Abstract

Neural models command state-of-the-art per- formance across NLP tasks, including ones involving ‘‘reasoning’’. Models claiming to reason about the evidence presented to them should attend to the correct parts of the input while avoiding spurious patterns therein, be self-consistent in their predictions across in- puts, and be immune to biases derived from their pre-training in a nuanced, context- sensitive fashion. Do the prevalent *BERT- family of models do so? In this paper, we study this question using the problem of rea- soning on tabular data. Tabular inputs are especially well-suited for the study—they ad- mit systematic probes targeting the properties listed above. Our experiments demonstrate that a RoBERTa-based model, representative of the current state-of-the-art, fails at reason- ing on the following counts: it (a) ignores relevant parts of the evidence, (b) is over- sensitive to annotation artifacts, and (c) relies on the knowledge encoded in the pre-trained language model rather than the evidence pre- sented in its tabular inputs. Finally, through inoculation experiments, we show that fine- tuning the model on perturbed data does not help it overcome the above challenges.

Abstract

The paper presents a seismic safety assessment of unreinforced masonry (URM) building using two approaches. The first approach uses the ‘Pier Analysis’ method, based on the concept of equivalent lateral stiffness, where in-plane and out-of-plane actions are considered independently. The second approach is developed with the program SAP2000, where the linear response is evaluated using continuum ‘finite element modelling’ (FEM). Both methods are compared to evaluate the safety of wall piers and the differences in the outcomes under combined gravitational and lateral seismic forces. The analysis results showed that few wall elements are unsafe in in-plane and out-of-plane tension. It is also observed that the pier analysis method is conservative compared to FEM, but can be used as a simplified and quick tool in design offices for safety assessment, with reasonable accuracy. To safeguard the URM wall piers under lateral loads, a retrofitting technique is adopted by providing vertical and horizontal belts called splints and bandages, respectively, using welded wire mesh (WWM) reinforcement. The study using the ‘Pier Analysis’ shows that the lateral load capacity of unsafe URM piers can be enhanced up to 3.67 times and made safe using the applied retrofitting technique. Further, the retrofitting design methodology and recommendations for application procedures to on-site URM buildings are discussed in detail.

Abstract

We devise a novel protocol to detect genuinely multipartite entangled states by harnessing quantum non-Markovian operations. We utilize a particular type of non-Markovianity known as the eternal non-Markovianity to construct a non-complete positive map to filter out the bi-separable states and detect genuine multipartite entanglement. We further propose a witness operator to detect gen- uinely multipartite entangled states experimentally based on this theory. Our study sheds light on a hitherto unexplored connection between entanglement theory and quantum non-Markovianity.

Abstract

Visual data obtained during driving scenarios usually con-tain large amounts of text that conveys semantic information necessary to analyse the urban environment and is integral to the traffic control plan. Yet, research on autonomous driving or driver assistance systems typically ignores this information. To advance research in this direction, we present RoadText-3K, a large driving video dataset with fully anno-tated text. RoadText-3K is three times bigger than its predecessor and contains data from varied geographical locations, unconstrained driving conditions and multiple languages and scripts. We offer a comprehensive analysis of tracking by detection and detection by tracking methods ex-ploring the limits of state-of-the-art text detection. Finally, we propose a new end-to-end trainable tracking model that yields state-of-the-art results on this challenging dataset. Our experiments demonstrate the complexity and variability of RoadText-3K and establish a new, realistic benchmark for scene text tracking in the wild.

Abstract

Maintaining the full scale of naturally occurring river flow is usually impossible due to the development of the water resources and variations of land and soil usage in the catchment. These developed resources can create differences in the balance of the ecosystem and socio-economic activities. These designed structures also cause a decrease in the minimum flow regime in downstream. Environmental flows (EFs) are the medium that help maintain river flow in healthy or ecological conditions. The river’s hydrologic (river mapping), hydraulic (cross-section, water depth, and velocity), and environmental conditions (riparian flora and fauna) are significant considerations for estimating environmental flows (EF’s). There are various desktop assessment methods for calculating the environmental flows. In this present study, the global environmental flow calculator (GEFC) method is used to estimate the ecological flows by using the flow duration curves (FDC) generated from the given monthly discharge data of the river. The FDC in this system contains 17 fixed percentile points concerning the discharge. In the current study, we analyze the environmental flows of the Tungabhadra River basin by considering the different discharge stations, which are Balehonuur, Haralahalli, Hosaritti, Shivamogga, Honalli, Rattihalli, and Tungabhadra Dam, with a mean annual flow (MAF) of 36%, 24.8%, 27.2%, 16.2%, 23.3%, 21.1%, and 12.2%, respectively, to maintain the ecological conditions of the river. The monthly discharge data from 1995 to 2017 for those stations are obtain from the Advance Center for Integrated Water Resource (ACIWR) Bengaluru, India. The river flow health is a study which helps in understanding the environmental variables that effects the habitat structure, flow regime, water quality, and biological conditions of the river. To estimate the Flow Health of Tungabhadra River, we used a tool called Flow Health which uses nine indicators to represent the Flow Health (FH) score for the stations Balehonnur, Haralahalli, Hosaritti, Shivamogga, Honalli, Rattihalli, and Tungabhadra. This tool uses the gauge discharge data in the form of reference (1995–2005) and test periods (2006–2017), with Flow Health score of 0.72, 0.4, 0.72,0.70, 0.58, 0.73, 0.71 and 0.72, 0.63, 0.63, 0.7, 0.66, 0.67, 0.66 for test and reference period with respect to stations. The study noted that majority of the discharge stations along the Tungabhadra River show a moderate to low flow variations for the reference and test periods. Overall, Tungabhadra river health, measured by the flow indices, had declined from 1995–2005 to 2006–2017.

