Abstract: This disclosure relates generally to long-form answer extraction and, more particularly, to long-form answer extraction based on combination of sentence index generation techniques. Existing answer extractions techniques have achieved significant progress for extractive short answers; however, less progress has been made for long form questions that require explanations. Further the state-of-art long-answer extractions techniques result in poorer long-form answers or not address sparsity which becomes an issue longer contexts. Additionally, pre-trained generative sequence-to-sequence models are gaining popularity for factoid answer extraction tasks. Hence the disclosure proposes a long-form answer extraction based on several steps including training a set of generative sequence-to-sequence models comprising a sentence indices generation model and a sentence index spans generation.

Abstract: Conventional gesture detection approaches demand large memory and computation power to run efficiently, thus limiting their use in power and memory constrained edge devices. Present application/disclosure provides a Spiking Neural Network based system which is a robust low power edge compatible ultrasound-based gesture detection system. The system uses a plurality of speakers and microphones that mimics a Multi Input Multi Output (MIMO) setup thus providing requisite diversity to effectively address fading. The system also makes use of distinctive Channel Impulse Response (CIR) estimated by imposing sparsity prior for robust gesture detection. A multi-layer Convolutional Neural Network (CNN) has been trained on these distinctive CIR images and the trained CNN model is converted into an equivalent Spiking Neural Network (SNN) via an ANN (Artificial Neural Network)-to-SNN conversion mechanism. The SNN is further configured to detect/classify gestures performed by user(s).

Abstract: Motion parameters estimation for localization of differential drive vehicles is an important part of robotics and autonomous navigation. Conventional methods require introceptive as well extroceptive sensors for localization. The present disclosure provides a control command based adaptive system and method for estimating motion parameters of differential drive vehicles. The method utilizes information from one or more time synchronized command signals and generate an experimental model for estimating one or more motion parameters of the differential drive vehicle by computing a mapping function. The experimental model is validated to determine change in the one or more motion parameters with change in one or more factors and adaptively updated to estimate updated value of the one or more motion parameters based on the validation. The system and method of present disclosure provide accurate results for localization with minimum use of extroceptive sensors.

Abstract: A method and system for automated continuous validation for regulatory compliance of CS with dynamic component. On identification of learning in the CS, a User Acceptance Testing (UAT) is performed using automated test cases of varying types in accordance with what-if scenarios and synthetic data generated using a unique approach. Thereafter, a base validation testing of the CS is performed with clean data (positive scenarios of outcome of the CS) and dirty data (negative scenarios) by conducting repeatability, stability (consistency) and reliability checks. The base validation testing is then followed by learning saturation testing on only if the dynamic component is validated, is rolled out in production environment else is rolled back to the earlier version.

Abstract: This disclosure relates generally to program synthesis for weakly-supervised multimodal question answering using filtered iterative back-translation (FIBT). Existing approaches for chart question answering mainly address structural, visual, relational, or simple data retrieval queries with fixed-vocabulary answers. The present disclosure implements a two-stage approach where, in first stage, a computer vision pipeline is employed to extract data from chart images and store in a generic schema. In second stage, SQL programs for Natural Language (NL) queries are generated in dataset by using FIBT. To adapt forward and backward models to required NL queries, a Probabilistic Context-Free Grammar is defined, whose probabilities are set to be inversely proportional to SQL programs in training data and sample programs from it.

Abstract: Conventional Question and Answer (QA) datasets are created for generating factoid questions only and the present disclosure generates longform technical QA dataset from textbooks. Initially, the system receives a technical textbook document and extracts a plurality of contexts. Further, a first plurality of questions are generated based on the plurality of contexts. A plurality of answerable questions are generated further based on the plurality of contexts using an unsupervised template-based matching technique. Further, a combined plurality of questions are generated by combining the first plurality of questions and the plurality of answerable questions. Further, an answer for the combined plurality of questions are generated using an autoregressive language model and a mapping score is computed. Further, a plurality of optimal answers are selected based on the corresponding mapping score.

Abstract: The present disclosure provides a method for surface wear inspection using millimeter wave radar. The system initially receives a plurality of uncompressed raw Synthetic Aperture Radar (SAR) images. Further, a plurality of reconstructed SAR images are generated based on the plurality of uncompressed raw SAR images using a variable focusing based Range Doppler Algorithm (RDA). Further, a master image and a slave image are selected from the reconstructed SAR images and corresponding anchor points are assigned. Further a plurality of fine level and coarse level shift coordinates are computed based on the corresponding anchor points. Further, a net shift value is computed based on the plurality of fine level and coarse level shift coordinates. The master and the slave images are aligned based on the net shift value and the interferogram is generated. The interferogram is further analyzed to profile the corresponding deformation pertaining to the surface under test.

