Showing 1–8 of 8 results for author: Baby, B

Arabic Sentiment Analysis with Noisy Deep Explainable Model

Authors: Md. Atabuzzaman, Md Shajalal, Maksuda Bilkis Baby, Alexander Boden

Abstract: Sentiment Analysis (SA) is an indispensable task for many real-world applications. Compared to limited resourced languages (i.e., Arabic, Bengali), most of the research on SA are conducted for high resourced languages (i.e., English, Chinese). Moreover, the reasons behind any prediction of the Arabic sentiment analysis methods exploiting advanced artificial intelligence (AI)-based approaches are l… ▽ More Sentiment Analysis (SA) is an indispensable task for many real-world applications. Compared to limited resourced languages (i.e., Arabic, Bengali), most of the research on SA are conducted for high resourced languages (i.e., English, Chinese). Moreover, the reasons behind any prediction of the Arabic sentiment analysis methods exploiting advanced artificial intelligence (AI)-based approaches are like black-box - quite difficult to understand. This paper proposes an explainable sentiment classification framework for the Arabic language by introducing a noise layer on Bi-Directional Long Short-Term Memory (BiLSTM) and Convolutional Neural Networks (CNN)-BiLSTM models that overcome over-fitting problem. The proposed framework can explain specific predictions by training a local surrogate explainable model to understand why a particular sentiment (positive or negative) is being predicted. We carried out experiments on public benchmark Arabic SA datasets. The results concluded that adding noise layers improves the performance in sentiment analysis for the Arabic language by reducing overfitting and our method outperformed some known state-of-the-art methods. In addition, the introduced explainability with noise layer could make the model more transparent and accountable and hence help adopting AI-enabled system in practice. △ Less

Submitted 29 November, 2023; v1 submitted 24 September, 2023; originally announced September 2023.

Comments: This is the pre-print version of our accepted paper at the 7th International Conference on Natural Language Processing and Information Retrieval~(ACM NLPIR'2023)

Journal ref: NLPIR 2023: 2023 7th International Conference on Natural Language Processing and Information Retrieval, Seoul, Republic of Korea, December 2023

From Forks to Forceps: A New Framework for Instance Segmentation of Surgical Instruments

Authors: Britty Baby, Daksh Thapar, Mustafa Chasmai, Tamajit Banerjee, Kunal Dargan, Ashish Suri, Subhashis Banerjee, Chetan Arora

Abstract: Minimally invasive surgeries and related applications demand surgical tool classification and segmentation at the instance level. Surgical tools are similar in appearance and are long, thin, and handled at an angle. The fine-tuning of state-of-the-art (SOTA) instance segmentation models trained on natural images for instrument segmentation has difficulty discriminating instrument classes. Our rese… ▽ More Minimally invasive surgeries and related applications demand surgical tool classification and segmentation at the instance level. Surgical tools are similar in appearance and are long, thin, and handled at an angle. The fine-tuning of state-of-the-art (SOTA) instance segmentation models trained on natural images for instrument segmentation has difficulty discriminating instrument classes. Our research demonstrates that while the bounding box and segmentation mask are often accurate, the classification head mis-classifies the class label of the surgical instrument. We present a new neural network framework that adds a classification module as a new stage to existing instance segmentation models. This module specializes in improving the classification of instrument masks generated by the existing model. The module comprises multi-scale mask attention, which attends to the instrument region and masks the distracting background features. We propose training our classifier module using metric learning with arc loss to handle low inter-class variance of surgical instruments. We conduct exhaustive experiments on the benchmark datasets EndoVis2017 and EndoVis2018. We demonstrate that our method outperforms all (more than 18) SOTA methods compared with, and improves the SOTA performance by at least 12 points (20%) on the EndoVis2017 benchmark challenge and generalizes effectively across the datasets. △ Less

Submitted 11 March, 2023; v1 submitted 26 November, 2022; originally announced November 2022.

