Malta

Sentence-level attacks craft adversarial sentences that are synonymous with correctly-classified sentences but are misclassified by the text classifiers. Under the black-box setting, classifiers are only accessible through their feedback to queried inputs, which is predominately available in the form of class probabilities. Even though utilizing class probabilities results in stronger attacks, due to the challenges of using them for sentence-level attacks, existing attacks use either no feedback or only the class labels. Overcoming the challenges, we develop a novel algorithm that uses class probabilities for black-box sentence-level attacks, investigate the effectiveness of using class probabilities on the attack&#39;s success, and examine the question if it is worthy or practical to use class probabilities by black-box sentence-level attacks. We conduct extensive evaluations of the proposed attack comparing with the baselines across various classifiers and benchmark datasets.

EACL 2024

Exploiting Class Probabilities for Black-box Sentence-level Attacks

score-based attacks

black-box attacks

sentence-level attacks

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

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

Sentence-level attacks craft adversarial sentences that are synonymous with correctly-classified sentences but are misclassified by the text classifiers. Under the black-box setting, classifiers are only accessible through their feedback to queried inputs, which is predominately available in the form of class probabilities. Even though utilizing class probabilities results in stronger attacks, due to the challenges of using them for sentence-level attacks, existing attacks use either no feedback or only the class labels. Overcoming the challenges, we develop a novel algorithm that uses class probabilities for black-box sentence-level attacks, investigate the effectiveness of using class probabilities on the attack's success, and examine the question if it is worthy or practical to use class probabilities by black-box sentence-level attacks. We conduct extensive evaluations of the proposed attack comparing with the baselines across various classifiers and benchmark datasets.

poster

Welcome to the 18th Conference of the European Chapter of the Association for Computational Linguistics (EACL). Continuing its mission of expanding and involving the science community of all European countries, EACL has selected the Malta community for the 18th EACL. Considering the importance of physical interaction among researchers, the conference will be held at the Hotel Radisson Blu, St. Julians, in Malta, from 17 to 22 of March, 2024. The conference will also feature the possibility to participate virtually. As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.<br><br>
#### [Conference Handbook](https://drive.google.com/file/d/13Wn38q0yev6U3RTgnf4Tn5T9GuAuZ-xW/view?usp=sharing) *downloadable*<br><br><br><br>

To access the EACL 2024 event page you need to register [**here**](https://acl.swoogo.com/EACL2024). 

Please register!

As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

Supervised contrastive learning (SCL) frameworks treat each class as independent and thus consider all classes to be equally important. This neglects the common scenario in which label hierarchy exists, where fine-grained classes under the same category show more similarity than very different ones. This paper introduces a family of Label-Aware SCL methods (LA-SCL) that incorporates hierarchical information to SCL by leveraging similarities between classes, resulting in creating a more well-structured and discriminative feature space. This is achieved by first adjusting the distance between instances based on measures of the proximity of their classes with the scaled instance-instance-wise contrastive. An additional instance-center-wise contrastive is introduced to move within-class examples closer to their centers, which are represented by a set of learnable label parameters. The learned label parameters can be directly used as a nearest neighbor classifier without further finetuning. In this way, a better feature representation is generated with improvements of intra-cluster compactness and inter-cluster separation. Experiments on three datasets show that the proposed LA-SCL works well on text classification of distinguishing a single label among multi-labels, outperforming the baseline supervised approaches. Our code is publicly available $^1$.

Learning Label Hierarchy with Supervised Contrastive Learning

Internet memes have gained immense traction as a medium for individuals to convey emotions, thoughts, and perspectives on social media. While memes often serve as sources of humor and entertainment, they can also propagate offensive, incendiary, or harmful content, deliberately targeting specific individuals or communities. Identifying such memes is challenging because of their satirical and cryptic characteristics. Most contemporary research on memes’ detrimental facets is skewed towards high-resource languages, often sidelining the unique challenges tied to low-resource languages, such as Bengali. To facilitate this research in low-resource languages, this paper presents a novel dataset MIMOSA (MultIMOdal aggreSsion dAtaset) in Bengali. MIMOSA encompasses 4,848 annotated memes across five aggression target categories: Political, Gender, Religious, Others, and non-aggressive. We also propose MAF (Multimodal Attentive Fusion), a simple yet effective approach that uses multimodal context to detect the aggression targets. MAF captures the selective modality-specific features of the input meme and jointly evaluates them with individual modality features. Experiments on MIMOSA exhibit that the proposed method outperforms several state-of-the-art rivaling approaches. Our code and data are available at https://github.com/shawlyahsan/Bengali-Aggression-Memes.