Abstract: Existing systems for behavioural tracking and identification have the disadvantage that they do not analyse data in behavioural aspects. As a result, they lack ability to pre-empt scenarios involving actions that adversely affect user health. The disclosure herein generally relates to behavior prediction, and, more particularly, to a method and system for identifying unhealthy behavior trigger and providing nudges. The system generates a casual inference model, which is a reverse causality model facilitating mapping of context with one or more behaviour of the user. The system further collects and processes real-time data using the casual inference model, to perform behavioral analysis of the user.

Abstract: Works in the literature fail to leverage embedding access patterns and memory units' access/storage capabilities, which when combined can yield high-speed heterogeneous systems by dynamically re-organizing embedding tables partitions across hardware during inference. A method and system for optimal deployment of embeddings tables across heterogeneous memory architecture for high-speed recommendations inference is disclosed, which dynamically partitions and organizes embedding tables across fast memory architectures to reduce access time. Partitions are chosen to take advantage of the past access patterns of those tables to ensure that frequently accessed data is available in the fast memory most of the time. Partition and replication is used to co-optimize memory access time and resources.

Abstract: Data analysis plays a crucial role to get significant information out of the data. A real time system and method for analyzing data streams have been provided. The system can utilize many different types of data formats such as numeric, text, video, audio, image, or combination thereof. The analysis takes place as per the requirement using an analytical engine and an intermediate output is generated. The intermediate output is further processed using a distributed real time business rule processing engine to determine required conditions in the data. The business rules comprise one or more set of meta data. On match of the business rule, the system triggers an alert or propagates the required information to integrating solution for required actions. The system and method are technology and communication protocol agnostic, and designed with highly efficient load balanced technique, thereby facilitating highly concurrent data processing with minimal latency.

Abstract: Elderly people suffer from health issues, and timely detection can save lives. State of the art techniques either make certain assumptions or require clinical data in order to perform the frailty detection, which affects the quality as well as cause inconvenience to the users. The disclosure herein generally relates to patient monitoring and, more particularly, to frailty detection using pedometer sensor data, PIR sensor data, and door sensor data. The system determines activity levels of the user being monitored, based on data from the pedometer sensors, PIR sensors, and door sensors, and based on the determined activity levels, further determines whether the user has frailty or not.

Abstract: The present disclosure provides a system and method for delay prediction for scheduled public transport. A multi-architectural deep learning approach has been used to predict the delays of a queried vehicle in the scheduled public transport. For this, historical operational data is transformed into temporal, and spatiotemporal data. While, the spatial data is obtained from geographical information. The system uses different combinations of neural networks architectures. A regressor model uses three separate kinds of architecture. One component is the Fully Connected Neural Network (FCNN), which is good at learning from static features, the second is the Long Short Term Memory (LSTM) network which is good at learning from temporal features, and the third is the 3D Convolutional Neural Network (3DCNN) which is good at learning from spatiotemporal features. Learned encoding from each are fed to another FCNN to produce the predicted delay value.

Abstract: The disclosure generally relates to an FPGA-based online 3D bin packing. Online 3D bin packing is the process of packing boxes into larger bins-Long Distance Containers (LDCs) such that the space inside each LDC is used to the maximum extent. The use of deep reinforcement learning (Deep RL) for this process is effective and popular. However, since the existing processor-based implementations are limited by Von-Neumann architecture and take a long time to evaluate each alignment for a box, only a few potential alignments are considered, resulting in sub-optimal packing efficiency. This disclosure describes an architecture for bin packing which leverages pipelining and parallel processing on FPGA for faster and exhaustive evaluation of all alignments for each box resulting in increased efficiency. In addition, a suitable generic purpose processor is employed to train the neural network within the algorithm to make the disclosed techniques computationally light, faster and efficient.

Abstract: Technical interviewing is important for organizations for assessing a candidate to make hiring decision. For effective technical interviewing, predicting difficulty of long form technical questions is crucial. The present disclosure provides systems and methods for predicting difficulty of long form technical questions using weak supervision from textbooks. Further, zero shot pre-trained large language models and unsupervised template-based technique are used for generating questions. Furthermore, a difficulty score is assigned to the generated questions based on context difficulty and task difficulty. The context difficulty for the generated questions is computed using hierarchical structure of the textbooks, and the task difficulty is computed by determining a similarity between the generated questions and Bloom's taxonomy levels.