Comments: WACV 2023

Textual Entailment Recognition with Semantic Features from Empirical Text Representation

Authors: Md Shajalal, Md Atabuzzaman, Maksuda Bilkis Baby, Md Rezaul Karim, Alexander Boden

Abstract: Textual entailment recognition is one of the basic natural language understanding(NLU) tasks. Understanding the meaning of sentences is a prerequisite before applying any natural language processing(NLP) techniques to automatically recognize the textual entailment. A text entails a hypothesis if and only if the true value of the hypothesis follows the text. Classical approaches generally utilize t… ▽ More Textual entailment recognition is one of the basic natural language understanding(NLU) tasks. Understanding the meaning of sentences is a prerequisite before applying any natural language processing(NLP) techniques to automatically recognize the textual entailment. A text entails a hypothesis if and only if the true value of the hypothesis follows the text. Classical approaches generally utilize the feature value of each word from word embedding to represent the sentences. In this paper, we propose a novel approach to identifying the textual entailment relationship between text and hypothesis, thereby introducing a new semantic feature focusing on empirical threshold-based semantic text representation. We employ an element-wise Manhattan distance vector-based feature that can identify the semantic entailment relationship between the text-hypothesis pair. We carried out several experiments on a benchmark entailment classification(SICK-RTE) dataset. We train several machine learning(ML) algorithms applying both semantic and lexical features to classify the text-hypothesis pair as entailment, neutral, or contradiction. Our empirical sentence representation technique enriches the semantic information of the texts and hypotheses found to be more efficient than the classical ones. In the end, our approach significantly outperforms known methods in understanding the meaning of the sentences for the textual entailment classification task. △ Less

Submitted 19 June, 2023; v1 submitted 18 October, 2022; originally announced October 2022.

Comments: Pre-print for our paper at International Conference on Speech & Language Technology for Low-resource Languages (SPELLL'2022)

Journal ref: Speech and Language Technologies for Low-Resource Languages . SPELLL 2022. Communications in Computer and Information Science, vol 1802. Springer, Cham

Swift J1753.5-0127 : Understanding the accretion geometry through frequency resolved spectroscopy

Authors: Blessy E. Baby, Ramadevi M. C

Abstract: The Black Hole Binary source Swift J175.5-0127 remained in outburst for $\sim$ 12 years from May 2005 to April 2017. For most part of the outburst, the source remained in the Low Hard State (LHS) displaying transitions to softer states only towards the end of the outburst for short periods of time. Quasi periodic Oscillations (QPOs) were observed in the Power Density Spectrum (PDS) only during the… ▽ More The Black Hole Binary source Swift J175.5-0127 remained in outburst for $\sim$ 12 years from May 2005 to April 2017. For most part of the outburst, the source remained in the Low Hard State (LHS) displaying transitions to softer states only towards the end of the outburst for short periods of time. Quasi periodic Oscillations (QPOs) were observed in the Power Density Spectrum (PDS) only during the decay. A soft thermal component was required to model the spectrum in LHS, which does not conform to the generally accepted disc truncation theory. In this work, we attempt to obtain a clearer picture of the accretion disc geometry by studying the QPO variability using frequency resolved spectroscopy (FRS). We obtain the QPO rms spectrum of the source during the bright-hard state and model it with physical components. We find that the QPO rms spectrum can be described only by a Comptonisation component with no contribution from the thermal disc. This indicates that the variability observed in the PDS originates in the Comptonisation component and the evolution of the QPOs is likely to be a result of localization of the variabilities to different radii of the hot inner flow rather than disc truncation. The minimal variation in disc parameters also points to the existence of a stable disc throughout the outburst. △ Less

Submitted 22 December, 2021; originally announced December 2021.

Comments: 9 pages; 6 figures; Accepted for publication in JoAA

Revealing the nature of the transient source MAXI J0637-430 through spectro-temporal analysis

Authors: Blessy E. Baby, Bhuvana G. R., Radhika D., Tilak Katoch, Samir Mandal, Anuj Nandi