A Multimodal Framework to Detect Target Aware Aggression in Memes

Cyber attacks cause over $1 trillion loss every year. An important task for cyber security analysts is attack forensics. It entails understanding malware behaviors and attack origins. However, existing automated or manual malware analysis can only disclose a subset of behaviors due to inherent difficulties (e.g., malware cloaking and obfuscation). As such, analysts often resort to text search techniques to identify existing malware reports based on the symptoms they observe, exploiting the fact that malware samples share a lot of similarity, especially those from the same origin. In this paper, we propose a novel malware behavior search technique that is based on graph isomorphism at the attention layers of Transformer models. We also compose a large dataset collected from various agencies to facilitate such research.
Our technique outperforms state-of-the-art methods, such as those based on sentence embeddings and keywords by 6-14%. In the case study of 10 real-world malwares, our technique can correctly attribute 8 of them to their ground truth origins while using Google only works for 3 cases.

Threat Behavior Textual Search by Attention Graph Isomorphism

State-of-the-art natural language generation evaluation metrics are based on black-box language models. Hence, recent works consider their explainability with the goals of better understandability for humans and better metric analysis, including failure cases. In contrast, we explicitly leverage explanations to boost the metrics' performance. In particular, we perceive feature importance explanations as word-level scores, which we convert, via power means, into a segment-level score. We then combine this segment-level score with the original metric to obtain a better metric. Our tests show improvements for multiple metrics across MT and summarization datasets. While improvements on machine translation are small, they are strong for summarization. Notably, BMX with the LIME explainer and preselected parameters achieves an average improvement of 0.087 points in Spearman correlation on the system-level evaluation of SummEval.

BMX: Boosting Natural Language Generation Metrics with Explainability

This presentation encompasses the overall content and notable aspects of our survey paper. It begins with the motivation behind writing a survey paper on the topic of 'Large Language Models for Mathematical Reasoning,' introducing how Math Datasets are organized within the paper. It covers a brief introduction to the topic of 'LLMs for Math,' including researchers' efforts to improve model performance, the models used, and various strategies. Finally, the presentation concludes by summarizing the current limitations in this field.

Large Language Models for Mathematical Reasoning: Progresses and Challenges

This paper presents multi-reference benchmarks for the Grammatical Error Correction (GEC) of Russian, based on two existing single-reference datasets, for a total of 7,444 learner sentences from a variety of first language backgrounds. Each sentence is corrected independently by two new raters, and their corrections are reviewed by a senior annotator, resulting in a total of three references per sentence. Analysis of the annotations reveals that the new raters tend to make more changes, compared to the original raters, especially at the lexical level. We conduct experiments with two popular GEC approaches and show competitive performance on the original datasets and the new benchmarks. We also compare system scores as evaluated against individual annotators and discuss the effect of using multiple references overall and on specific error types. We find that using the union of the references increases system scores by more than 10 points and decreases the gap between system and human performance, thereby providing a more realistic evaluation of GEC system performance, although the effect is not the same across the error types. The annotations are available for research.

