Dominican Republic

While large language models a la BERT are used ubiquitously in NLP, pretraining them is considered a luxury that only a few well-funded industry labs can afford. How can one train such models with a more modest budget? We present a recipe for pretraining a masked language model in 24 hours using a single low-end deep learning server. We demonstrate that through a combination of software optimizations, design choices, and hyperparameter tuning, it is possible to produce models that are competitive with BERT-base on GLUE tasks at a fraction of the original pretraining cost.

EMNLP 2021

How to Train BERT with an Academic Budget

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.

technical paper

Transformers have outperformed recurrent neural networks (RNNs) in natural language generation. This comes with a significant computational overhead, as the attention mechanism scales with a quadratic complexity in sequence length. Efficient transformer variants have received increasing interest from recent works. Among them, a linear-complexity recurrent variant has proven well suited for autoregressive generation. It approximates the softmax attention with randomized or heuristic feature maps, but can be difficult to train or yield suboptimal accuracy. This work aims to convert a pretrained transformer into its efficient recurrent counterpart, improving the efficiency while retaining the accuracy. Specifically, we propose a swap-then-finetune procedure: in an off-the-shelf pretrained transformer, we replace the softmax attention with its linear-complexity recurrent alternative and then finetune. With a learned feature map, our approach provides an improved tradeoff between efficiency and accuracy over the standard transformer and other recurrent variants. We also show that the finetuning process needs lower training cost than training these recurrent variants from scratch. As many recent models for natural language tasks are increasingly dependent on large-scale pretrained transformers, this work presents a viable approach to improving inference efficiency without repeating the expensive pretraining process.

Finetuning Pretrained Transformers into RNNs

Various temporal knowledge graph (KG) completion models have been proposed in the recent literature. The models usually contain two parts, a temporal embedding layer and a score function derived from existing static KG modeling approaches. Since the approaches differ along several dimensions, including different score functions and training strategies, the individual contributions of different temporal embedding techniques to model performance are not always clear. In this work, we systematically study six temporal embedding approaches and empirically quantify their performance across a wide range of configurations with about 4000 experiments and 19000 GPU hours. We classify the temporal embeddings into two classes: (1) timestamp embeddings and (2) time-dependent entity embeddings. Despite the common belief that the latter is more expressive, an extensive experimental study shows that timestamp embeddings can achieve on-par or even better performance with significantly fewer parameters. Moreover, we find that when trained appropriately, the relative performance differences between various temporal embeddings often shrink and sometimes even reverse when compared to prior results. For example, TTransE (Leblay and Chekol, 2018), one of the first temporal KG models, can outperform more recent architectures on ICEWS datasets. To foster further research, we provide the first unified open-source framework for temporal KG completion models with full composability, where temporal embeddings, score functions, loss functions, regularizers, and the explicit modeling of reciprocal relations can be combined arbitrarily.

Time-dependent Entity Embedding is not All You Need: A Re-evaluation of Temporal Knowledge Graph Completion Models under a Unified Framework

Information Extraction (IE) aims to extract structural information from unstructured texts. In practice, long-tailed distributions caused by the selection bias of a dataset, may lead to incorrect correlations, also known as spurious correlations, between entities and labels in the conventional likelihood models. This motivates us to propose counterfactual IE (CFIE), a novel framework that aims to uncover the main causalities behind data in the view of causal inference. Specifically, 1) we first introduce a unified structural causal model (SCM) for various IE tasks, describing the relationships among variables; 2) with our SCM, we then generate counterfactuals based on an explicit language structure to better calculate the direct causal effect during the inference stage; 3) we further propose a novel debiasing approach to yield more robust predictions. Experiments on three IE tasks across five public datasets show the effectiveness of our CFIE model in mitigating the spurious correlation issues.

Uncovering Main Causalities for Long-tailed Information Extraction

Pretrained Language Models (PTLMs) yield state-of-the-art performance on many Natural Language Processing tasks, including syntax, semantics and commonsense reasoning. In this paper, we focus on identifying to what extent do PTLMs capture semantic attributes and their values, e.g. the relation between rich and high net worth. We use PTLMs to predict masked tokens using patterns and lists of items from Wikidata in order to verify how likely PTLMs encode semantic attributes along with their values. Such inferences based on seman- tics are intuitive for us humans as part of our language understanding. Since PTLMs are trained on large amount of Wikipedia data, we would assume that they can generate similar predictions. However, our findings reveal that PTLMs perform still much worse than humans on this task. We show an analysis which ex- plains how to exploit our methodology to inte- grate better context and semantics into PTLMs using knowledge bases.

Probing Pre-trained Language Models for Semantic Attributes and their Values

Understanding when a text snippet does not provide a sought after information is an essential part of natural language understanding. Recent work (SQuAD 2.0, Rajpurkar et al., 2018) has attempted to make some progress in this direction by enriching the SQuAD dataset for the Extractive QA task with unanswerable
questions. However, as we show, the performance of a top system trained on SQuAD 2.0 drops considerably in out-of-domain scenarios, limiting its use in practical situations. In order to study this we build an out-of-domain corpus, focusing on simple event-based questions and distinguish between two types of IDK questions: competitive questions, where the context includes an entity of the same
type as the expected answer, and simpler, non-competitive questions where there is no entity of the same type in the context. We find that SQuAD 2.0-based models fail even in the case of the simpler questions. We then analyze the similarities and differences between the IDK phenomenon in Extractive QA and the Recognizing Textual Entailments task (RTE, Dagan et al., 2013) and investigate the extent to which the latter can be used to improve the performance.

