P Deepak

Anoop Kadan, Deepak P., Sahely Bhadra et al.

Groundbreaking inventions and highly significant performance improvements in deep learning based Natural Language Processing are witnessed through the development of transformer based large Pre-trained Language Models (PLMs). The wide availability of unlabeled data within human generated data deluge along with self-supervised learning strategy helps to accelerate the success of large PLMs in language generation, language understanding, etc. But at the same time, latent historical bias/unfairness in human minds towards a particular gender, race, etc., encoded unintentionally/intentionally into the corpora harms and questions the utility and efficacy of large PLMs in many real-world applications, particularly for the protected groups. In this paper, we present an extensive investigation towards understanding the existence of "Affective Bias" in large PLMs to unveil any biased association of emotions such as anger, fear, joy, etc., towards a particular gender, race or religion with respect to the downstream task of textual emotion detection. We conduct our exploration of affective bias from the very initial stage of corpus level affective bias analysis by searching for imbalanced distribution of affective words within a domain, in large scale corpora that are used to pre-train and fine-tune PLMs. Later, to quantify affective bias in model predictions, we perform an extensive set of class-based and intensity-based evaluations using various bias evaluation corpora. Our results show the existence of statistically significant affective bias in the PLM based emotion detection systems, indicating biased association of certain emotions towards a particular gender, race, and religion.

Deepak P

This article critically examines the recent hype around AI safety. We first start with noting the nature of the AI safety hype as being dominated by governments and corporations, and contrast it with other avenues within AI research on advancing social good. We consider what 'AI safety' actually means, and outline the dominant concepts that the digital footprint of AI safety aligns with. We posit that AI safety has a nuanced and uneasy relationship with transparency and other allied notions associated with societal good, indicating that it is an insufficient notion if the goal is that of societal good in a broad sense. We note that the AI safety debate has already influenced some regulatory efforts in AI, perhaps in not so desirable directions. We also share our concerns on how AI safety may normalize AI that advances structural harm through providing exploitative and harmful AI with a veneer of safety.

Sujit Mandava, Deepak P, Sahely Bhadra

While there has been much research into developing AI techniques for fake news detection aided by various benchmark datasets, it has often been pointed out that fake news in different geo-political regions traces different contours. In this work we uncover, through analytical arguments and empirical evidence, the existence of an important characteristic in news originating from the Global South viz., the geo-political veracity gradient. In particular, we show that Global South news about topics from Global North -- such as news from an Indian news agency on US elections -- tend to be less likely to be fake. Observing through the prism of the political economy of fake news creation, we posit that this pattern could be due to the relative lack of monetarily aligned incentives in producing fake news about a different region than the regional remit of the audience. We provide empirical evidence for this from benchmark datasets. We also empirically analyze the consequences of this effect in applying AI-based fake news detection models for fake news AI trained on one region within another regional context. We locate our work within emerging critical scholarship on geo-political biases within AI in general, particularly with AI usage in fake news identification; we hope our insight into the geo-political veracity gradient could help steer fake news AI scholarship towards positively impacting Global South societies.

Deepak P, Sowmya S Sundaram

Pervasiveness of tracking devices and enhanced availability of spatially located data has deepened interest in using them for various policy interventions, through computational data analysis tasks such as spatial hot spot detection. In this paper, we consider, for the first time to our best knowledge, fairness in detecting spatial hot spots. We motivate the need for ensuring fairness through statistical parity over the collective population covered across chosen hot spots. We then characterize the task of identifying a diverse set of solutions in the noteworthiness-fairness trade-off spectrum, to empower the user to choose a trade-off justified by the policy domain. Being a novel task formulation, we also develop a suite of evaluation metrics for fair hot spots, motivated by the need to evaluate pertinent aspects of the task. We illustrate the computational infeasibility of identifying fair hot spots using naive and/or direct approaches and devise a method, codenamed {\it FiSH}, for efficiently identifying high-quality, fair and diverse sets of spatial hot spots. FiSH traverses the tree-structured search space using heuristics that guide it towards identifying effective and fair sets of spatial hot spots. Through an extensive empirical analysis over a real-world dataset from the domain of human development, we illustrate that FiSH generates high-quality solutions at fast response times.

Martins Samuel Dogo, Deepak P, Anna Jurek-Loughrey

The spread of fake news remains a serious global issue; understanding and curtailing it is paramount. One way of differentiating between deceptive and truthful stories is by analyzing their coherence. This study explores the use of topic models to analyze the coherence of cross-domain news shared online. Experimental results on seven cross-domain datasets demonstrate that fake news shows a greater thematic deviation between its opening sentences and its remainder.

