NLP research has explored different neural model architectures and sizes, datasets, training objectives, and transfer learning techniques. However, the choice of optimizer during training has not been explored as extensively. Typically, some variant of Stochastic Gradient Descent (SGD) is employed, selected among numerous variants, using unclear criteria, often with minimal or no tuning of the optimizer’s hyperparameters. Experimenting with five GLUE datasets, two models (DistilBERT and DistilRoBERTa), and seven popular optimizers (SGD, SGD with Momentum, Adam, AdaMax, Nadam, AdamW, and AdaBound), we find that when the hyperparameters of the optimizers are tuned, there is no substantial difference in test performance across the five more elaborate (adaptive) optimizers, despite differences in training loss. Furthermore, tuning just the learning rate is in most cases as good as tuning all the hyperparameters. Hence, we recommend picking any of the best-behaved adaptive optimizers (e.g., Adam) and tuning only its learning rate. When no hyperparameter can be tuned, SGD with Momentum is the best choice.

Prompting Large Language Models (LLMs) performs impressively in zero- and few-shot settings. Hence, small and medium-sized enterprises (SMEs) that cannot afford the cost of creating large task-specific training datasets, but also the cost of pretraining their own LLMs, are increasingly turning to third-party services that allow them to prompt LLMs. However, such services currently require a payment per call, which becomes a significant operating expense (OpEx). Furthermore, customer inputs are often very similar over time, hence SMEs end-up prompting LLMs with very similar instances. We propose a framework that allows reducing the calls to LLMs by caching previous LLM responses and using them to train a local inexpensive model on the SME side. The framework includes criteria for deciding when to trust the local model or call the LLM, and a methodology to tune the criteria and measure the tradeoff between performance and cost. For experimental purposes, we instantiate our framework with two LLMs, GPT-3.5 or GPT-4, and two inexpensive students, a k-NN classifier or a Multi-Layer Perceptron, using two common business tasks, intent recognition and sentiment analysis. Experimental results indicate that significant OpEx savings can be obtained with only slightly lower performance.

In the era of billion-parameter-sized Language Models (LMs), start-ups have to follow trends and adapt their technology accordingly. Nonetheless, there are open challenges since the development and deployment of large models comes with a need for high computational resources and has economical consequences. In this work, we follow the steps of the R&D group of a modern legal-tech start-up and present important insights on model development and deployment. We start from ground zero by pre-training multiple domain-specific multi-lingual LMs which are a better fit to contractual and regulatory text compared to the available alternatives (XLM-R). We present benchmark results of such models in a half-public half-private legal benchmark comprising 5 downstream tasks showing the impact of larger model size. Lastly, we examine the impact of a full-scale pipeline for model compression which includes: a) Parameter Pruning, b) Knowledge Distillation, and c) Quantization: The resulting models are much more efficient without sacrificing performance at large.

We study the effect of seven data augmentation (DA) methods in factoid question answering, focusing on the biomedical domain, where obtaining training instances is particularly difficult. We experiment with data from the BIOASQ challenge, which we augment with training instances obtained from an artificial biomedical machine reading comprehension dataset, or via back-translation, information retrieval, word substitution based on WORD2VEC embeddings, or masked language modeling, question generation, or extending the given passage with additional context. We show that DA can lead to very significant performance gains, even when using large pre-trained Transformers, contributing to a broader discussion of if/when DA benefits large pre-trained models. One of the simplest DA methods, WORD2VEC-based word substitution, performed best and is recommended. We release our artificial training instances and code.

Publicly traded companies are required to submit periodic reports with eXtensive Business Reporting Language (XBRL) word-level tags. Manually tagging the reports is tedious and costly. We, therefore, introduce XBRL tagging as a new entity extraction task for the financial domain and release FiNER-139, a dataset of 1.1M sentences with gold XBRL tags. Unlike typical entity extraction datasets, FiNER-139 uses a much larger label set of 139 entity types. Most annotated tokens are numeric, with the correct tag per token depending mostly on context, rather than the token itself. We show that subword fragmentation of numeric expressions harms BERT’s performance, allowing word-level BILSTMs to perform better. To improve BERT’s performance, we propose two simple and effective solutions that replace numeric expressions with pseudo-tokens reflecting original token shapes and numeric magnitudes. We also experiment with FIN-BERT, an existing BERT model for the financial domain, and release our own BERT (SEC-BERT), pre-trained on financial filings, which performs best. Through data and error analysis, we finally identify possible limitations to inspire future work on XBRL tagging.

