Hate speech, quite common in the age of social media, at times harmless but can also cause mental trauma to someone or even riots in communities. Image of a religious symbol with derogatory comment or video of a man abusing a particular community, all become hate speech with its every modality (such as text, image, and audio) contributing towards it. Models based on a particular modality of hate speech post on social media are not useful, rather, we need models like multi-modal fusion models that consider both image and text while classifying hate speech. Text-image fusion models are heavily parameterized, hence we propose a quaternion neural network-based model having additional fusion components for each pair of modalities. The model is tested on the MMHS150K twitter dataset for hate speech classification. The model shows an almost 75% reduction in parameters and also benefits us in terms of storage space and training time while being at par in terms of performance as compared to its real counterpart.

Neural Machine Translation (NMT) models have been effective on large bilingual datasets. However, the existing methods and techniques show that the model's performance is highly dependent on the number of examples in training data. For many languages, having such an amount of corpora is a far-fetched dream. Taking inspiration from monolingual speakers exploring new languages using bilingual dictionaries, we investigate the applicability of bilingual dictionaries for languages with extremely low, or no bilingual corpus. In this paper, we explore methods using bilingual dictionaries with an NMT model to improve translations for extremely low resource languages. We extend this work to multilingual systems, exhibiting zero-shot properties. We present a detailed analysis of the effects of the quality of dictionaries, training dataset size, language family, etc., on the translation quality. Results on multiple low-resource test languages show a clear advantage of our bilingual dictionary-based method over the baselines.

In this work, we explored the eigenspectra of minimally doubled fermions, in both Karsten-Wilczek and Borici-Creutz realizations. We generated 4-dim $SU(3)$ gauge fields with a definite topological charge and calculated the chiralities of the eigenmodes for KW and BC fermions. We used the spectral flow of the eigenvalues for this purpose and demonstrated the Index theorem.

We develop a new refinement of the Kato's inequality and using this refinement we obtain several upper bounds for the numerical radius of a bounded linear operator as well as the product of operators, which improve the well known existing bounds. Further, we obtain a necessary and sufficient condition for the positivity of $2\times 2$ certain block matrices and using this condition we deduce an upper bound for the numerical radius involving a contraction operator. Furthermore, we study the Schatten $p$-norm inequalities for the sum of two $n\times n$ complex matrices via singular values and from the inequalities we obtain the $p$-numerical radius and the classical numerical radius bounds. We show that for every $p>0$, the $p$-numerical radius $w_p(\cdot): \mathcal{M}_n(\mathbb C)\to \mathbb R$ satisfies $w_p(T) \leq \frac12 \sqrt{\left\| |T|^{2(1-t)}+|T^*|^{2(1-t)} \right\|^{} \, \big \||T|^{2t}+|T^*|^{2t} \big\|_{p/2}^{} }$ for all $t\in [0,1]$. Considering $p\to \infty$, we get a nice refinement of the well known classical numerical radius bound $w(T) \leq \sqrt{\frac12 \left\| T^*T+TT^* \right \|}.$ As an application of the Schatten $p$-norm inequalities we develop a bound for the energy of graph. We show that $\mathcal{E}(G) \geq \frac{2m}{ \sqrt{ \max_{1\leq i \leq n} \left\{ \sum_{j, v_i \sim v_j}d_j\right\}} },$ where $\mathcal{E}(G)$ is the energy of a simple graph $G$ with $m$ edges and $n$ vertices $v_1,v_2,\ldots,v_n$ such that degree of $v_i$ is $d_i$ for each $i=1,2,\ldots,n.$

We present a lattice calculation of the electromagnetic (EM) effects on the masses of light pseudoscalar mesons. The simulations employ 2+1 dynamical flavors of asqtad QCD quarks, and quenched photons. Lattice spacings vary from $\approx 0.12$ fm to $\approx 0.045$ fm. We compute the quantity $\epsilon$, which parameterizes the corrections to Dashen's theorem for the $K^+$-$K^0$ EM mass splitting, as well as $\epsilon_{K^0}$, which parameterizes the EM contribution to the mass of the $K^0$ itself. An extension of the nonperturbative EM renormalization scheme introduced by the BMW group is used in separating EM effects from isospin-violating quark mass effects. We correct for leading finite-volume effects in our realization of lattice electrodynamics in chiral perturbation theory, and remaining finite-volume errors are relatively small. While electroquenched effects are under control for $\epsilon$, they are estimated only qualitatively for $\epsilon_{K^0}$, and constitute one of the largest sources of uncertainty for that quantity. We find $\epsilon = 0.78(1)_{\rm stat}({}^{+\phantom{1}8}_{-11})_{\rm syst}$ and $\epsilon_{K^0}=0.035(3)_{\rm stat}(20)_{\rm syst}$. We then use these results on 2+1+1 flavor pure QCD HISQ ensembles and find $m_u/m_d = 0.4529(48)_{\rm stat}( {}_{-\phantom{1}67}^{+150})_{\rm syst}$.