Abstract

This paper describes the Neural Machine Translation systems of IIIT-Hyderabad (LTRC-MT) for WAT 2019 Hindi-English shared task. We experimented with both Recurrent Neural Networks & Transformer architectures. We also show the results of our experiments of training NMT models using additional data via backtranslation.

Abstract

Neural text-to-speech synthesis (NTTS) models have shown significant progress in generating high-quality speech, however they require a large quantity of training data. This makes creating models for multiple styles expensive and time-consuming. In this paper different styles of speech are analysed based on prosodic variations, from this a model is proposed to synthesise speech in the style of a newscaster, with just a few hours of supplementary data. We pose the problem of synthesising in a target style using limited data as that of creating a bi-style model that can synthesise both neutral-style and newscaster-style speech via a one-hot vector which factorises the two styles. We also propose conditioning the model on contextual word embeddings, and extensively evaluate it against neutral NTTS, and neutral concatenative-based synthesis. This model closes the gap in perceived style-appropriateness between natural recordings for newscaster-style of speech, and neutral speech synthesis by approximately two-thirds.

Abstract

Musical training causes structural and functional changes in the brain due to its sensory-motor demands. This leads to differences in how musicians perceive and process music as compared to nonmusicians, thereby providing insights into brain adaptations and plasticity. Correlational studies and network analysis investigations have indicated the presence of large-scale brain networks involved in the processing of music and have highlighted the differences between musicians and non-musicians. However, studies on functional connectivity in the brain during music listening tasks have thus far focused solely on static network analysis. Dynamic Functional Connectivity (DFC) studies have lately been found useful in unearthing meaningful, time-varying functional connectivity information in both resting-state and task-based experimental settings. In this study, we examine DFC in the fMRI obtained from two groups of participants, 18 musicians and 18 non-musicians, while they listened to a musical stimulus in a naturalistic setting. We utilize spatial Group Independent Component Analysis (ICA), sliding time window correlations, and a deterministic agglomerative clustering of windowed correlation matrices to identify quasi-stable Functional Connectivity (FC) states in the two groups. To compute cluster centroids that represent FC states, we device and present a method that primarily utilizes windowed correlation matrices occurring repeatedly over time and across participants, while excluding matrices corresponding to spontaneous fluctuations. Preliminary analysis indicate states with greater visuo-sensorimotor integration in musicians, larger presence of DMN states in non-musicians, and variability in states found in musicians due to differences in training and prior experiences.

Abstract

. Automatic research paper summarization is a fairly interesting topic that has garnered significant interest in the research community in recent years. In this paper, we introduce team Helium’s system description for the CL-SciSumm shared task colocated with SIGIR 2019. We specifically attempt the first task, targeting in building an improved recall system of reference text spans from a given citing research paper (Task 1A) and constructing better models for comprehension of scientific facets (Task 1B). Our architecture incorporates transfer learning by utilising a combination of pretrained embeddings which are subsequently used for building models for the given tasks. In particular - for task 1A, we locate the related text spans referred to by the citation text by creating paired text representations and employ pre-trained embedding mechanisms in conjunction with XGBoost, a gradient boosted decision tree algorithm to identify textual entailment. For task 1B, we make use of the same pretrained embeddings and use the RAKEL algorithm for multi-label classification. Our goal is to enable better scientific research comprehension and we believe that a new approach involving transfer learning will certainly add value to the research community working on these tasks.

Abstract

Lately, Indian online platforms have witnessed recurring consumer boycott calls in forms of uninstalling/down-voting applications on the app stores. For our exploratory study, we conduct a longitudinal analysis of one of these controversies involving online boycotting of an e-commerce company (Snapdeal) over a controversial statement issued by their brand ambassador (Aamir Khan) which hurt religious and nationalist sentiments of users on Twitter. Through the lens of this study, we try to understand emergent collective user behaviour and how collective action has begun to play out in online (troll) communities. We call the user behaviour in our study as trolling behaviour as - the call to boycott appears in order to silence an influential voice which challenges a nationalist narrative of these users. A broader implication of such behaviour seems to be strong arming any counter narrative with a threat of potential backlash and financial harm. This analysis is important as online deviant and trolling behaviour by group of users is increasingly influencing socio-political agendas online. It contributes to broader CSCW understanding of online platforms and collective behaviour. We also situate our work in online consumer boycotting behaviour.

Abstract

No Results Found

Abstract

Recent revision of Indian standard code on earthquake resistant design suggested certain recommendations for inclusion of effects of vertical ground acceleration for design of buildings. Clause 6.3.2.1 of IS 1893 (Part 1) suggests that for non-orthogonal frames, load combinations should include 30% of earthquake load in other directions along with the direction considered and Clause 6.3.2.1.1 suggests that for orthogonal frames, full load has to be applied. Besides, clause 6.3.3 mentions that all the buildings present in seismic zone IV and V have to consider vertical ground motion effects. This results in 73 load combinations for calculating the design forces in structural members. However, the necessity to consider the effects of vertical ground motions for regular buildings, although present in seismic zones IV and V, is not verified. To verify the above clauses for regular buildings, a case study is performed on a 5 storied regular framed RC building. Using, linear static analysis, design forces in structural members for the contradicting clauses on load combinations are compared. From the study, it is concluded that 3-dimensional ground motion effects are not necessary for every building even if the building is located in seismic zone IV or V.

Abstract

We introduce a novel model for future/next page prediction in on- line user journeys that uses a combination of doc2vec webpage representations with an LSTM-based neural network to mine pat- terns from users’ online navigational paths combined with their content preferences. Empirical explorations show promise towards creating customized user experiences leveraging this work.

Abstract

In this paper, the problem of path planning through tight spaces, for the task of spatial payload transportation, using a formation of mobile manipulators is addressed. Due to the high dimensional configuration space of the system, efficient and geometrically stable path planning through tight spaces is challenging. We resolve this by planning the path for the system in two phases. First, an obstacle-free trajectory in R 3 for the payload being transported is determined using RRT. Next, near-energy optimal and quasi-statically stable paths are planned for the formation of robots along this trajectory using non-linear multi-objective optimization. We validate the proposed approach in simulation experiments and compare different multi-objective optimization algorithms to find energy optimal and geometrically stable robot path plans.

Abstract

In the field of hyperspectral remote sensing, the spectral data is derived from hyperspectral images and are matched with the spectral library. Till date there are no studies that have matched with the hand–held spectrum collected by spectroradiometer at temporal scale. Creating a spectral library for the same species at various regions is a challenge, as acquiring spectrum at temporal sets are practically difficult and expensive. The absorption features are the unique key print for the identification of the two different samples of same species and also helps to discriminate two different species. The intensity and position of strong absorption region helps to match the spectrum accurately. In the current study probability based matching technique is used where the dips of a species of a season is matched to the other within a predefined abundance window. This study helps to know the seasonal changes of the tree due to its environmental factors, stress and deficiency.