Interpretability or explainability is an emerging research field in NLP. From a user-centric point of view, the goal is to build models that provide proper justification for their decisions, similar to those of humans, by requiring the models to satisfy additional constraints. To this end, we introduce a new application on legal text where, contrary to mainstream literature targeting word-level rationales, we conceive rationales as selected paragraphs in multi-paragraph structured court cases. We also release a new dataset comprising European Court of Human Rights cases, including annotations for paragraph-level rationales. We use this dataset to study the effect of already proposed rationale constraints, i.e., sparsity, continuity, and comprehensiveness, formulated as regularizers. Our findings indicate that some of these constraints are not beneficial in paragraph-level rationale extraction, while others need re-formulation to better handle the multi-label nature of the task we consider. We also introduce a new constraint, singularity, which further improves the quality of rationales, even compared with noisy rationale supervision. Experimental results indicate that the newly introduced task is very challenging and there is a large scope for further research.

Major scandals in corporate history have urged the need for regulatory compliance, where organizations need to ensure that their controls (processes) comply with relevant laws, regulations, and policies. However, keeping track of the constantly changing legislation is difficult, thus organizations are increasingly adopting Regulatory Technology (RegTech) to facilitate the process. To this end, we introduce regulatory information retrieval (REG-IR), an application of document-to-document information retrieval (DOC2DOC IR), where the query is an entire document making the task more challenging than traditional IR where the queries are short. Furthermore, we compile and release two datasets based on the relationships between EU directives and UK legislation. We experiment on these datasets using a typical two-step pipeline approach comprising a pre-fetcher and a neural re-ranker. Experimenting with various pre-fetchers from BM25 to k nearest neighbors over representations from several BERT models, we show that fine-tuning a BERT model on an in-domain classification task produces the best representations for IR. We also show that neural re-rankers under-perform due to contradicting supervision, i.e., similar query-document pairs with opposite labels. Thus, they are biased towards the pre-fetcher’s score. Interestingly, applying a date filter further improves the performance, showcasing the importance of the time dimension.

We release EDGAR-CORPUS, a novel corpus comprising annual reports from all the publicly traded companies in the US spanning a period of more than 25 years. To the best of our knowledge, EDGAR-CORPUS is the largest financial NLP corpus available to date. All the reports are downloaded, split into their corresponding items (sections), and provided in a clean, easy-to-use JSON format. We use EDGAR-CORPUS to train and release EDGAR-W2V, which are WORD2VEC embeddings for the financial domain. We employ these embeddings in a battery of financial NLP tasks and showcase their superiority over generic GloVe embeddings and other existing financial word embeddings. We also open-source EDGAR-CRAWLER, a toolkit that facilitates downloading and extracting future annual reports.

BERT has achieved impressive performance in several NLP tasks. However, there has been limited investigation on its adaptation guidelines in specialised domains. Here we focus on the legal domain, where we explore several approaches for applying BERT models to downstream legal tasks, evaluating on multiple datasets. Our findings indicate that the previous guidelines for pre-training and fine-tuning, often blindly followed, do not always generalize well in the legal domain. Thus we propose a systematic investigation of the available strategies when applying BERT in specialised domains. These are: (a) use the original BERT out of the box, (b) adapt BERT by additional pre-training on domain-specific corpora, and (c) pre-train BERT from scratch on domain-specific corpora. We also propose a broader hyper-parameter search space when fine-tuning for downstream tasks and we release LEGAL-BERT, a family of BERT models intended to assist legal NLP research, computational law, and legal technology applications.

Although BERT is widely used by the NLP community, little is known about its inner workings. Several attempts have been made to shed light on certain aspects of BERT, often with contradicting conclusions. A much raised concern focuses on BERT’s over-parameterization and under-utilization issues. To this end, we propose o novel approach to fine-tune BERT in a structured manner. Specifically, we focus on Large Scale Multilabel Text Classification (LMTC) where documents are assigned with one or more labels from a large predefined set of hierarchically organized labels. Our approach guides specific BERT layers to predict labels from specific hierarchy levels. Experimenting with two LMTC datasets we show that this structured fine-tuning approach not only yields better classification results but also leads to better parameter utilization.

