COMPARATIVE ANALYSIS OF TRANSFORMER BASED LANGUAGE MODELS

arXiv (Cornell University), 2023

Transformer-based pretrained models like BERT, GPT-2 and T5 have been finetuned for a large number of natural language processing (NLP) tasks, and have been shown to be very effective. However, while finetuning, what changes across layers in these models with respect to pretrained checkpoints is under-studied. Further, how robust are these models to perturbations in input text? Does the robustness vary depending on the NLP task for which the models have been finetuned? While there exists some work on studying the robustness of BERT finetuned for a few NLP tasks, there is no rigorous study that compares this robustness across encoder only, decoder only and encoderdecoder models. In this paper, we characterize changes between pretrained and finetuned language model representations across layers using two metrics: CKA and STIR. Further, we study the robustness of three language models (BERT, GPT-2 and T5) with eight different text perturbations on classification tasks from the General Language Understanding Evaluation (GLUE) benchmark, and generation tasks like summarization, free-form generation and question generation. GPT-2 representations are more robust than BERT and T5 across multiple types of input perturbation. Although models exhibit good robustness broadly, dropping nouns, verbs or changing characters are the most impactful. Overall, this study provides valuable insights into perturbation-specific weaknesses of popular Transformer-based models, which should be kept in mind when passing inputs. We make the code and models publicly available 1 .

TELKOMNIKA Telecommunication Computing Electronics and Control, 2024

This review provides a concise overview of key transformer-based language models, including bidirectional encoder representations from transformers (BERT), generative pre-trained transformer 3 (GPT-3), robustly optimized BERT pretraining approach (RoBERTa), a lite BERT (ALBERT), text-to-text transfer transformer (T5), generative pre-trained transformer 4 (GPT-4), and extra large neural network (XLNet). These models have significantly advanced natural language processing (NLP) capabilities, each bringing unique contributions to the field. We delve into BERT's bidirectional context understanding, GPT-3's versatility with 175 billion parameters, and RoBERTa's optimization of BERT. ALBERT emphasizes model efficiency, T5 introduces a text-to-text framework, and GPT-4, with 170 trillion parameters, excels in multimodal tasks. Safety considerations are highlighted, especially in GPT-4. Additionally, XL-Net's permutation-based training achieves bidirectional context understanding. The motivations, advancements, and challenges of these models are explored, offering insights into the evolving landscape of large-scale language models. This is an open access article under the CC BY-SA license.

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)

Transformer-based language models (TLMs), such as BERT, ALBERT and GPT-3, have shown strong performance in a wide range of NLP tasks and currently dominate the field of NLP. However, many researchers wonder whether these models can maintain their dominance forever. Of course, we do not have answers now, but, as an attempt to find better neural architectures and training schemes, we pretrain a simple CNN using a GAN-style learning scheme and Wikipedia data, and then integrate it with standard TLMs. We show that on the GLUE tasks, the combination of our pretrained CNN with ALBERT outperforms the original ALBERT and achieves a similar performance to that of SOTA. Furthermore, on open-domain QA (Quasar-T and SearchQA), the combination of the CNN with ALBERT or RoBERTa achieved stronger performance than SOTA and the original TLMs. We hope that this work provides a hint for developing a novel strong network architecture along with its training scheme. Our source code and models are available at https://github.com/nict-wisdom/bertac.

Machine Learning and Knowledge Discovery in Databases. Research Track, 2021

Recent advances in neural architectures, such as the Transformer, coupled with the emergence of large-scale pre-trained models such as BERT, have revolutionized the field of Natural Language Processing (NLP), pushing the state of the art for a number of NLP tasks. A rich family of variations of these models has been proposed, such as RoBERTa, ALBERT, and XLNet, but fundamentally, they all remain limited in their ability to model certain kinds of information, and they cannot cope with certain information sources, which was easy for preexisting models. Thus, here we aim to shed light on some important theoretical limitations of pre-trained BERT-style models that are inherent in the general Transformer architecture. First, we demonstrate in practice on two general types of tasks-segmentation and segment labeling-and on four datasets that these limitations are indeed harmful and that addressing them, even in some very simple and naïve ways, can yield sizable improvements over vanilla RoBERTa and XLNet models. Then, we offer a more general discussion on desiderata for future additions to the Transformer architecture that would increase its expressiveness, which we hope could help in the design of the next generation of deep NLP architectures.