Abstract

Is the capability to improvise only available to humans? Is it possible to generate musical phrases using natural language processing and machine learning? Is it possible to improvise in North Indian classical music using constructs like N-gram models and neural networks? In this paper we propose two methods to generate improvised sequences for North Indian classical music. In first method, we propose a bi-gram language model to generate improvised sequences. We discuss the limitations of this approach. In second method, we propose a LSTM-RNN based model to generate improvised sequences. We discuss the motivation for this method. We discuss model architecture and training settings. We also discuss the limitations of this approach. To assess these approaches, we introduce Pt. Ajoy Chakraborty Bhoopali Rāga Compositions Dataset. We evaluate the bigram model on this dataset to achieve a TOP-1 accuracy of 46.71 percent and a TOP-3 accuracy of 83.81 percent. We also evaluate LSTM-RNN based model on Pt. Ajoy Chakraborty Bhoopali Compositions Dataset. We achieve TOP-1 accuracy of 51.43 percent and TOP-3 accuracy of 81.21 percent.

Abstract

Is the capability to improvise only available to humans? Is it possible to generate musical phrases using Automata Theory? Is it possible to improvise in North Indian classical music using constructs like Context free grammars? In this paper we propose a method to generate improvised sequences for North Indian classical music using context free grammar. We choose to focus on yamanrāga as a case study. We discuss scope of the context free grammar, primary rules for context free grammar based on yamanrāga, rules for stationary, ascending, descending, circulating and microtonal patterns. We also discuss shortcomings of this approach in the discussion section.

Abstract

Political propaganda in recent times has been amplified by media news portals through bi- ased reporting, creating untruthful narratives on serious issues causing misinformed public opinions with interests of siding and helping a particular political party. This issue pro- poses a challenging NLP task of detecting po- litical bias in news articles. We propose a transformer-based transfer learning method to fine-tune the pre-trained network on our data for this bias detection. As the required dataset for this particular task was not available, we created our dataset comprising 1388 Hindi news articles and their headlines from various Hindi news media outlets. We marked them on whether they are biased towards, against, or neutral to BJP, a political party, and the current ruling party at the centre in India.

Abstract

This paper presents our solutions for Task4 at SemEval2022: Patronizing and Condescend- ing Language Detection. This shared task con- tains two sub-tasks. The first sub-task is a bi- nary classification task whose goal is to predict whether a given paragraph contains any form of patronising or condescending language(PCL). For the second sub-task, given a paragraph, we have to find which PCL categories express the condescension. Here we have a total of 7 overlapping sub-categories for PCL. Our pro- posed solution uses BERT based ensembled models with hard voting and techniques applied to take care of class imbalances. Our paper de- scribes the system architecture of the submitted solution and other experiments that we con- ducted. Our best performing models achieve an F1 score of 59.4 and 15.7 on sub-tasks 1 and 2 respectively.

Abstract

We investigate Referring Image Segmentation (RIS), which outputs a segmentation map corresponding to the natural language description. Addressing RIS efficiently requires considering the interactions happening across visual and linguistic modalities and the interactions within each modality. Existing methods are limited because they either compute different forms of interactions sequentially (leading to error propagation) or ignore intramodal interactions. We address this limitation by performing all three interactions simultaneously through a Synchronous Multi-Modal Fusion Module (SFM). Moreover, to produce refined segmentation masks, we propose a novel Hierarchical Cross-Modal Aggregation Module (HCAM), where linguistic features facilitate the exchange of contextual information across the visual hierarchy. We present thorough ablation studies and validate our approach’s performance on four benchmark datasets, showing considerable performance gains over the existing state-of-the-art (SOTA) methods.