Search

With social media usage growing exponentially in the past few years, fake news has also become extremely prevalent. The detrimental impact of fake news emphasizes the need for research focused on automating the detection of false information and verifying its accuracy. In this work, we present the outcome of the Factify 2 shared task, which provides a multi-modal fact verification and satire news dataset, as part of the DeFactify 2 workshop at AAAI'23. The data calls for a comparison based approach to the task by pairing social media claims with supporting documents, with both text and image, divided into 5 classes based on multi-modal relations. In the second iteration of this task we had over 60 participants and 9 final test-set submissions. The best performances came from the use of DeBERTa for text and Swinv2 and CLIP for image. The highest F1 score averaged for all five classes was 81.82%., Defactify2 @AAAI 2023

Word embeddings, i.e., semantically meaningful vector representation of words, are largely influenced by the distributional hypothesis "You shall know a word by the company it keeps" (Harris, 1954), whereas modern prediction-based neural network embeddings rely on design choices and hyperparameter optimization. Word embeddings like Word2Vec, GloVe etc. well capture the contextuality and real-world analogies but contemporary convolution-based image embeddings such as VGGNet, AlexNet, etc. do not capture contextual knowledge. The popular king-queen analogy does not hold true for most commonly used vision embeddings. In this paper, we introduce a pre-trained joint embedding (JE), named IMAGINATOR, trained on 21K distinct image objects level from 1M image+text pairs. JE is a way to encode multimodal data into a vector space where the text modality serves as the ground-ing key, which the complementary modality (in this case, the image) is anchored with. IMAGINATOR encapsulates three individual representations: (i) object-object co-location, (ii) word-object co-location, and (iii) word-object correlation. These three ways capture complementary aspects of the two modalities which are further combined to obtain the final JEs. Generated JEs are intrinsically evaluated to assess how well they capture the contextuality and real-world analogies. We also evaluate pre-trained IMAGINATOR JEs on three downstream tasks: (i) image captioning, (ii) Image2Tweet, and (iii) text-based image retrieval. IMAGINATOR establishes a new standard on the aforementioned down-stream tasks by outperforming the current SoTA on all the selected tasks. IMAGINATOR will be made publicly available. The codes are available at https://github.com/varunakk/IMAGINATOR

The internet gives the world an open platform to express their views and share their stories. While this is very valuable, it makes fake news one of our society's most pressing problems. Manual fact checking process is time consuming, which makes it challenging to disprove misleading assertions before they cause significant harm. This is he driving interest in automatic fact or claim verification. Some of the existing datasets aim to support development of automating fact-checking techniques, however, most of them are text based. Multi-modal fact verification has received relatively scant attention. In this paper, we provide a multi-modal fact-checking dataset called FACTIFY 2, improving Factify 1 by using new data sources and adding satire articles. Factify 2 has 50,000 new data instances. Similar to FACTIFY 1.0, we have three broad categories - support, no-evidence, and refute, with sub-categories based on the entailment of visual and textual data. We also provide a BERT and Vison Transformer based baseline, which acheives 65% F1 score in the test set. The baseline codes and the dataset will be made available at https://github.com/surya1701/Factify-2.0., Defactify@AAAI2023

Memes are the new-age conveyance mechanism for humor on social media sites. Memes often include an image and some text. Memes can be used to promote disinformation or hatred, thus it is crucial to investigate in details. We introduce Memotion 3, a new dataset with 10,000 annotated memes. Unlike other prevalent datasets in the domain, including prior iterations of Memotion, Memotion 3 introduces Hindi-English Codemixed memes while prior works in the area were limited to only the English memes. We describe the Memotion task, the data collection and the dataset creation methodologies. We also provide a baseline for the task. The baseline code and dataset will be made available at https://github.com/Shreyashm16/Memotion-3.0, Defactify2 @AAAI