Interspeech 2019

We explore deep autoregressive Transformer models in language modeling for speech recognition. We focus on two aspects. First, we revisit Transformer model configurations specifically for language modeling. We show that well configured Transformer models outperform our baseline models based on the shallow stack of LSTM recurrent neural network layers. We carry out experiments on the open-source LibriSpeech 960hr task, for both 200K vocabulary word-level and 10K byte-pair encoding subword-level language modeling. We apply our wordlevel models to conventional hybrid speech recognition by lattice rescoring, and the subword-level models to attention based encoder-decoder models by shallow fusion. Second, we show that deep Transformer language models do not require positional encoding. The positional encoding is an essential augmentation for the self-attention mechanism which is invariant to sequence ordering. However, in autoregressive setup, as is the case for language modeling, the amount of information increases along the position dimension, which is a positional signal by its own. The analysis of attention weights shows that deep autoregressive selfattention models can automatically make use of such positional information. We find that removing the positional encoding even slightly improves the performance of these models.

arXiv (Cornell University), 2020

There has been significant progress in recent years in the field of Natural Language Processing thanks to the introduction of the Transformer architecture. Current state-of-the-art models, via a large number of parameters and pre-training on massive text corpus, have shown impressive results on several downstream tasks. Many researchers have studied previous (non-Transformer) models to understand their actual behavior under different scenarios, showing that these models are taking advantage of clues or failures of datasets and that slight perturbations on the input data can severely reduce their performance. In contrast, recent models have not been systematically tested with adversarial-examples in order to show their robustness under severe stress conditions. For that reason, this work evaluates three Transformer-based models (RoBERTa, XLNet, and BERT) in Natural Language Inference (NLI) and Question Answering (QA) tasks to know if they are more robust or if they have the same flaws as their predecessors. As a result, our experiments reveal that RoBERTa, XLNet and BERT are more robust than recurrent neural network models to stress tests for both NLI and QA tasks. Nevertheless, they are still very fragile and demonstrate various unexpected behaviors, thus revealing that there is still room for future improvement in this field.

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021

In human-level NLP tasks, such as predicting mental health, personality, or demographics, the number of observations is often smaller than the standard 768+ hidden state sizes of each layer within modern transformer-based language models, limiting the ability to effectively leverage transformers. Here, we provide a systematic study on the role of dimension reduction methods (principal components analysis, factorization techniques, or multi-layer auto-encoders) as well as the dimensionality of embedding vectors and sample sizes as a function of predictive performance. We first find that fine-tuning large models with a limited amount of data pose a significant difficulty which can be overcome with a pre-trained dimension reduction regime. RoBERTa consistently achieves top performance in humanlevel tasks, with PCA giving benefit over other reduction methods in better handling users that write longer texts. Finally, we observe that a majority of the tasks achieve results comparable to the best performance with just 1 12 of the embedding dimensions.

2019

Although n-gram language models (LMs) have been outperformed by the state-of-the-art neural LMs, they are still widely used in speech recognition due to its high efficiency in inference. In this paper, we demonstrate that n-gram LM can be improved by neural LMs through a text generation based data augmentation method. In contrast to previous approaches, we employ a large-scale general domain pre-training followed by in-domain fine-tuning strategy to construct deep Transformer based neural LMs. Large amount of in-domain text data is generated with the well trained deep Transformer to construct new n-gram LMs, which are then interpolated with baseline n-gram systems. Empirical studies on different speech recognition tasks show that the proposed approach can effectively improve recognition accuracy. In particular, our proposed approach brings significant relative word error rate reduction up to 6.0% for domains with limited in-domain data.

2020 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 2020

Transfer learning and Transformer-based language models play important roles in modern natural language processing research community. In this paper, we propose Transformer model's fine-tuning and data augmentation (TMFTDA) techniques for conversational texts and noisy usergenerated content. We use two NTCIR-15 tasks, namely the first Dialogue Evaluation (DialEval-1) task and the second Numeral Attachment in Financial Tweets (FinNum-2) task, to evaluate the efficacy of TMFTDA. Experimental results show that TMFTDA substantially outperforms the baselines model of Bidirectional Long Short-Term Memory (Bi-LSTM) in multi-turn dialogue system evaluation at DialEval-1's Dialogue Quality (DQ) and Nugget Detection (ND) subtasks. Moreover, TMFTDA performs to a satisfactory level at FinNum-2 with a model of Cross-lingual Language Models using a Robustly Optimized BERT Pretraining Approach (XLM-RoBERTa). The research contribution of this paper is that, we help shed some light on the usefulness of TMFTDA, for conversational texts and noisy user-generated content in social media text analytics.

2021

Large pretrained language models using the transformer neural network architecture are becoming a dominant methodology for many natural language processing tasks, such as question answering, text classification, word sense disambiguation, text completion and machine translation. Commonly comprising hundreds of millions of parameters, these models offer state-of-the-art performance, but at the expense of interpretability. The attention mechanism is the main component of transformer networks. We present AttViz, a method for exploration of self-attention in transformer networks, which can help in explanation and debugging of the trained models by showing associations between text tokens in an input sequence. We show that existing deep learning pipelines can be explored with AttViz, which offers novel visualizations of the attention heads and their aggregations. We implemented the proposed methods in an online toolkit and an offline library. Using examples from news analysis, we demonst...