Abstract As NLP models are used by a growing number of end-users, an area of increasing importance is NLP Security (NLPSec): assessing the vulnerability of models to malicious attacks and developing comprehensive countermeasures against them. While work at the intersection of NLP and cybersecurity has the potential to create safer NLP for all, accidental oversights can result in tangible harm (e.g., breaches of privacy or proliferation of malicious models). In this emerging field, however, the research ethics of NLP have not yet faced many of the long-standing conundrums pertinent to cybersecurity, until now. We thus examine contemporary works across NLPSec, and explore their engagement with cybersecurity’s ethical norms. We identify trends across the literature, ultimately finding alarming gaps on topics like harm minimization and responsible disclosure. To alleviate these concerns, we provide concrete recommendations to help NLP researchers navigate this space more ethically, bridging the gap between traditional cybersecurity and NLP ethics, which we frame as “white hat NLP”. The goal of this work is to help cultivate an intentional culture of ethical research for those working in NLP Security.

Few-shot classification is a powerful technique, but training requires substantial computing power and data.We propose an efficient method with small model sizes and less training data with only 2-8 training instances per class.Our proposed method, AncSetFit, targets lowdata scenarios by anchoring the task and label information through sentence embeddings in fine-tuning a Sentence Transformer model.It uses contrastive learning and a triplet loss to enforce training instances of a class to be closest to its own textual semantic label information in the embedding space -and thereby learning to embed different class instances more distinct.AncSetFit obtains strong performance in data-sparse scenarios compared to existing methods across SST-5, Emotion detection, and AG News data, even with just two examples per class.

Prediction of a $\textbf{varying number}$ of $\textbf{ordered clusters}$ from sets of $\textbf{any cardinality}$ is a challenging task for neural networks, combining elements of set representation, clustering and learning to order. This task arises in many diverse areas, ranging from medical triage and early discharge, through machine part management and multi-channel signal analysis for petroleum exploration to product catalog structure prediction. This paper focuses on that last area, which exemplifies a number of challenges inherent to adaptive ordered clustering, referred to further as the eponymous $\textit{Catalog Problem}$. These include learning variable cluster constraints, exhibiting relational reasoning and managing combinatorial complexity. Despite progress in both neural clustering and set-to-sequence methods, no joint, fully differentiable model exists to-date. We develop such a modular architecture, referred to further as Neural Ordered Clusters (NOC), enhance it with a specific mechanism for learning cluster-level cardinality constraints, and provide a robust comparison of its performance in relation to alternative models. We test our method on three datasets, including synthetic catalog structures and PROCAT, a dataset of real-world catalogs consisting of over 1.5M products, achieving state-of-the-art results on a new, more challenging formulation of the underlying problem, which has not been addressed before. Additionally, we examine the network’s ability to learn higher-order interactions.

Abstract Recent work in natural language processing (NLP) has yielded appealing results from scaling model parameters and training data; however, using only scale to improve performance means that resource consumption also grows. Such resources include data, time, storage, or energy, all of which are naturally limited and unevenly distributed. This motivates research into efficient methods that require fewer resources to achieve similar results. This survey synthesizes and relates current methods and findings in efficient NLP. We aim to provide both guidance for conducting NLP under limited resources, and point towards promising research directions for developing more efficient methods.

The task of learning to map an input set onto a permuted sequence of its elements is challenging for neural networks. Set-to-sequence problems occur in natural language processing, computer vision and structure prediction, where interactions between elements of large sets define the optimal output. Models must exhibit relational reasoning, handle varying cardinalities and manage combinatorial complexity. Previous attention-based methods require n layers of their set transformations to explicitly represent n-th order relations. Our aim is to enhance their ability to efficiently model higher-order interactions through an additional interdependence component. We propose a novel neural set encoding method called the Set Interdependence Transformer, capable of relating the set's permutation invariant representation to its elements within sets of any cardinality. We combine it with a permutation learning module into a complete, 3-part set-to-sequence model and demonstrate its state-of-the-art performance on a number of tasks. These range from combinatorial optimization problems, through permutation learning challenges on both synthetic and established NLP datasets for sentence ordering, to a novel domain of product catalog structure prediction. Additionally, the network's ability to generalize to unseen sequence lengths is investigated and a comparative empirical analysis of the existing methods' ability to learn higher-order interactions is provided.

Philine Zeinert, Nanna Inie, Leon Derczynski. 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). 2021.