Over the past decade, analogies, in the form of word-level analogies, have played a significant role as an intrinsic measure of evaluating the quality of word embedding methods such as word2vec. Modern large language models (LLMs), however, are primarily evaluated on extrinsic measures based on benchmarks such as GLUE and SuperGLUE, and there are only a few investigations on whether LLMs can draw analogies between long texts. In this paper, we present ANALOGICAL, a new benchmark to intrinsically evaluate LLMs across a taxonomy of analogies of long text with six levels of complexity -- (i) word, (ii) word vs. sentence, (iii) syntactic, (iv) negation, (v) entailment, and (vi) metaphor. Using thirteen datasets and three different distance measures, we evaluate the abilities of eight LLMs in identifying analogical pairs in the semantic vector space. Our evaluation finds that it is increasingly challenging for LLMs to identify analogies when going up the analogy taxonomy., Comment: Accepted as a long paper at Findings of ACL 2023

Combating disinformation is one of the burning societal crises -- about 67% of the American population believes that disinformation produces a lot of uncertainty, and 10% of them knowingly propagate disinformation. Evidence shows that disinformation can manipulate democratic processes and public opinion, causing disruption in the share market, panic and anxiety in society, and even death during crises. Therefore, disinformation should be identified promptly and, if possible, mitigated. With approximately 3.2 billion images and 720,000 hours of video shared online daily on social media platforms, scalable detection of multimodal disinformation requires efficient fact verification. Despite progress in automatic text-based fact verification (e.g., FEVER, LIAR), the research community lacks substantial effort in multimodal fact verification. To address this gap, we introduce FACTIFY 3M, a dataset of 3 million samples that pushes the boundaries of the domain of fact verification via a multimodal fake news dataset, in addition to offering explainability through the concept of 5W question-answering. Salient features of the dataset include: (i) textual claims, (ii) ChatGPT-generated paraphrased claims, (iii) associated images, (iv) stable diffusion-generated additional images (i.e., visual paraphrases), (v) pixel-level image heatmap to foster image-text explainability of the claim, (vi) 5W QA pairs, and (vii) adversarial fake news stories., Comment: arXiv admin note: text overlap with arXiv:2305.04329

Automatic fact verification has received significant attention recently. Contemporary automatic fact-checking systems focus on estimating truthfulness using numerical scores which are not human-interpretable. A human fact-checker generally follows several logical steps to verify a verisimilitude claim and conclude whether its truthful or a mere masquerade. Popular fact-checking websites follow a common structure for fact categorization such as half true, half false, false, pants on fire, etc. Therefore, it is necessary to have an aspect-based (delineating which part(s) are true and which are false) explainable system that can assist human fact-checkers in asking relevant questions related to a fact, which can then be validated separately to reach a final verdict. In this paper, we propose a 5W framework (who, what, when, where, and why) for question-answer-based fact explainability. To that end, we present a semi-automatically generated dataset called FACTIFY-5WQA, which consists of 391, 041 facts along with relevant 5W QAs - underscoring our major contribution to this paper. A semantic role labeling system has been utilized to locate 5Ws, which generates QA pairs for claims using a masked language model. Finally, we report a baseline QA system to automatically locate those answers from evidence documents, which can serve as a baseline for future research in the field. Lastly, we propose a robust fact verification system that takes paraphrased claims and automatically validates them. The dataset and the baseline model are available at https: //github.com/ankuranii/acl-5W-QA, Comment: Accepted at ACL main conference 2023

Joint embedding (JE) is a way to encode multi-modal data into a vector space where text remains as the grounding key and other modalities like image are to be anchored with such keys. Meme is typically an image with embedded text onto it. Although, memes are commonly used for fun, they could also be used to spread hate and fake information. That along with its growing ubiquity over several social platforms has caused automatic analysis of memes to become a widespread topic of research. In this paper, we report our initial experiments on Memotion Analysis problem through joint embeddings. Results are marginally yielding SOTA., Accepted as Student Abstract at AAAI-22