Do We Know What We Don’t Know? Studying Unanswerable Questions beyond SQuAD 2.0

Linguistic representations derived from text alone have been criticized for their lack of grounding, i.e., connecting words to their meanings in the physical world. Vision-and-Language (VL) models, trained jointly on text and image or video data, have been offered as a response to such criticisms. However, while VL pretraining has shown success on multimodal tasks such as visual question answering, it is not yet known how the internal linguistic representations themselves compare to their text-only counterparts. This paper compares the semantic representations learned via VL vs. text-only pretraining for two recent VL models using a suite of analyses (clustering, probing, and performance on a commonsense question answering task) in a language-only setting. We find that the multimodal models fail to significantly outperform the text-only variants, suggesting that future work is required if multimodal pretraining is to be pursued as a means of improving NLP in general.

Does Vision-and-Language Pretraining Improve Lexical Grounding?

To alleviate human efforts from obtaining large-scale annotations, Semi-Supervised Relation Extraction methods aim to leverage unlabeled data in addition to learning from limited samples. Existing self-training methods suffer from the gradual drift problem, where noisy pseudo labels on unlabeled data are incorporated during training. To alleviate the noise in pseudo labels, we propose a method called MetaSRE, where a Relation Label Generation Network generates quality assessment on pseudo labels by (meta) learning from the successful and failed attempts on Relation Classification Network as an additional meta-objective. To reduce the influence of noisy pseudo labels, MetaSRE adopts a pseudo label selection and exploitation scheme which assesses pseudo label quality on unlabeled samples and only exploits high-quality pseudo labels in a self-training fashion to incrementally augment labeled samples for both robustness and accuracy. Experimental results on two public datasets demonstrate the effectiveness of the proposed approach.

Semi-supervised Relation Extraction via Incremental Meta Self-Training

Computational models of human language often involve combinatorial problems. For instance, a probabilistic parser may marginalize over exponentially many trees to make predictions. Algorithms for such problems often employ dynamic programming and are not always unique. Finding one with optimal asymptotic runtime can be unintuitive, time-consuming, and error-prone. Our work aims to automate this laborious process. Given an initial correct declarative program, we search for a sequence of semantics-preserving transformations to improve its running time as much as possible. To this end, we describe a set of program transformations, a simple metric for assessing the efficiency of a transformed program, and a heuristic search procedure to improve this metric. We show that in practice, automated search -- like the mental search performed by human programmers -- can find substantial improvements to the initial program. Empirically, we show that many common speed-ups described in the NLP literature could have been discovered automatically by our system.

Searching for More Efficient Dynamic Programs

Phrase representations derived from BERT often do not exhibit complex phrasal compositionality, as the model relies instead on lexical similarity to determine semantic relatedness. In this paper, we propose a contrastive fine-tuning objective that enables BERT to produce more powerful phrase embeddings. Our approach (Phrase-BERT) relies on a dataset of diverse phrasal paraphrases, which is automatically generated using a paraphrase generation model, as well as a large-scale dataset of phrases in context mined from the Books3 corpus. Phrase-BERT outperforms baselines across a variety of phrase-level similarity tasks, while also demonstrating increased lexical diversity between nearest neighbors in the vector space. Finally, as a case study, we show that Phrase-BERT embeddings can be easily integrated with a simple autoencoder to build a phrase-based neural topic model that interprets topics as mixtures of words and phrases by performing a nearest neighbor search in the embedding space. Crowdsourced evaluations demonstrate that this phrase-based topic model produces more coherent and meaningful topics than baseline word and phrase-level topic models, further validating the utility of Phrase-BERT.

Phrase-BERT: Improved Phrase Embeddings from BERT with an Application to Corpus Exploration

This study investigates the applicability of semantic change detection methods in descriptively oriented linguistic research. It specifically focuses on contact-induced semantic shifts in Quebec English. We contrast synchronic data from different regions in order to identify the meanings that are specific to Quebec and potentially related to language contact. Type-level embeddings are used to detect new semantic shifts, and token-level embeddings to isolate regionally specific occurrences. We introduce a new 80-item test set and conduct both quantitative and qualitative evaluations.
We demonstrate that diachronic word embedding methods can be applied to contactinduced semantic shifts observed in synchrony, obtaining results comparable to the state of the art on similar tasks in diachrony. However, we show that encouraging evaluation results do not translate to practical value in detecting new semantic shifts. Finally, our application of token-level embeddings accelerates manual data exploration and provides an efficient way of scaling up sociolinguistic analyses.

Downloads

Next from EMNLP 2021

Finetuning Pretrained Transformers into RNNs

Similar lecture

Efficient Domain Adaptation of Language Models via Adaptive Tokenization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES