The debate around vaccines has been going on for decades, but the COVID-19 pandemic showed how crucial it is to understand and mitigate anti-vaccine sentiments. While the pandemic may be over, it is still important to understand how the pandemic affected the anti-vaccine discourse, and whether the arguments against non-COVID vaccines have also changed due to the pandemic. This study attempts to answer these questions through a large-scale study of anti-vaccine posts on Twitter. Almost all prior works that utilized social media to understand anti-vaccine opinions considered only the three broad stances of Anti-Vax, Pro-Vax, and Neutral. There has not been any effort to identify the specific reasons/concerns behind the anti-vax sentiments (e.g., side-effects, conspiracy theories, political reasons) on social media at scale. In this work, we propose two novel methods for classifing tweets into 11 different anti-vax concerns -- a discriminative approach (entailment-based) and a generative approach (based on instruction tuning of LLMs) -- which outperform several strong baselines. We then apply this classifier on anti-vaccine tweets posted over a 5-year period (Jan 2018 - Jan 2023) to understand how the COVID-19 pandemic has impacted the anti-vaccine concerns among the masses. We find that the pandemic has made the anti-vaccine discourse far more complex than in the pre-COVID times, and increased the variety of concerns being voiced. Alarmingly, we find that concerns about COVID vaccines are now being projected onto the non-COVID vaccines, thus making more people hesitant in taking them in post-COVID times.

Subhendu Khatuya, Rajdeep Mukherjee, Akash Ghosh, Manjunath Hegde, Koustuv Dasgupta, Niloy Ganguly, Saptarshi Ghosh, Pawan Goyal. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Products on e-commerce websites are usually organized based on seller-provided product attributes. Customers looking for a product typically have certain needs or product use-cases in mind, for e.g., a headphone for gym classes, or a printer for a small business. However, they often struggle to map these use-cases to product attributes and subsequently fail to find the product they need. In this talk, we present a use-case based shopping (UBS) ML system that facilitates use-case based customer experiences (CXs). The UBS system recommends dominant product use-cases to customers along with most relevant products for those use-cases. Use-cases and their definitions vary across product categories and marketplaces (MPs). This makes training supervised models for thousands of e-commerce categories and multiple MPs infeasible by collecting large amount training data needed to train these models. In this talk, we present our work on scaling the UBS model by instruction tuning an LLM for our task.

While social media platforms play an important role in our daily lives in obtaining the latest news and trends from across the globe, they are known to be prone to widespread proliferation of harmful information in different forms leading to misconceptions among the masses. Accordingly, several prior works have attempted to tag social media posts with labels/classes reflecting their veracity, sentiments, hate content, and so on. However, in order to have a convincing impact, it is important to additionally extract the post snippets on which the labelling decision is based. We call such a post snippet the rationale . These rationales significantly improve human trust and debuggability of the predictions, especially when detecting misinformation or stigmas from social media posts. These rationale spans or snippets are also helpful in post-classification social analysis, such as for finding out the target communities in hate-speech, or for understanding the arguments or concerns against the intake of vaccines. Also it is observed that a post may express multiple notions of misinformation, hate, sentiment, and the like. Thus, the task of determining (one or multiple) labels for a given piece of text, along with the text snippets explaining the rationale behind each of the identified labels is a challenging multi-label, multi-rationale classification task, which is still nascent in the literature. While transformer -based encoder-decoder generative models such as BART and T5 are well suited for the task, in this work we show how a relatively simpler encoder-only discriminative question-answering (QA) model can be effectively trained using simple template-based questions to accomplish the task. We thus propose MuLX-QA and demonstrate its utility in producing (label, rationale span) pairs in two different settings: multi-class (on the HateXplain dataset related to hate speech on social media), and multi-label (on the CAVES dataset related to COVID-19 anti-vaccine concerns). MuLX-QA outperforms heavier generative models in both settings. We also demonstrate the relative advantage of our proposed model MuLX-QA over strong baselines when trained with limited data. We perform several ablation studies, and experiments to better understand the effect of training MuLX-QA with different question prompts, and draw interesting inferences. Additionally, we show that MuLX-QA is effective on social media posts in resource-poor non-English languages as well. Finally, we perform a qualitative analysis of our model predictions and compare them with those of our strongest baseline.

Automatic summarization of legal case judgments is a practically important problem that has attracted substantial research efforts in many countries.In the context of the Indian judiciary, there is an additional complexity -Indian legal case judgments are mostly written in complex English, but a significant portion of India's population lacks command of the English language.Hence, it is crucial to summarize the legal documents in Indian languages to ensure equitable access to justice.While prior research primarily focuses on summarizing legal case judgments in their source languages, this study presents a pioneering effort toward cross-lingual summarization of English legal documents into Hindi, the most frequently spoken Indian language.We construct the first high-quality legal corpus comprising of 3,122 case judgments from prominent Indian courts in English, along with their summaries in both English and Hindi, drafted by legal practitioners.We benchmark the performance of several diverse summarization approaches on our corpus and demonstrate the need for further research in cross-lingual summarization in the legal domain.

In the dynamic realm of AI, integrating Large Language Models (LLMs) with security systems reshape cybersecurity. LLMs bolster defense against cyber threats but also introduce risks, aiding adversaries in generating malicious content, discovering vulnerabilities, and distorting perceptions. This paper presents Net-GPT, an LLM-empowered offensive chatbot that understands network protocols and launches Unmanned Aerial Vehicles (UAV)-based Man-in-the-middle (MITM) attacks against a hijack communication between UAV and Ground Control Stations (GCS). Facilitated by an edge server equipped with finely tuned LLMs, Net-GPT crafts mimicked network packets between UAV and GCS. Leveraging the adaptability of popular LLMs, Net-GPT produces context-aligned network packets. We fine-tune and assess Net-GPT's LLM-based efficacy, showing its impressive generative accuracy: 95.3% for Llama-2-13B and 94.1% for Llama-2-7B. Smaller LLMs, such as Distil-GPT-2, reach 77.9% predictive capability of Llama-2-7B but are 47× faster. Cost-efficiency tests highlight model quality's impact on accuracy while fine-tuning data quantity enhances predictability on specific metrics. It holds great potential to be used in edge-computing environments with amplified computing capability.

Amazon Web Services (AWS) is a comprehensive and broadly adopted cloud provider. AWS SDKs provide access to AWS services through API endpoints. However, incorrect use of these APIs can lead to code defects, crashes, performance issues, and other problems. AWS best practices are a set of guidelines for correct and secure use of these APIs to access cloud services, allowing conformant clients to fully reap the benefits of cloud computing. We present static analyses, developed in the context of a commercial service for detection of code defects and security vulnerabilities, to identify deviations from AWS best practices. We focus on applications that use the AWS SDK for Python, called Boto3. Precise static analysis of Python cloud applications requires robust type inference for inferring the types of cloud service clients. However, Boto3’s "Pythonic" APIs pose unique challenges for type resolution, as does the interprocedural style in which service clients are used. We offer a layered approach that combines multiple type-resolution and tracking strategies in a staged manner: (i) general-purpose type inference augmented by type annotations, (ii) interprocedural dataflow analysis expressed in a domain-specific language, and (iii) name-based resolution as a low-confidence fallback. Across >3,000 popular Python GitHub repos that make use of the AWS SDK, our layered type inference system achieves 85% precision and 100% recall in inferring Boto3 clients in Python client code. Additionally, we use real-world developer feedback to assess a representative sample of eight AWS best-practice rules. These rules detect a wide range of issues including pagination, polling, and batch operations. Developers have accepted more than 85% of the recommendations made by five out of eight Python rules, and almost 83% of all recommendations.

Rajdeep Mukherjee, Abhinav Bohra, Akash Banerjee, Soumya Sharma, Manjunath Hegde, Afreen Shaikh, Shivani Shrivastava, Koustuv Dasgupta, Niloy Ganguly, Saptarshi Ghosh, Pawan Goyal. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

Convincing people to get vaccinated against COVID-19 is a key societal challenge in the present times. As a first step towards this goal, many prior works have relied on social media analysis to understand the specific concerns that people have towards these vaccines, such as potential side-effects, ineffectiveness, political factors, and so on. Though there are datasets that broadly classify social media posts into Anti-vax and Pro-Vax labels, there is no dataset (to our knowledge) that labels social media posts according to the specific anti-vaccine concerns mentioned in the posts. In this paper, we have curated CAVES, the first large-scale dataset containing about 10k COVID-19 anti-vaccine tweets labelled into various specific anti-vaccine concerns in a multi-label setting. This is also the first multi-label classification dataset that provides explanations for each of the labels. Additionally, the dataset also provides class-wise summaries of all the tweets. We also perform preliminary experiments on the dataset and show that this is a very challenging dataset for multi-label explainable classification and tweet summarization, as is evident by the moderate scores achieved by some state-of-the-art models.

Occurrences of catastrophes such as natural or man-made disasters trigger the spread of rumours over social media at a rapid pace. Presenting a trustworthy and summarized account of the unfolding event in near real-time to the consumers of such potentially unreliable information thus becomes an important task. In this work, we propose MTLTS, the first end-to-end solution for the task that jointly determines the credibility and summary-worthiness of tweets. Our credibility verifier is designed to recursively learn the structural properties of a Twitter conversation cascade, along with the stances of replies towards the source tweet. We then take a hierarchical multi-task learning approach, where the verifier is trained at a lower layer, and the summarizer is trained at a deeper layer where it utilizes the verifier predictions to determine the salience of a tweet. Different from existing disaster-specific summarizers, we model tweet summarization as a supervised task. Such an approach can automatically learn summary-worthy features, and can therefore generalize well across domains. When trained on the PHEME dataset [29], not only do we outperform the strongest baselines for the auxiliary task of verification/rumour detection, we also achieve 21 - 35% gains in the verified ratio of summary tweets, and 16 - 20% gains in ROUGE1-F1 scores over the existing state-of-the-art solutions for the primary task of trustworthy summarization.