Large-scale Multi-label Text Classification (LMTC) has a wide range of Natural Language Processing (NLP) applications and presents interesting challenges. First, not all labels are well represented in the training set, due to the very large label set and the skewed label distributions of datasets. Also, label hierarchies and differences in human labelling guidelines may affect graph-aware annotation proximity. Finally, the label hierarchies are periodically updated, requiring LMTC models capable of zero-shot generalization. Current state-of-the-art LMTC models employ Label-Wise Attention Networks (LWANs), which (1) typically treat LMTC as flat multi-label classification; (2) may use the label hierarchy to improve zero-shot learning, although this practice is vastly understudied; and (3) have not been combined with pre-trained Transformers (e.g. BERT), which have led to state-of-the-art results in several NLP benchmarks. Here, for the first time, we empirically evaluate a battery of LMTC methods from vanilla LWANs to hierarchical classification approaches and transfer learning, on frequent, few, and zero-shot learning on three datasets from different domains. We show that hierarchical methods based on Probabilistic Label Trees (PLTs) outperform LWANs. Furthermore, we show that Transformer-based approaches outperform the state-of-the-art in two of the datasets, and we propose a new state-of-the-art method which combines BERT with LWAN. Finally, we propose new models that leverage the label hierarchy to improve few and zero-shot learning, considering on each dataset a graph-aware annotation proximity measure that we introduce.

We consider Large-Scale Multi-Label Text Classification (LMTC) in the legal domain. We release a new dataset of 57k legislative documents from EUR-LEX, annotated with ∼4.3k EUROVOC labels, which is suitable for LMTC, few- and zero-shot learning. Experimenting with several neural classifiers, we show that BIGRUs with label-wise attention perform better than other current state of the art methods. Domain-specific WORD2VEC and context-sensitive ELMO embeddings further improve performance. We also find that considering only particular zones of the documents is sufficient. This allows us to bypass BERT’s maximum text length limit and fine-tune BERT, obtaining the best results in all but zero-shot learning cases.

We propose SUM-QE, a novel Quality Estimation model for summarization based on BERT. The model addresses linguistic quality aspects that are only indirectly captured by content-based approaches to summary evaluation, without involving comparison with human references. SUM-QE achieves very high correlations with human ratings, outperforming simpler models addressing these linguistic aspects. Predictions of the SUM-QE model can be used for system development, and to inform users of the quality of automatically produced summaries and other types of generated text.

We consider the task of Extreme Multi-Label Text Classification (XMTC) in the legal domain. We release a new dataset of 57k legislative documents from EURLEX, the European Union’s public document database, annotated with concepts from EUROVOC, a multidisciplinary thesaurus. The dataset is substantially larger than previous EURLEX datasets and suitable for XMTC, few-shot and zero-shot learning. Experimenting with several neural classifiers, we show that BIGRUs with self-attention outperform the current multi-label state-of-the-art methods, which employ label-wise attention. Replacing CNNs with BIGRUs in label-wise attention networks leads to the best overall performance.

Experimenting with a new dataset of 1.6M user comments from a Greek news portal and existing datasets of EnglishWikipedia comments, we show that an RNN outperforms the previous state of the art in moderation. A deep, classification-specific attention mechanism improves further the overall performance of the RNN. We also compare against a CNN and a word-list baseline, considering both fully automatic and semi-automatic moderation.

Experimenting with a dataset of approximately 1.6M user comments from a Greek news sports portal, we explore how a state of the art RNN-based moderation method can be improved by adding user embeddings, user type embeddings, user biases, or user type biases. We observe improvements in all cases, with user embeddings leading to the biggest performance gains.

Experimenting with a new dataset of 1.6M user comments from a news portal and an existing dataset of 115K Wikipedia talk page comments, we show that an RNN operating on word embeddings outpeforms the previous state of the art in moderation, which used logistic regression or an MLP classifier with character or word n-grams. We also compare against a CNN operating on word embeddings, and a word-list baseline. A novel, deep, classificationspecific attention mechanism improves the performance of the RNN further, and can also highlight suspicious words for free, without including highlighted words in the training data. We consider both fully automatic and semi-automatic moderation.

We introduce the BIOASQ suite, a set of open-source Web tools for the creation, assessment and community-driven improvement of question answering benchmarks. The suite comprises three main tools: (1) the annotation tool supports the creation of benchmarks per se. In particular, this tool allows a team of experts to create questions and answers as well as to annotate the latter with documents, document snippets, RDF triples and ontology concepts. While the creation of questions is supported by different views and contextual information pertaining to the same question, the creation of answers is supported by the integration of several search engines and context information to facilitate the retrieval of the said answers as well as their annotation. (2) The assessment tool allows comparing several answers to the same question. Therewith, it can be used to assess the inter-annotator agreement as well as to manually evaluate automatically generated answers. (3) The third tool in the suite, the social network, aims to ensure the sustainability and iterative improvement of the benchmark by empowering communities of experts to provide insights on the questions in the benchmark. The BIOASQ suite has already been used successfully to create the 311 questions comprised in the BIOASQ question answering benchmark. It has also been evaluated by the experts who used it to create the BIOASQ benchmark.