-
FusionMind -- Improving question and answering with external context fusion
Authors:
Shreyas Verma,
Manoj Parmar,
Palash Choudhary,
Sanchita Porwal
Abstract:
Answering questions using pre-trained language models (LMs) and knowledge graphs (KGs) presents challenges in identifying relevant knowledge and performing joint reasoning.We compared LMs (fine-tuned for the task) with the previously published QAGNN method for the Question-answering (QA) objective and further measured the impact of additional factual context on the QAGNN performance. The QAGNN met…
▽ More
Answering questions using pre-trained language models (LMs) and knowledge graphs (KGs) presents challenges in identifying relevant knowledge and performing joint reasoning.We compared LMs (fine-tuned for the task) with the previously published QAGNN method for the Question-answering (QA) objective and further measured the impact of additional factual context on the QAGNN performance. The QAGNN method employs LMs to encode QA context and estimate KG node importance, and effectively update the question choice entity representations using Graph Neural Networks (GNNs). We further experimented with enhancing the QA context encoding by incorporating relevant knowledge facts for the question stem. The models are trained on the OpenbookQA dataset, which contains ~6000 4-way multiple choice questions and is widely used as a benchmark for QA tasks. Through our experimentation, we found that incorporating knowledge facts context led to a significant improvement in performance. In contrast, the addition of knowledge graphs to language models resulted in only a modest increase. This suggests that the integration of contextual knowledge facts may be more impactful for enhancing question answering performance compared to solely adding knowledge graphs.
△ Less
Submitted 30 December, 2023;
originally announced January 2024.
-
Transformer Based Implementation for Automatic Book Summarization
Authors:
Siddhant Porwal,
Laxmi Bewoor,
Vivek Deshpande
Abstract:
Document Summarization is the procedure of generating a meaningful and concise summary of a given document with the inclusion of relevant and topic-important points. There are two approaches: one is picking up the most relevant statements from the document itself and adding it to the Summary known as Extractive and the other is generating sentences for the Summary known as Abstractive Summarizatio…
▽ More
Document Summarization is the procedure of generating a meaningful and concise summary of a given document with the inclusion of relevant and topic-important points. There are two approaches: one is picking up the most relevant statements from the document itself and adding it to the Summary known as Extractive and the other is generating sentences for the Summary known as Abstractive Summarization. Training a machine learning model to perform tasks that are time-consuming or very difficult for humans to evaluate is a major challenge. Book Abstract generation is one of such complex tasks. Traditional machine learning models are getting modified with pre-trained transformers. Transformer based language models trained in a self-supervised fashion are gaining a lot of attention; when fine-tuned for Natural Language Processing(NLP) downstream task like text summarization. This work is an attempt to use Transformer based techniques for Abstract generation.
△ Less
Submitted 17 January, 2023;
originally announced January 2023.
-
Quantum approach on linear combination of harmonic univalent mappings
Authors:
Omendra Mishra,
Saurabh Porwal
Abstract:
In this paper using $q$ calculus operator we obtain some sufficient conditions on $f_1$ and $f_2$ so that their linear combination $% f=tf_{1}+(1-t)f_{2},\ t\in \left[ 0,1\right] $, is univalent and convex in the direction of the real axis. Some examples are also illustrated to support our main results.
In this paper using $q$ calculus operator we obtain some sufficient conditions on $f_1$ and $f_2$ so that their linear combination $% f=tf_{1}+(1-t)f_{2},\ t\in \left[ 0,1\right] $, is univalent and convex in the direction of the real axis. Some examples are also illustrated to support our main results.
△ Less
Submitted 10 August, 2021;
originally announced August 2021.
-
Univalent functions with positive coefficients involving Touchard polynomials
Authors:
G. Murugusundaramoorthy,
Saurabh Porwal
Abstract:
The purpose of the present paper is to establish connections between various subclasses of analytic univalent functions by applying certain convolution operator involving Touchard polynomials. To be more precise, we investigate such connections with the classes of analytic univalent functions with positive coefficients in the open unit disk U.