Soumya Suvra Ghosal, Deepak P, Anna Jurek-Loughrey

Disinformation is often presented in long textual articles, especially when it relates to domains such as health, often seen in relation to COVID-19. These articles are typically observed to have a number of trustworthy sentences among which core disinformation sentences are scattered. In this paper, we propose a novel unsupervised task of identifying sentences containing key disinformation within a document that is known to be untrustworthy. We design a three-phase statistical NLP solution for the task which starts with embedding sentences within a bespoke feature space designed for the task. Sentences represented using those features are then clustered, following which the key sentences are identified through proximity scoring. We also curate a new dataset with sentence level disinformation scorings to aid evaluation for this task; the dataset is being made publicly available to facilitate further research. Based on a comprehensive empirical evaluation against techniques from related tasks such as claim detection and summarization, as well as against simplified variants of our proposed approach, we illustrate that our method is able to identify core disinformation effectively.

Deepak P, Savitha Sam Abraham

Incorporating fairness constructs into machine learning algorithms is a topic of much societal importance and recent interest. Clustering, a fundamental task in unsupervised learning that manifests across a number of web data scenarios, has also been subject of attention within fair ML research. In this paper, we develop a novel notion of fairness in clustering, called representativity fairness. Representativity fairness is motivated by the need to alleviate disparity across objects' proximity to their assigned cluster representatives, to aid fairer decision making. We illustrate the importance of representativity fairness in real-world decision making scenarios involving clustering and provide ways of quantifying objects' representativity and fairness over it. We develop a new clustering formulation, RFKM, that targets to optimize for representativity fairness along with clustering quality. Inspired by the $K$-Means framework, RFKM incorporates novel loss terms to formulate an objective function. The RFKM objective and optimization approach guides it towards clustering configurations that yield higher representativity fairness. Through an empirical evaluation over a variety of public datasets, we establish the effectiveness of our method. We illustrate that we are able to significantly improve representativity fairness at only marginal impact to clustering quality.

Deepak P

Within the relatively busy area of fair machine learning that has been dominated by classification fairness research, fairness in clustering has started to see some recent attention. In this position paper, we assess the existing work in fair clustering and observe that there are several directions that are yet to be explored, and postulate that the state-of-the-art in fair clustering has been quite parochial in outlook. We posit that widening the normative principles to target for, characterizing shortfalls where the target cannot be achieved fully, and making use of knowledge of downstream processes can significantly widen the scope of research in fair clustering research. At a time when clustering and unsupervised learning are being increasingly used to make and influence decisions that matter significantly to human lives, we believe that widening the ambit of fair clustering is of immense significance.

Deepak P, Savitha Sam Abraham

An outlier detection method may be considered fair over specified sensitive attributes if the results of outlier detection are not skewed towards particular groups defined on such sensitive attributes. In this task, we consider, for the first time to our best knowledge, the task of fair outlier detection. In this work, we consider the task of fair outlier detection over multiple multi-valued sensitive attributes (e.g., gender, race, religion, nationality, marital status etc.). We propose a fair outlier detection method, FairLOF, that is inspired by the popular LOF formulation for neighborhood-based outlier detection. We outline ways in which unfairness could be induced within LOF and develop three heuristic principles to enhance fairness, which form the basis of the FairLOF method. Being a novel task, we develop an evaluation framework for fair outlier detection, and use that to benchmark FairLOF on quality and fairness of results. Through an extensive empirical evaluation over real-world datasets, we illustrate that FairLOF is able to achieve significant improvements in fairness at sometimes marginal degradations on result quality as measured against the fairness-agnostic LOF method.

Sridhama Prakhya, Deepak P

Quite often, words from one language are adopted within a different language without translation; these words appear in transliterated form in text written in the latter language. This phenomenon is particularly widespread within Indian languages where many words are loaned from English. In this paper, we address the task of identifying loanwords automatically and in an unsupervised manner, from large datasets of words from agglutinative Dravidian languages. We target two specific languages from the Dravidian family, viz., Malayalam and Telugu. Based on familiarity with the languages, we outline an observation that native words in both these languages tend to be characterized by a much more versatile stem - stem being a shorthand to denote the subword sequence formed by the first few characters of the word - than words that are loaned from other languages. We harness this observation to build an objective function and an iterative optimization formulation to optimize for it, yielding a scoring of each word's nativeness in the process. Through an extensive empirical analysis over real-world datasets from both Malayalam and Telugu, we illustrate the effectiveness of our method in quantifying nativeness effectively over available baselines for the task.