Contending hate speech in social media is one of the most challenging social problems of our time. There are various types of anti-social behavior in social media. Foremost of them is aggressive behavior, which is causing many social issues such as affecting the social lives and mental health of social media users. In this paper, we propose an end-to-end ensemble-based architecture to automatically identify and classify aggressive tweets. Tweets are classified into three categories - Covertly Aggressive, Overtly Aggressive, and Non-Aggressive. The proposed architecture is an ensemble of smaller subnetworks that are able to characterize the feature embeddings effectively. We demonstrate qualitatively that each of the smaller subnetworks is able to learn unique features. Our best model is an ensemble of Capsule Networks and results in a 65.2% F1 score on the Facebook test set, which results in a performance gain of 0.95% over the TRAC-2018 winners. The code and the model weights are publicly available at https://github.com/parthpatwa/Hater-O-Genius-Aggression-Classification-using-Capsule-Networks., Comment: Accepted at the 17th International Conference on Natural Language Processing (ICON 2020)

NLP applications for code-mixed (CM) or mix-lingual text have gained a significant momentum recently, the main reason being the prevalence of language mixing in social media communications in multi-lingual societies like India, Mexico, Europe, parts of USA etc. Word embeddings are basic build-ing blocks of any NLP system today, yet, word embedding for CM languages is an unexplored territory. The major bottleneck for CM word embeddings is switching points, where the language switches. These locations lack in contextually and statistical systems fail to model this phenomena due to high variance in the seen examples. In this paper we present our initial observations on applying switching point based positional encoding techniques for CM language, specifically Hinglish (Hindi - English). Results are only marginally better than SOTA, but it is evident that positional encoding could bean effective way to train position sensitive language models for CM text., Comment: Accepted as Student Abstract at AAAI 2022

Along with COVID-19 pandemic we are also fighting an `infodemic'. Fake news and rumors are rampant on social media. Believing in rumors can cause significant harm. This is further exacerbated at the time of a pandemic. To tackle this, we curate and release a manually annotated dataset of 10,700 social media posts and articles of real and fake news on COVID-19. We benchmark the annotated dataset with four machine learning baselines - Decision Tree, Logistic Regression, Gradient Boost, and Support Vector Machine (SVM). We obtain the best performance of 93.46% F1-score with SVM. The data and code is available at: https://github.com/parthpatwa/covid19-fake-news-dectection, Published at CONSTRAINT-2021, Collocated with AAAI-2021

In this paper, we present a novel hostility detection dataset in Hindi language. We collect and manually annotate ~8200 online posts. The annotated dataset covers four hostility dimensions: fake news, hate speech, offensive, and defamation posts, along with a non-hostile label. The hostile posts are also considered for multi-label tags due to a significant overlap among the hostile classes. We release this dataset as part of the CONSTRAINT-2021 shared task on hostile post detection.

Additional file 1. Methods, data on identification and biological activity profile of fungi and isolated active secondary metabolite.

Users in social networks either unicast or broadcast their messages. At mention is the popular way of unicasting for Twitter whereas general tweeting could be considered as broadcasting method. Understanding the information flow and dynamics within a Social Network and modeling the same is a promising and an open research area called Information Diffusion. This paper seeks an answer to a fundamental question - understanding if the at-mention network or the unicasting pattern in social media is purely random in nature or is there any user specific selectional preference? To answer the question we present an empirical analysis to understand the sociological aspects of Twitter mentions network within a social network community. To understand the sociological behavior we analyze the values (Schwartz model: Achievement, Benevolence, Conformity, Hedonism, Power, Security, Self-Direction, Stimulation, Traditional and Universalism) of all the users. Empirical results suggest that values traits are indeed salient cue to understand how the mention-based communication network functions. For example, we notice that individuals possessing similar values unicast among themselves more often than with other value type people. We also observe that traditional and self-directed people do not maintain very close relationship in the network with the people of different values traits.