The purpose of the present paper is to establish connections between various subclasses of analytic univalent functions by applying certain convolution operator involving Touchard polynomials. To be more precise, we investigate such connections with the classes of analytic univalent functions with positive coefficients in the open unit disk U.
△ Less
Submitted 10 July, 2020;
originally announced July 2020.
-
Fabrication of plasmonic surface relief gratings for the application of band-pass filter in UV-Visible spectral range
Authors:
Sudheer,
S. Porwal,
S. Bhartiya,
C. Mukherjee,
P. Tiwari,
T. K. Sharma,
V. N. Rai,
A. K. Srivastava
Abstract:
The measured experimental results of optical diffraction of 10, 5 and 3.4 micrometer period plasmonic surface relief grating are presented for the application of band-pass filter in visible spectral range. Conventional scanning electron microscopic (SEM) is used to fabricate the grating structures on the silver halide based film (substrate) by exposing the electron beam in raster scan fashion. Mor…
▽ More
The measured experimental results of optical diffraction of 10, 5 and 3.4 micrometer period plasmonic surface relief grating are presented for the application of band-pass filter in visible spectral range. Conventional scanning electron microscopic (SEM) is used to fabricate the grating structures on the silver halide based film (substrate) by exposing the electron beam in raster scan fashion. Morphological characterization of the gratings is performed by atomic force microscopy (AFM) shows that the period, height and profile depends on the line per frame, beam spot, single line dwell time, beam current, and accelerating voltage of the electron beam. Optical transmission spectra of 10 micrometer period grating shows a well-defined localized surface plasmon resonance (LSPR) dip at ~366 nm wavelength corresponding to gelatin embedded silver nanoparticles of the grating structure. As the period of the grating reduces LSPR dip becomes prominent. The maximum first order diffraction efficiency (DE) and bandwidth for 10 micrometer period grating are observed as 4% and 400 nm in 350 nm to 800 nm wavelength range respectively. The DE and bandwidth are reduced up to 0.03% and 100 nm for 3.4 micrometer period grating. The profile of DE is significantly flat within the diffraction bandwidth for each of the gratings. An assessment of the particular role of LSPR absorption and varied grating period in the development of the profile of first order DE v/s wavelength are studied. Fabrication of such nano-scale structures in a large area using conventional SEM and silver halide based films may provide the simple and efficient technique for various optical devices applications.
△ Less
Submitted 4 July, 2017;
originally announced August 2017.
-
Development of high power quantum well lasers at RRCAT
Authors:
T. K. Sharma,
Tapas Ganguli,
V. K. Dixit,
S. D. Singh,
S. Pal,
S. Porwal,
Ravi Kumar,
Alexander Khakha,
R. Jangir,
V. Kheraj,
P. Rawat,
A. K. Nath
Abstract:
We at RRCAT have recently developed high power laser diodes in the wavelength range of 740 to 1000 nm. A typical semiconductor laser structure is consisted of about 10 epilayers with different composition, thickness and doping values. For example, a laser diode operating at 0.8 micron has either GaAs or GaAsP quantum well as an active layer. The quantum well is sandwiched between AlGaAs wider band…
▽ More
We at RRCAT have recently developed high power laser diodes in the wavelength range of 740 to 1000 nm. A typical semiconductor laser structure is consisted of about 10 epilayers with different composition, thickness and doping values. For example, a laser diode operating at 0.8 micron has either GaAs or GaAsP quantum well as an active layer. The quantum well is sandwiched between AlGaAs wider bandgap waveguide and cladding layers. The complete laser structure is grown by metal organic vapour phase epitaxy technique and devices are fabricated through standard procedure using photolithography. We recently achieved about 5.3 Watt peak power at 853 nm. These laser diodes were tested under pulsed operation at room temperature for 500 nanosecond pulse duration with a duty cycle of 1:1000. Laser diode arrays consisting of 6-10 elements were also developed and tested for operation in pulsed mode at room temperature.
△ Less
Submitted 4 December, 2014;
originally announced December 2014.