The literature in automated sarcasm detection has mainly focused on lexical, syntactic and semantic-level analysis of text. However, a sarcastic sentence can be expressed with contextual presumptions, background commonsense knowledge. In this paper, we propose CASCADE (a ContextuAl SarCasm DEtector) that adopts hybrid approach both content context-driven modeling for online social media discussions. For the latter, aims at extracting information from discourse discussion thread. Also, since...

Multimodal emotion recognition is the task of detecting emotions present in user-generated multimedia content. Such resources contain complementary information multiple modalities. A stiff challenge often faced complexity associated with feature-level fusion these heterogeneous modes. In this paper, we propose a new method based on self-attention mechanism. We also compare it traditional methods such as concatenation, outer-product, etc. Analyzed using textual and speech (audio) modalities,...

Multi-Disease Predictive Analytics (MDPA) models simultaneously predict the risks of multiple diseases in patients and are valuable early diagnoses. Patients tend to have or develop complications over time, MDPA can learn effectively utilize such correlations between diseases. Data from large-scale Electronic Health Records (EHR) be used through Multi-Label Learning (MLL) methods models. However, data-driven approaches for face challenge data imbalance, because rare much less than common...

Underserved communities face critical health challenges due to lack of access timely and reliable information. Nongovernmental organizations are leveraging the widespread use cellphones combat these healthcare spread preventative awareness. The workers at reach out individually beneficiaries; however such programs still suffer from declining engagement. We have deployed SAHELI, a system efficiently utilize limited availability for improving maternal child in India. SAHELI uses Restless...

Search and recommendation systems, such as search engines, recruiting tools, online marketplaces, news, social media, output ranked lists of content, products, sometimes, people. Credit ratings, standardized tests, risk assessments only a score, but are also used implicitly for ranking. Bias in ranking especially among the top ranks, can worsen economic inequalities, polarize opinions, reinforce stereotypes. On other hand, bias correction minority groups cause more harm if perceived favoring...

We investigate the problem of probably approximately correct and fair (PACF) ranking items by adaptively evoking pairwise comparisons. Given a set $n$ that belong to disjoint groups, our goal is find an $(\epsilon, \delta)$-PACF-Ranking according objective function we propose. assume access oracle, wherein, for each query, learner can choose pair receive stochastic winner feedback from oracle. Our proposed asks minimize $\ell_q$ norm error where group $\ell_p$ all within group, $p, q \geq...

Randomized rankings have been of recent interest to achieve ex-ante fairer exposure and better robustness than deterministic rankings. We propose a set natural axioms for randomized group-fair prove that there exists unique distribution D satisfies our is supported only over ex-post rankings, i.e., satisfy given lower upper bounds on group-wise representation in the top-k ranks. Our problem formulation works even when implicit bias, incomplete relevance information, or ordinal ranking...

Abstract Underserved communities face critical health challenges due to lack of access timely and reliable information. Nongovernmental organizations are leveraging the widespread use cellphones combat these healthcare spread preventative awareness. The workers at reach out individually beneficiaries; however, such programs still suffer from declining engagement. We have deployed Saheli , a system efficiently utilize limited availability for improving maternal child in India. uses Restless...

With the multitude of companies and organizations abound today, ranking them choosing one out many is a difficult cumbersome task. Although there are available metrics that rank companies, an inherent need for generalized metric takes into account different aspects constitute employee opinions companies. In this work, we aim to overcome aforementioned problem by generating aspect-sentiment based embedding looking reliable reviews them. We created comprehensive dataset company from famous...

Rankings on online platforms help their end-users find the relevant information -- people, news, media, and products quickly. Fair ranking tasks, which ask to rank a set of items maximize utility subject satisfying group-fairness constraints, have gained significant interest in Algorithmic Fairness, Information Retrieval, Machine Learning literature. Recent works, however, identify uncertainty utilities as primary cause unfairness propose introducing randomness output. This is carefully...

In learning-to-rank (LTR), optimizing only the relevance (or expected ranking utility) can cause representational harm to certain categories of items. Moreover, if there is implicit bias in scores, LTR models may fail optimize for true relevance. Previous works have proposed efficient algorithms train stochastic that achieve fairness exposure groups ex-ante (or, expectation), which not guarantee representation ex-post, is, after realizing a from model. Typically, ex-post achieved by...

Rankings on online platforms help their end-users find the relevant information—people, news, media, and products—quickly. Fair ranking tasks, which ask to rank a set of items maximize utility subject satisfying group-fairness constraints, have gained significant interest in Algorithmic Fairness, Information Retrieval, Machine Learning literature. Recent works, however, identify uncertainty utilities as primary cause unfairness propose introducing randomness output. This is carefully chosen...

Randomized rankings have been of recent interest to achieve ex-ante fairer exposure and better robustness than deterministic rankings. We propose a set natural axioms for randomized group-fair prove that there exists unique distribution $D$ satisfies our is supported only over ex-post rankings, i.e., satisfy given lower upper bounds on group-wise representation in the top-$k$ ranks. Our problem formulation works even when implicit bias, incomplete relevance information, or ordinal ranking...

Center-based clustering (e.g., $k$-means, $k$-medians) and using linear subspaces are two most popular techniques to partition real-world data into smaller clusters. However, when the consists of sensitive demographic groups, significantly different cost per point for groups can lead fairness-related harms quality-of-service). The goal socially fair is minimize maximum over all groups. In this work, we propose a unified framework solve center-based subspace clustering, give practical,...