Abstract: We study the spectral and temporal properties of MAXI J0637-430 during its 2019-2020 outburst using \textit{NICER}, \textit{AstroSat} and \textit{Swift-XRT} data. The source was in a disc dominant state within a day of its detection and traces out a `c' shaped profile in the HID, similar to the `mini'-outbursts of the recurrent BHB 4U 1630-472. Energy spectrum is obtained in the $0.5-10$ keV band… ▽ More We study the spectral and temporal properties of MAXI J0637-430 during its 2019-2020 outburst using \textit{NICER}, \textit{AstroSat} and \textit{Swift-XRT} data. The source was in a disc dominant state within a day of its detection and traces out a `c' shaped profile in the HID, similar to the `mini'-outbursts of the recurrent BHB 4U 1630-472. Energy spectrum is obtained in the $0.5-10$ keV band with \textit{NICER} and \textit{Swift-XRT}, and $0.5-25$ keV with \textit{AstroSat}. The spectra can be modelled using a multicolour disc emission (\textit{diskbb}) convolved with a thermal Comptonisation component (\textit{thcomp}). The disc temperature decreases from 0.6 keV to 0.1 keV during the decay with a corresponding decrease in photon index ($Γ$) from 4.6 to 1.8. The fraction of Compton scattered photons ($f_{cov}$) remains $<$ 0.3 during the decay upto mid-January 2020 and gradually increases to 1 as the source reaches hard state. Power Density Spectra (PDS) generated in the 0.01-100 Hz range display no Quasi-periodic Oscillations (QPOs) although band-limited noise (BLN) is seen towards the end of January 2020. During \textit{AstroSat} observations, $Γ$ lies in the range $2.3-2.6$ and rms increases from 11 to 20\%, suggesting that the source was in an intermediate state till 21 November 2019. Spectral fitting with the relativistic disc model (\textit{kerrbb}), in conjunction with the soft-hard transition luminosity, favour a black hole with mass $3-19$ $M_{\odot}$ with retrograde spin at a distance $<15$ kpc. Finally, we discuss the possible implications of our findings. △ Less

Submitted 21 October, 2021; v1 submitted 17 September, 2021; originally announced September 2021.

Comments: References updated, final version

AstroSat view of IGR J17091-3624 and GRS 1915+105: decoding the 'pulse' in the 'Heartbeat State'

Authors: Tilak Katoch, Blessy E. Baby, Anuj Nandi, V. K. Agrawal, H. M. Antia, Kallol Mukerjee

Abstract: IGR J17091--3624 is a transient galactic black hole which has a distinct quasi-periodic variability known as `heartbeat', similar to the one observed in GRS 1915+105. In this paper, we report the results of $\sim 125$ ks \textit{AstroSat} observations of this source during the 2016 outburst. For the first time a double peaked QPO (DPQ) is detected in a few time segments of this source with a diffe… ▽ More IGR J17091--3624 is a transient galactic black hole which has a distinct quasi-periodic variability known as `heartbeat', similar to the one observed in GRS 1915+105. In this paper, we report the results of $\sim 125$ ks \textit{AstroSat} observations of this source during the 2016 outburst. For the first time a double peaked QPO (DPQ) is detected in a few time segments of this source with a difference of $δf ~\sim12$ mHz between the two peaks. The nature of the DPQ was studied based on hardness ratios and using the static as well as the dynamic power spectrum. Additionally, a low frequency (25--48 mHz) `heartbeat' single peak QPO (SPQ) was observed at different intervals of time along with harmonics ($50-95$ mHz). Broadband spectra in the range $0.7-23$ keV, obtained with \textit{SXT} and \textit{LAXPC}, could be fitted well with combination of a thermal Comptonisation and a multicolour disc component model. During \textit{AstroSat} observation, the source was in the Soft-Intermediate State (SIMS) as observed with \textit{Swift/XRT}. We present a comparative study of the `heartbeat' state variability in IGR J17091--3624 with GRS 1915+105. Significant difference in the timing properties is observed although spectral parameters ($Γ\sim2.1-2.4$ and $T_\mathrm{max}\sim0.6-0.8$ keV) in the broad energy band remain similar. Spectral properties of segments exhibiting SPQ and DPQ are further studied using simple phase resolved spectroscopy which does not show a significant difference. Based on the model parameters, we obtain the maximum ratio of mass accretion rate in GRS 1915+105 to that in IGR J17091--3624 as $\sim25:1$. We discuss the implications of our findings and comment on the physical origin of these exotic variabilities. △ Less

Submitted 26 November, 2020; originally announced November 2020.