Machine learning on sets towards sequential output is an important and ubiquitous task, with applications ranging from language modelling and meta-learning to multi-agent strategy games and power grid optimization. Combining elements of representation learning and structured prediction, its two primary challenges include obtaining a meaningful, permutation invariant set representation and subsequently utilizing this representation to output a complex target permutation. This paper provides a comprehensive introduction to the _eld as well as an overview of important machine learning methods tackling both of these key challenges, with a detailed qualitative comparison of selected model architectures.

Word clustering groups words that exhibit similar properties. One popular method for this is Brown clustering, which uses short-range distributional information to construct clusters. Specifically, this is a hard hierarchical clustering with a fixed-width beam that employs bi-grams and greedily minimizes global mutual information loss. The result is word clusters that tend to outperform or complement other word representations, especially when constrained by small datasets. However, Brown clustering has high computational complexity and does not lend itself to parallel computation. This, together with the lack of efficient implementations, limits their applicability in NLP. We present efficient implementations of Brown clustering and the alternative Exchange clustering as well as a number of methods to accelerate the computation of both hierarchical and flat clusters. We show empirically that clusters obtained with the accelerated method match the performance of clusters computed using the original methods.

The presence of offensive language on social media platforms and the implications this poses is becoming a major concern in modern society. Given the enormous amount of content created every day, automatic methods are required to detect and deal with this type of content. Until now, most of the research has focused on solving the problem for the English language, while the problem is multilingual. We construct a Danish dataset DKhate containing user-generated comments from various social media platforms, and to our knowledge, the first of its kind, annotated for various types and target of offensive language. We develop four automatic classification systems, each designed to work for both the English and the Danish language. In the detection of offensive language in English, the best performing system achieves a macro averaged F1-score of 0.74, and the best performing system for Danish achieves a macro averaged F1-score of 0.70. In the detection of whether or not an offensive post is targeted, the best performing system for English achieves a macro averaged F1-score of 0.62, while the best performing system for Danish achieves a macro averaged F1-score of 0.73. Finally, in the detection of the target type in a targeted offensive post, the best performing system for English achieves a macro averaged F1-score of 0.56, and the best performing system for Danish achieves a macro averaged F1-score of 0.63. Our work for both the English and the Danish language captures the type and targets of offensive language, and present automatic methods for detecting different kinds of offensive language such as hate speech and cyberbullying.

Information extraction systems and techniques have been largely used to deal with the increasing amount of unstructured data available nowadays. Time is among the different kinds of information that may be extracted from such unstructured data sources, including text documents. However, the inability to correctly identify and extract temporal information from text makes it difficult to understand how the extracted events are organised in a chronological order. Furthermore, in many situations, the meaning of temporal expressions (timexes) is imprecise, such as in “less than 2 years” and “several weeks”, and cannot be accurately normalised, leading to interpretation errors. Although there are some approaches that enable representing imprecise timexes, they are not designed to be applied to specific scenarios and difficult to generalise. This paper presents a novel methodology to analyse and normalise imprecise temporal expressions by representing temporal imprecision in the form of membership functions, based on human interpretation of time in two different languages (Portuguese and English). Each resulting model is a generalisation of probability distributions in the form of trapezoidal and hexagonal fuzzy membership functions. We use an adapted F1-score to guide the choice of the best models for each kind of imprecise timex and a weighted F1-score ( $$\hbox {\textit{F}1}_{3\mathrm{D}}$$ ) as a complementary metric in order to identify relevant differences when comparing two normalisation models. We apply the proposed methodology for three distinct classes of imprecise timexes, and the resulting models give distinct insights in the way each kind of temporal expression is interpreted.

Named Entity Recognition (NER) is a key NLP task, which is all the more challenging on Web and user-generated content with their diverse and continuously changing language. This paper aims to quantify how this diversity impacts state-of-the-art NER methods, by measuring named entity (NE) and context variability, feature sparsity, and their effects on precision and recall. In particular, our findings indicate that NER approaches struggle to generalise in diverse genres with limited training data. Unseen NEs, in particular, play an important role, which have a higher incidence in diverse genres such as social media than in more regular genres such as newswire. Coupled with a higher incidence of unseen features more generally and the lack of large training corpora, this leads to significantly lower F 1 scores for diverse genres as compared to more regular ones. We also find that leading systems rely heavily on surface forms found in training data, having problems generalising beyond these, and offer explanations for this observation.