Sentiment analysis is essential in many real-world applications such as stance detection, review analysis, recommendation system, and so on. Sentiment analysis becomes more difficult when the data is noisy and collected from social media. India is a multilingual country; people use more than one languages to communicate within themselves. The switching in between the languages is called code-switching or code-mixing, depending upon the type of mixing. This paper presents overview of the shared task on sentiment analysis of code-mixed data pairs of Hindi-English and Bengali-English collected from the different social media platform. The paper describes the task, dataset, evaluation, baseline and participant's systems.

Bio-activity directed investigation of hexane extract from the leaves of Premna tomentosa led to the isolation of three new clerodane diterpenes (1-3) along with four known compounds (4-7). The structures of new compounds were established using IR, MS, 1D, and 2D NMR techniques. The in vitro cytotoxicity of the crude hexane extract and the isolated compounds were evaluated against seven human cancer cell lines and results indicated that compounds 2 and 4 depicted significant cytotoxicity against hepatocellular carcinoma cell line.

Forlagets publiserte versjon

Automatic understanding of noisy social media text is one of the prime present-day research areas. Most research has so far concentrated on English texts ; however, more than half of the users are writing in other languages, making language identification a pre-requisite for comprehensive processing of social media text. Though language identification has been considered an almost solved problem in other applications, language detectors fail in the social media context due to phenomena such as code-mixing, code-switching, lexical borrowings, Anglicisms, and phonetic typing. This paper reports an initial study to understand the characteristics of code-mixing in the social media context and presents a system developed to automatically detect language boundaries in code-mixed social media text, here exemplified by Facebook messages in mixed English-Bengali and English-Hindi. RÉSUMÉ: La compréhension automatique du texte bruyant des médias sociaux est l’un des secteurs de recherche contemporaine principaux. Jusqu’ici, la plupart des recherches se sont concentrées sur les textes en anglais ; mais plus de la moitié des utilisateurs écrivent dans d’autres langues, ce qui rend l’identification de la langue préalable au traitement complet du texte des médias sociaux. Bien que l’identification de la langue ait été considérée comme un problème presque résolu dans d’autres applications, les détecteurs de langue échouent dans le contexte des médias sociaux, et cela est dû aux phénomènes tels que le mélange et l’alternance de code linguistique, les emprunts lexicaux, les anglicismes et la dactylographie phonétique. Cet article présente une étude initiale pour comprendre les caractéristiques de mélange des codes dans le contexte des médias sociaux ainsi qu’un système développé pour détecter automatiquement les barrières linguistiques en texte « code-mélangé » de médias sociaux, ici illustrées par des messages de Facebook en mixte anglais-bengali et anglais-hindi. Published article - an open access journal

At an injury site, efficient clearance of apoptotic cells by wound macrophages or efferocytosis is a prerequisite for the timely resolution of inflammation. Emerging evidence indicates that microRNA-21 (miR-21) may regulate the inflammatory response. In this work, we sought to elucidate the significance of miR-21 in the regulation of efferocytosis-mediated suppression of innate immune response, a key process implicated in resolving inflammation following injury. An increased expression of inducible miR-21 was noted in postefferocytotic peripheral blood monocyte-derived macrophages. Such induction of miR-21 was associated with silencing of its target genes PTEN and PDCD4. Successful efferocytosis of apoptotic cells by monocyte-derived macrophages resulted in the suppression of LPS-induced NF-κB activation and TNF-α expression. Interestingly, bolstering of miR-21 levels alone, using miR mimic, resulted in significant suppression of LPS-induced TNF-α expression and NF-κB activation. We report that efferocytosis-induced miR-21, by silencing PTEN and GSK3β, tempers the LPS-induced inflammatory response. Macrophage efferocytosis is known to trigger the release of anti-inflammatory cytokine IL-10. This study demonstrates that following successful efferocytosis, miR-21 induction in macrophages silences PDCD4, favoring c-Jun-AP-1 activity, which in turn results in elevated production of anti-inflammatory IL-10. In summary, this work provides direct evidence implicating miRNA in the process of turning on an anti-inflammatory phenotype in the postefferocytotic macrophage. Elevated macrophage miR-21 promotes efferocytosis and silences target genes PTEN and PDCD4, which in turn accounts for a net anti-inflammatory phenotype. Findings of this study highlight the significance of miRs in the resolution of wound inflammation.