Abstract: Performance of a machine learning (ML) model in production, is heavily dependent on underlying distribution of data or underlying process generating labels from attributes. Any change in either one or both impacts the ML model performance heavily and inhibits knowledge of true labels. This in turn affects ML model uncertainty. Thus, performance monitoring of ML models in production becomes necessary. Embodiments of the present disclosure provide estimates operating model accuracy at production stage by constructing the correlations between the model accuracy, model uncertainty and deviation of the distributions in absence of ground truth. In the method of present disclosure, the model performance of the machine learning (ML) model deployed in production is estimated in absence of ground truths. Moreover, this can be done without retraining the model, thus saving computational costs and resources. The method of the present disclosure can be used and performed in real time.

Abstract: The disclosure relates to sequencing of asset segments of privacy policies. The asset segments are sequenced based on a set of constraints. In an embodiment the asset segments are sequenced using a set of pre-defined predecessors and a set of pre-defined successors of each asset segment through a feasible sequence generation technique and a sequence generation technique based on the constraints, wherein the constraints are preferences associated with the source entity and the target entity. Hence the disclosure bridges a communication gap between the source entity and the target entity by optimally displaying the most relevant privacy policy (mapped to the asset segments) based on the constraints associated with the source entity and the target entity. Further the disclosed system also determines a violation factor that represents a conflict between the preferences associated with the source entity and the target entity.

Abstract: Existing approaches in pharmaceutical pricing, fail to provide visibility on nature of pricing followed by pharma players such as pharmacy benefit managers (PBMs), manufacturer, distributor, insurer, and so on, due to involvement of many players in pharma value chain and their complicated pricing strategies. The disclosure herein generally relates to decision support system for pharmaceutical pricing, and, more particularly, to a method and system for providing visibility on nature of pricing followed by different entities of pharma players. The system, by performing pricing analysis, extracts a magnitude of interrelationship between the plurality of entities in the pharmaceutical domain to form a pharmaceutical pricing guide. The pharmaceutical pricing guide is further processed to maximize a measured quality of the pharmaceutical pricing guide in real time and used to choose entities of pharma players associated with retail pharmacy.

Abstract: This disclosure relates generally to system and method for ambient intelligence based user interaction. Prior methods for touchless user interaction are sensitive to ambient temperature in a lab environment, susceptible to noise from metallic surfaces and ambient radio waves and are dependent on ambient lighting. Embodiments of the present disclosure provides a multi-modal sensor fusion method which captures touchless gestures from a user or a group of users with their physical context information fused and tagged to these gestures for user interaction. Further pose graphs are generated for user interaction systems using a data association technique and Gaussian mixture model technique. The disclosed method provides a hands-free interface to operate instruments in a smart space, using principles of ambient intelligence.

Abstract: Model-based image reconstruction (MBIR) methods using convolutional neural networks (CNNs) as priors have demonstrated superior image quality and robustness compared to conventional methods. Studies have explored MBIR combined with supervised and unsupervised denoising techniques for image reconstruction in magnetic resonance imaging (MRI) and positron emission tomography (PET). Unsupervised methods like the deep image prior (DIP) have shown promising results and are less prone to hallucinations. However, since the noisy image is used as a reference, strategies to prevent overfitting are unclear. Recently, Bayesian DIP (BDIP) networks that model uncertainty tend to prevent overfitting without requiring early stopping. However, BDIP has not been studied with data-fidelity term for image reconstruction. Present disclosure provides systems and method that implement a MBIR framework with a modified BDIP.

Abstract: This disclosure relates generally to the field of Electroencephalogram (EEG) classification, and, more particularly, to method and system for EEG motor imagery classification. Existing deep learning works employ the sensor-space for EEG graph representations wherein the channels of the EEG are considered as nodes and connection between the nodes are either predefined or are based on certain heuristics. However, these representations are ineffective and fail to accurately capture the underlying brain's functional networks. Embodiments of present disclosure provide a method of training a weighted adjacency matrix and a Graph Neural Network (GNN) to accurately represent the EEG signals. The method also trains a graph, a node, and an edge classifier to perform graph classification (i.e. motor imagery classification), node and edge classification. Thus, representations generated by the GNN can be additionally used for node and edge classification unlike state of the art methods.