiple benefits. However, this process is solely based on pre-training data statistics, making it hard for the tokenizer to handle infrequent spellings. On the other hand, though robust to misspellings, pure character-level models often lead to unreasonably long sequences and make it harder for the model to learn meaningful words. To alleviate these challenges, we propose a character-based subword module (char2subword) that learns the subword embedding table in pre-trained models like BERT. Our char2subword module builds representations from characters out of the subword vocabulary, and it can be used as a drop-in replacement of the subword embedding table. The module is robust to character-level alterations such as misspellings, word inflection, casing, and punctuation. We integrate it further with BERT through pre-training while keeping BERT transformer parameters fixed--and thus, providing a practical method. Finally, we show that incorporating our module to mBERT significantly improves the performance on the social media linguistic code-switching evaluation (LinCE) benchmark.

Char2Subword: Extending the Subword Embedding Space Using Robust Character Compositionality

Code-switching (CS), a ubiquitous phe-nomenon due to the ease of communicationit offers in multilingual communities stillremains an understudied problem in languageprocessing. The primary reasons behind thisare: (1) minimal efforts in leveraging largepretrained multilingual models, and (2) thelack of annotated data.The distinguishingcase of low performance of multilingualmodels in CS is the intra-sentence mixingof languages leading to switch points. Wefirst benchmark two sequence labeling tasks –POS and NER on 4 different language pairswith a suite of pretrained models to identifythe problems and select the best performingmodel, char-BERT, among them (addressing(1)). We then propose a self training methodto repurpose the existing pretrained modelsusing a switch-point bias by leveragingunannotated data (addressing (2)). We finallydemonstrate that our approach performs wellon both tasks by reducing the gap between theswitch point performance while retaining theoverall performance on two distinct languagepairs in both the tasks

Switch Point biased Self-Training: Re-purposing Pretrained Models for Code-Switching

We present AEDA (An Easier Data Augmentation) technique to help improve the performance on text classification tasks. AEDA includes only random insertion of punctuation marks into the original text. This is an easier technique to implement for data augmentation than EDA method (Wei et al. 2019) with which we compare our results. In addition, it keeps the order of the words while changing their positions in the sentence leading to a better generalized performance. Furthermore, the deletion operation in EDA can cause loss of information which, in turn, misleads the network, whereas AEDA preserves all the input information. Following the baseline, we perform experiments on five different datasets for text classification. We show that using the AEDA-augmented data for training, the models show superior performance compared to using the EDA-augmented data in all five datasets.

Augmenting BERT-style Models with Predictive Coding to Improve Discourse-level Representations

Downloads

Next from EMNLP 2021

A Simple and Effective Positional Encoding for Transformers

Similar lecture

Self-training Improves Pre-training for Few-shot Learning in Task-oriented Dialog Systems