Savitha Sam Abraham, Deepak P, Sowmya S Sundaram

A clustering may be considered as fair on pre-specified sensitive attributes if the proportions of sensitive attribute groups in each cluster reflect that in the dataset. In this paper, we consider the task of fair clustering for scenarios involving multiple multi-valued or numeric sensitive attributes. We propose a fair clustering method, \textit{FairKM} (Fair K-Means), that is inspired by the popular K-Means clustering formulation. We outline a computational notion of fairness which is used along with a cluster coherence objective, to yield the FairKM clustering method. We empirically evaluate our approach, wherein we quantify both the quality and fairness of clusters, over real-world datasets. Our experimental evaluation illustrates that the clusters generated by FairKM fare significantly better on both clustering quality and fair representation of sensitive attribute groups compared to the clusters from a state-of-the-art baseline fair clustering method.

Iknoor Singh, Deepak P, Anoop K

The generation and spread of fake news within new and online media sources is emerging as a phenomenon of high societal significance. Combating them using data-driven analytics has been attracting much recent scholarly interest. In this study, we analyze the textual coherence of fake news articles vis-a-vis legitimate ones. We develop three computational formulations of textual coherence drawing upon the state-of-the-art methods in natural language processing and data science. Two real-world datasets from widely different domains which have fake/legitimate article labellings are then analyzed with respect to textual coherence. We observe apparent differences in textual coherence across fake and legitimate news articles, with fake news articles consistently scoring lower on coherence as compared to legitimate news ones. While the relative coherence shortfall of fake news articles as compared to legitimate ones form the main observation from our study, we analyze several aspects of the differences and outline potential avenues of further inquiry.

Sumit Bhatia, Deepak P

Ideological leanings of an individual can often be gauged by the sentiment one expresses about different issues. We propose a simple framework that represents a political ideology as a distribution of sentiment polarities towards a set of topics. This representation can then be used to detect ideological leanings of documents (speeches, news articles, etc.) based on the sentiments expressed towards different topics. Experiments performed using a widely used dataset show the promise of our proposed approach that achieves comparable performance to other methods despite being much simpler and more interpretable.

Deepak P

Differentiating intrinsic language words from transliterable words is a key step aiding text processing tasks involving different natural languages. We consider the problem of unsupervised separation of transliterable words from native words for text in Malayalam language. Outlining a key observation on the diversity of characters beyond the word stem, we develop an optimization method to score words based on their nativeness. Our method relies on the usage of probability distributions over character n-grams that are refined in step with the nativeness scorings in an iterative optimization formulation. Using an empirical evaluation, we illustrate that our method, DTIM, provides significant improvements in nativeness scoring for Malayalam, establishing DTIM as the preferred method for the task.

Oluwaseun Ajao, Deepak P, Jun Hong

The impact of social media and its growing association with the sharing of ideas and propagation of messages remains vital in everyday communication. Twitter is one effective platform for the dissemination of news and stories about recent events happening around the world. It has a continually growing database currently adopted by over 300 million users. In this paper we propose a novel grid-based approach employing supervised Multinomial Naive Bayes while extracting geographic entities from relevant user descriptions metadata which gives a spatial indication of the user location. To the best of our knowledge our approach is the first to make location inference from tweets using geo-enriched grid-based classification. Our approach performs better than existing baselines achieving more than 57% accuracy at city-level granularity. In addition we present a novel framework for content-based estimation of user locations by specifying levels of granularity required in pre-defined location grids.

V. S. Anoop, S. Asharaf, P. Deepak

With the advent of semantic web, various tools and techniques have been introduced for presenting and organizing knowledge. Concept hierarchies are one such technique which gained significant attention due to its usefulness in creating domain ontologies that are considered as an integral part of semantic web. Automated concept hierarchy learning algorithms focus on extracting relevant concepts from unstructured text corpus and connect them together by identifying some potential relations exist between them. In this paper, we propose a novel approach for identifying relevant concepts from plain text and then learns hierarchy of concepts by exploiting subsumption relation between them. To start with, we model topics using a probabilistic topic model and then make use of some lightweight linguistic process to extract semantically rich concepts. Then we connect concepts by identifying an "is-a" relationship between pair of concepts. The proposed method is completely unsupervised and there is no need for a domain specific training corpus for concept extraction and learning. Experiments on large and real-world text corpora such as BBC News dataset and Reuters News corpus shows that the proposed method outperforms some of the existing methods for concept extraction and efficient concept hierarchy learning is possible if the overall task is guided by a probabilistic topic modeling algorithm.