Language identification at the document level has been considered an almost solved problem in some application areas, but language detectors fail in the social media context due to phenomena such as utterance internal code-switching, lexical borrowings, and phonetic typing; all implying that language identification in social media has to be carried out at the word level. The paper reports a study to detect language boundaries at the word level in chat message corpora in mixed EnglishBengali and English-Hindi. We introduce a code-mixing index to evaluate the level of blending in the corpora and describe the performance of a system developed to separate multiple languages. Forlagets publiserte versjon

Cirrhosis is associated with prominent changes in sinusoidal structure and function. Although the resident pericyte in liver, the hepatic stellate cell (HSC), is well characterized in the process of fibrogenesis, signaling pathways that regulate HSC vascular function are less developed. Because pericyte populations outside the liver are increasingly being recognized as a key cell type for angiogenesis and changes in vascular structure, in this study, we explore new HSC-signaling pathways that regulate sinusoidal structure and function.Real-time video microscopy and quantitative software analysis of vascular tube formation were used to measure HSC angiogenesis in vitro. Platelet-derived growth factor (PDGF) and ephrin-signaling pathways were modulated using molecular and pharmacologic techniques. Complementary whole animal studies were performed to correlate in vitro findings with pericyte functions in vivo.We show that PDGF promotes a phenotype of HSC evidenced by enhanced HSC-driven vascular tube formation in vitro and enhanced HSC coverage of sinusoids in vivo. This angiogenic phenotype modulates specific pericyte vascular functions including permeability and pressure regulation. Furthermore, we identify a key role for ephrin-B2 as a downstream effector of PDGF signaling.These studies elucidate novel HSC-signaling pathways that regulate microvascular structure and function in liver.

Purpose: To evaluate the efficacy of topical (1%) and systemic itraconazole against common fungi such as Aspergillus and other filamentous fungi that cause mycotic corneal ulcer. Methods: A prospective randomised, controlled study was done in 54 clinically suspected cases of fungal keratitis of which 44 were culture proven. Half the cases (n=27) with superficial involvement were treated with only topical itraconazole (1%) and the other half were treated with both topical and systemic itraconazole. Results:Aspergillus, Penicillium and Fusarium were the most common fungi isolated. The ulcer resolved in 42 eyes (77%) and 12 eyes (23%) did not respond well to treatment. Four of 12 non-responding eyes were caused by Fusarium species. Conclusion: Itraconazole, given either topically or systemically, is effective in treating mycotic corneal ulcers.

Wooden intraorbital foreign body is characteristic for delayed manifestation, silent progression and unpredictable outcome. A silent wooden intraorbital foreign body is difficult to diagnose clinically. Spontaneous expulsion of entire foreign body is rare.

The work described in this report is directed at understanding transport physics in sub-micron heterostructure devices, at developing computational techniques for modeling such devices, and at applying these techniques to investigate new device concepts. The focus of the past year’s work has been on extending our collisionless, quantum device models to treat elastic scattering processes and at applying previously-developed models to the design and study of AlGaAs/GaAs heterojunction bipolar transistors. This report describes the past year’s progress in these two areas. As a by-product of the research, several heterostructure device models have been developed, 1- and 2-D equilibrium models, 1- and 2-D drift-diffusion models, a 1-D Monte Carlo simulator and a 1-D, collisionless quantum device model. These simulation programs are being applied to advanced device analysis at a number of laboratories and are available to SRC members on request.