Comments: 17 pages, 13 Figures, 6 Tables; Accepted for publication in MNRAS

Neuro-Endo-Trainer-Online Assessment System (NET-OAS) for Neuro-Endoscopic Skills Training

Authors: Vinkle Srivastav, Britty Baby, Ramandeep Singh, Prem Kalra, Ashish Suri

Abstract: Neuro-endoscopy is a challenging minimally invasive neurosurgery that requires surgical skills to be acquired using training methods different from the existing apprenticeship model. There are various training systems developed for imparting fundamental technical skills in laparoscopy where as limited systems for neuro-endoscopy. Neuro-Endo-Trainer was a box-trainer developed for endo-nasal transs… ▽ More Neuro-endoscopy is a challenging minimally invasive neurosurgery that requires surgical skills to be acquired using training methods different from the existing apprenticeship model. There are various training systems developed for imparting fundamental technical skills in laparoscopy where as limited systems for neuro-endoscopy. Neuro-Endo-Trainer was a box-trainer developed for endo-nasal transsphenoidal surgical skills training with video based offline evaluation system. The objective of the current study was to develop a modified version (Neuro-Endo-Trainer-Online Assessment System (NET-OAS)) by providing a stand-alone system with online evaluation and real-time feedback. The validation study on a group of 15 novice participants shows the improvement in the technical skills for handling the neuro-endoscope and the tool while performing pick and place activity. △ Less

Submitted 16 July, 2020; originally announced July 2020.

Comments: Published at Federated Conference on Computer Science and Information Systems - FedCSIS 2017

Journal ref: IEEE (2017)

AstroSat and MAXI view of the Black Hole binary 4U 1630-472 during 2016 and 2018 Outbursts

Authors: Blessy E. Baby, V. K. Agrawal, Ramadevi M. C., Tilak Katoch, H. M. Antia, Samir Mandal, Anuj Nandi

Abstract: We present an in-depth spectral and timing analysis of the Black Hole binary 4U 1630-472 during 2016 and 2018 outbursts as observed by \textit{AstroSat} and \textit{MAXI}. The extensive coverage of the outbursts with \textit{MAXI} is used to obtain the Hardness Intensity Diagram (HID). The source follows a `c'-shaped profile in agreement with earlier findings. Based on the HIDs of previous outburs… ▽ More We present an in-depth spectral and timing analysis of the Black Hole binary 4U 1630-472 during 2016 and 2018 outbursts as observed by \textit{AstroSat} and \textit{MAXI}. The extensive coverage of the outbursts with \textit{MAXI} is used to obtain the Hardness Intensity Diagram (HID). The source follows a `c'-shaped profile in agreement with earlier findings. Based on the HIDs of previous outbursts, we attempt to track the evolution of the source during a `super'-outburst and `mini'-outbursts. We model the broadband energy spectra ($0.7-20.0$ keV) of \textit{AstroSat} observations of both outbursts using phenomenological and physical models. No Keplerian disc signature is observed at the beginning of 2016 outburst. However, the disc appears within a few hours after which it remains prominent with temperature ($T_{in}$) $\sim$ 1.3 keV and increase in photon index ($Γ$) from 1.8 to 2.0, whereas the source was at a disc dominant state throughout the \textit{AstroSat} campaign of 2018 outburst. Based on the HIDs and spectral properties, we classify the outbursts into three different states - the `canonical' hard and soft states along with an intermediate state. Evolution of rms along different states is seen although no Quasi-periodic Oscillations (QPOs) are detected. We fit the observed spectra using a dynamical accretion model and estimate the accretion parameters. Mass of the black hole is estimated using inner disc radius, bolometric luminosity and two component flow model to be $3-9$ $M_{\odot}$. Finally, we discuss the possible implications of our findings. △ Less

Submitted 8 July, 2020; v1 submitted 2 July, 2020; originally announced July 2020.

Comments: 16 pages, 11 figures, Accepted for publication in MNRAS, typos fixed and figures updated