Aspect Sentiment Triplet Extraction (ASTE) deals with extracting opinion triplets, consisting of an opinion target or aspect, its associated sentiment, and the corresponding opinion term/span explaining the rationale behind the sentiment.Existing research efforts are majorly tagging-based.Among the methods taking a sequence tagging approach, some fail to capture the strong interdependence between the three opinion factors, whereas others fall short of identifying triplets with overlapping aspect/opinion spans.A recent grid tagging approach on the other hand fails to capture the span-level semantics while predicting the sentiment between an aspect-opinion pair.Different from these, we present a tagging-free solution for the task, while addressing the limitations of the existing works.We adapt an encoder-decoder architecture with a Pointer Network-based decoding framework that generates an entire opinion triplet at each time step thereby making our solution end-to-end.Interactions between the aspects and opinions are effectively captured by the decoder by considering their entire detected spans while predicting their connecting sentiment.Extensive experiments on several benchmark datasets establish the better efficacy of our proposed approach, especially in recall, and in predicting multiple and aspect/opinion-overlapped triplets from the same review sentence.We report our results both with and without BERT and also demonstrate the utility of domainspecific BERT post-training for the task.

We propose VADEC, a multi-task framework that exploits the correlation between the categorical and dimensional models of emotion representation for better subjectivity analysis. Focusing primarily on the effective detection of emotions from tweets, we jointly train multi-label emotion classification and multi-dimensional emotion regression, thereby utilizing the inter-relatedness between the tasks. Co-training especially helps in improving the performance of the classification task as we outperform the strongest baselines with 3.4%, 11%, and 3.9% gains in Jaccard Accuracy, Macro-F1, and Micro-F1 scores respectively on the AIT dataset [17]. We also achieve state-of-the-art results with 11.3% gains averaged over six different metrics on the SenWave dataset [27]. For the regression task, VADEC, when trained with SenWave, achieves 7.6% and 16.5% gains in Pearson Correlation scores over the current state-of-the-art on the EMOBANK dataset [5] for the Valence (V) and Dominance (D) affect dimensions respectively. We conclude our work with a case study on COVID-19 tweets posted by Indians that further helps in establishing the efficacy of our proposed solution.

Manually extracting relevant aspects and opinions from large volumes of user-generated text is a time-consuming process. Summaries, on the other hand, help readers with limited time budgets to quickly consume the key ideas from the data. State-of-the-art approaches for multi-document summarization, however, do not consider user preferences while generating summaries. In this work, we argue the need and propose a solution for generating personalized aspect-based opinion summaries from large collections of online tourist reviews. We let our readers decide and control several attributes of the summary such as the length and specific aspects of interest among others. Specifically, we take an unsupervised approach to extract coherent aspects from tourist reviews posted onTripAdvisor. We then propose an Integer Linear Programming (ILP) based extractive technique to select an informative subset of opinions around the identified aspects while respecting the user-specified values for various control parameters. Finally, we evaluate and compare our summaries using crowdsourcing and ROUGE-based metrics and obtain competitive results.

We consider the problem of efficiently checking a set of safety properties P1,...,Pkof one design. We introduce a new approach called JA-verification, where JA stands for “JustAssume” (as opposed to “assume-guarantee”). In this approach, when proving a property Pi, one assumes that every property Pjfor j ≠ i holds. The process of proving properties either results in showing that P1,...,Pkhold without any assumptions or finding a “debugging set” of properties. The latter identifies a subset of failed properties that are the first to break. The design behaviors that cause the properties in the debugging set to fail must be fixed first. Importantly, in our approach, there is no need to prove the assumptions used. We describe the theory behind our approach and report experimental results that demonstrate substantial gains in performance, especially in the cases where a small debugging set exists.

Verification is indispensable for building reliable of hardware/software co-designs. However, the scope of formal methods in this domain is limited. This is attributed to the lack of unified property specification languages, the semantic gap between hardware and software components, and the lack of verifiers that support both C and Verilog/VHDL. To address these limitations, we present an approach that uses a bounded co-verification tool, HW-CBMC, for formally validating hardware/software co-designs written in Verilog and C. Properties are expressed in C enriched with special-purpose primitives that capture temporal correlation between hardware and software events. We present an industrial case-study, proving bounded safety properties as well as discovering critical co-design bugs on a large and complex text analytics FPGA accelerator from IBM®.

This paper reports for the first time, the use of a hardware-software combined bounded model checking approach for hardware-software mixed implementations of power management logic. We report significant performance gains as compared to our earlier attempt of extracting a finite quotient transition system from the control software.

With the growing complexity and gradually shrinking power requirements in the system-on-chip designs, sophisticated global power management policies (which orchestrate the switching between power states of multiple power domains) are commonplace. Recent research has paved some novel ways to verify the sophisticated on-chip architectural power management decisions and analyze the verification coverage. However, one of the primary challenges in verifying such power management architectures stems from the mixed implementation of such strategies, where the local power controllers are in hardware and the global power management is implemented in software/firmware. There has been lack of effort to build a unified and automated framework for power intent verification and coverage analysis for generic power management logics. This paper tries to develop an end-to-end automated framework enabled by a tool named POWER-TRUCTOR for power intent validation.