Brown clustering is an established technique, used in hundreds of computational linguistics papers each year, to group word types that have similar distributional information. It is unsupervised and can be used to create powerful word representations for machine learning. Despite its improbable success relative to more complex methods, few have investigated whether Brown clustering has really been applied optimally. In this paper, we present a subtle but profound generalisation of Brown clustering to improve the overall quality by decoupling the number of output classes from the computational active set size. Moreover, the generalisation permits a novel approach to feature selection from Brown clusters: We show that the standard approach of shearing the Brown clustering output tree at arbitrary bitlengths is lossy and that features should be chosen insead by rolling up Generalised Brown hierarchies. The generalisation and corresponding feature generation is more principled, challenging the way Brown clustering is currently understood and applied.

Determining the relative order of events and times described in text is an important problem in natural language processing. It is also a difficult one: general state-of-the-art performance has been stuck at a relatively low ceiling for years. We investigate the representation of temporal relations, and empirically evaluate the effect that various temporal relation representations have on machine learning performance. While machine learning performance decreases with increased representational expressiveness, not all representation simplifications have equal impact.

One of the main obstacles, hampering method development and comparative evaluation of named entity recognition in social media, is the lack of a sizeable, diverse, high quality annotated corpus, analogous to the CoNLL’2003 news dataset. For instance, the biggest Ritter tweet corpus is only 45,000 tokens – a mere 15% the size of CoNLL’2003. Another major shortcoming is the lack of temporal, geographic, and author diversity. This paper introduces the Broad Twitter Corpus (BTC), which is not only significantly bigger, but sampled across different regions, temporal periods, and types of Twitter users. The gold-standard named entity annotations are made by a combination of NLP experts and crowd workers, which enables us to harness crowd recall while maintaining high quality. We also measure the entity drift observed in our dataset (i.e. how entity representation varies over time), and compare to newswire. The corpus is released openly, including source text and intermediate annotations.

Abstract: Intuitive and effective access to large volumes of information is increasingly important. As social media explodes as a useful source of information, so are methods required to access these large volumes of user-generated content. Word clouds are an effective information access tool. However, those generated over social media data often depict redundant and mis-ranked entries. This limits the users ’ ability to browse and explore datasets. This paper proposes a method for improving word cloud generation over social streams. Named entity expressions in tweets are detected, disambiguated and aggregated into entity clusters. A word cloud is generated from terms that represent the most relevant entity clusters. We find that word clouds with grouped named entities attain significantly broader coverage and significantly decreased content duplication. Further, access to relevant entries in the collection is improved. An extrinsic crowdsourced user evaluation of generated word clouds was performed. Word clouds with grouped named entities are rated as significantly more relevant and more diverse with respect to the baseline. In addition, we found that word clouds with higher levels of Mean Average Precision (MAP) are more likely to be rated by users as being relevant to the concepts reflected. Critically, this supports MAP as a tool for predicting word cloud quality without requiring a human in the loop. 1

Kalina Bontcheva, Ian Roberts, Leon Derczynski, Samantha Alexander-Eames. Proceedings of the Demonstrations at the 14th Conference of the European Chapter of the Association for Computational Linguistics. 2014.

Recognising entities in social media text is difficult.NER on newswire text is conventionally cast as a sequence labeling problem.This makes implicit assumptions regarding its textual structure.Social media text is rich in disfluency and often has poor or noisy structure, and intuitively does not always satisfy these assumptions.We explore noise-tolerant methods for sequence labeling and apply discriminative post-editing to exceed state-of-the-art performance for person recognition in tweets, reaching an F1 of 84%.

Danish is a major Scandinavian language spoken daily by around six million people.However, it lacks a unified, open set of NLP tools.This demonstration will introduce DKIE, an extensible open-source toolkit for processing Danish text.We implement an information extraction architecture for Danish within GATE, including integrated third-party tools.This implementation includes the creation of a substantial set of corpus annotations for dataintensive named entity recognition.The final application and dataset is made are openly available, and the part-of-speech tagger and NER model also operate independently or with the Stanford NLP toolkit.

Social media has changed the way we communicate. Social media data capture our social interactions and utterances in machine readable format. Searching and analysing massive and frequently updated social media data brings significant and diverse rewards across many different application domains, from politics and business to social science and epidemiology.