Multi-Reference Benchmarks for Russian Grammatical Error Correction

Depression is a global concern suffered by millions of people, significantly impacting their thoughts and behavior. Over the years, heightened awareness, spurred by health campaigns and other initiatives, has driven the study of this disorder using data collected from social media platforms. In our research, we aim to gauge the severity of symptoms related to depression among social media users. The ultimate goal is to estimate the user's responses to a well-known standardized psychological questionnaire, the Beck Depression Inventory-II (BDI). This is a 21-question multiple-choice self-report inventory that covers multiple topics about how the subject has been feeling. Mining users' social media interactions and understanding psychological states represents a challenging goal. To that end, we present here an approach based on search and summarization that extracts multiple BDI-biased summaries from the thread of users' publications. We also leverage a robust large language model to estimate the potential answer for each BDI item. Our method involves several steps. First, we employ a search strategy based on sentence similarity to obtain pertinent extracts related to each topic in the BDI questionnaire. Next, we compile summaries of the content of these groups of extracts. Last, we exploit chatGPT to respond to the 21 BDI questions, using the summaries as contextual information in the prompt. Our model has undergone rigorous evaluation across various depression datasets, yielding encouraging results. The experimental report includes a comparison against an assessment done by expert humans and competes favorably with state-of-the-art methods.

Delving into the Depths: Evaluating Depression Severity through BDI-biased Summaries

Within Motivational Interviewing (MI), client utterances are coded as for or against a certain behaviour change, along with commitment strength; this is essential to ensure therapists soften rather than persisting goal-related actions in the face of resistance. Prior works in MI agents have been scripted or semi-scripted, limiting users' natural language expressions. With the aim of automating the MI interactions, we propose and explore the task of automated identification of client motivational language. Employing Large Language Models (LLMs), we compare in-context learning (ICL) and instruction fine-tuning (IFT) with varying training sizes for this identification task. Our experiments show that both approaches can learn under low-resourced settings. Our results demonstrate that IFT, though cheaper, is more stable to prompt choice, and yields better performance with more data. Given the detected motivation, we further present an approach to the analysis of therapists' strategies for balancing building rapport with clients with advancing the treatment plan. A framework of MI agents is developed using insights from the data and the psychotherapy literature.

How Can Client Motivational Language Inform Psychotherapy Agents?

Therapist Self-Disclosure (TSD) within the context of psychotherapy entails the revelation of personal information by the therapist. The ongoing scholarly discourse surrounding the utility of TSD, spanning from the inception of psychotherapy to the present day, has underscored the need for greater specificity in conceptualizing TSD. This inquiry has yielded more refined classifications within the TSD domain, with a consensus emerging on the distinction between immediate and non-immediate TSD, each of which plays a distinct role in the therapeutic process. Despite this progress in the field of psychotherapy, the Natural Language Processing (NLP) domain currently lacks methodological solutions or explorations for such scenarios. This lacuna can be partly due to the difficulty of attaining publicly available clinical data. To address this gap, this paper presents an innovative NLP-based approach that formalizes TSD as an NLP task. The proposed methodology involves the creation of publicly available, expert-annotated test sets designed to simulate therapist utterances, and the employment of NLP techniques for evaluation purposes. By integrating insights from psychotherapy research with NLP methodologies, this study aims to catalyze advancements in both NLP and psychotherapy research.

Therapist Self-Disclosure as a Natural Language Processing Task

Objective: A thematic and topic modelling analysis of sleep concerns in a social media derived, privacy-preserving, suicidality dataset. This forms the basis for an exploration of sleep as a potential computational linguistic signal in suicide prevention. Background: Suicidal ideation is a limited signal for suicide. Developments in computational linguistics and mental health datasets afford an opportunity to investigate additional signals and to consider the broader clinical ethical design implications. Methodology: A clinician-led integration of reflexive thematic analysis, with machine learning topic modelling (Bertopic), and the purposeful sampling of the University of Maryland Suicidality Dataset. Results: Sleep as a place of refuge and escape, revitalisation for exhaustion, and risk and vulnerability were generated as core themes in an initial thematic analysis of 546 posts. Bertopic analysing 21,876 sleep references in 16791 posts facilitated the production of 40 topics that were clinically interpretable, relevant, and thematically aligned to a level that exceeded original expectations. Privacy and synthetic representative data, reproducibility, validity and stochastic variability of results, and a multi-signal formulation perspective, are highlighted as key research and clinical issues.

Downloads

Next from EACL 2024

Learning Label Hierarchy with Supervised Contrastive Learning

Similar lecture

A Strong Baseline for Query Efficient Attacks in a Black Box Setting