The work described in this report is directed at understanding quantum transport phenomena in sub-micron heterostructure devices, at developing computational techniques for modeling such devices, and at applying these techniques to develop new device concepts. During the past year we have (l) applied a previously developed collisionless quantum device model (SEQUAL) and Monte Carlo model (DEMON) to the design and study of heterojunction bipolar transistors (Chapter 2); (2) developed a technique for the analysis of arbitrarily shaped quantum devices with elastic scattering (Chapter 3); and (3) developed an approach for incorporating inelastic dissipative processes in quantum transport theory (Chapter 4). As a by-product of the research, several heterostructure device models have been developed: 1- and 2-D equilibrium models, 1- and 2-D drift-diffusion models, a I-D Monte Carlo simulator and a 1-D collisionless quantum device model. These simulation programs are being applied to advanced device analysis at a number of laboratories and are available to SRC members on request

Electron transport in AlGaAs/GaAs heterojunction bipolar transistors with compositionally abrupt emitter‐base junctions is examined. Transport across the abrupt emitter‐base heterojunction is treated quantum mechanically, and the Monte Carlo technique is used to study transport through the base. Although there is a sizeable population of upper‐valley electrons in the bulk emitter, the AlGaAs/GaAs heterojunction is found to favor the injection of Γ‐valley electrons into the base. This valley filtering effect enhances device performance by reducing base transit time, but quantum mechanical tunneling lowers the average energy of the injected flux which increases base transit time. The design of a heterojunction bipolar transistor for minimum base transit time involves a careful tradeoff between these competing factors. We examine the influence of varying aluminum fraction and bias on base transit time. The results suggest that a moderately doped emitter with high aluminum mole fraction produces the shortest base transit time.

Leishmaniasis, a vector-borne disease, is caused by intracellular parasite Leishmania donovani. Unlike most intracellular pathogens, Leishmania donovani are lodged in parasitophorous vacuoles and replicate within the phagolysosomes in macrophages. Effective vaccines against this disease are still under development, while the efficacy of the available drugs is being questioned owing to the toxicity for nonspecific distribution in human physiology and the reported drug-resistance developed by Leishmania donovani. Thus, a stimuli-responsive nanocarrier that allows specific localization and release of the drug in the lysosome has been highly sought after for addressing two crucial issues, lower drug toxicity and a higher drug efficacy. We report here a unique lysosome targeting polymeric nanocapsules, formed via inverse mini-emulsion technique, for stimuli-responsive release of the drug miltefosine in the lysosome of macrophage RAW 264.7 cell line. A benign polymeric backbone, with a disulfide bonding susceptible to an oxidative cleavage, is utilized for the organelle-specific release of miltefosine. Oxidative rupture of the disulfide bond is induced by intracellular glutathione (GSH) as an endogenous stimulus. Such a stimuli-responsive release of the drug miltefosine in the lysosome of macrophage RAW 264.7 cell line over a few hours helped in achieving an improved drug efficacy by 200 times as compared to pure miltefosine. Such a drug formulation could contribute to a new line of treatment for leishmaniasis. A. Das acknowledges SERB (India) Grants (CRG/2020/000492 and JCB/2017/000004) and DBT Grant (BT/PR22251/NNT/28/1274/2017) for supporting this research. N. Mukherjee acknowledges SERB (India) Grant PDF/2016/001437 and K. Das acknowledges the grant EMR/2015/001674 for supporting this research. Financial support from DST (DST/INSPIRE/03/2017/002477) is acknowledged by R.T. This manuscript bears CSMCRI registration no 7/2021. Pramanik, SK (corresponding author), CSIR Cent Salt & Marine Chem Res Inst, Bhavnagar 364002, Gujarat, India. Mukherjee, N (corresponding author), CSIR Indian Inst Chem Biol, Canc Biol & Inflammatory Disorder Div, Kolkata 700032, India. Chattopadhy, S (corresponding author), BITS Pilani, Pilani 403726, Goa, India. Das, A (corresponding author), Indian Inst Sci Educ & Res Kolkata, Mohanpur 741246, W Bengal, India. niladrizoology@gmail.com; sumitpramanik@csmcri.res.in; a.das@csmri.res.in