Proceedings of the Pakistan Academy of Sciences: A. Physical and Computational Sciences

Advancements in Word Embeddings: A Comprehensive Survey and Analysis

Khushal Das — 2024-09-23

In recent years, the field of Natural Language Processing (NLP) has seen significant growth in the study of word representation, with word embeddings proving valuable for various NLP tasks by providing representations that encapsulate prior knowledge. We reviewed word embedding models, their applications, cross-lingual embeddings, model analyses, and techniques for model compression. We offered insights into the evolving landscape of word representations in NLP, focusing on the models and algorithms used to estimate word embeddings and their analysis strategies. To address this, we conducted a detailed examination and categorization of these evaluations and models, highlighting their significant strengths and weaknesses. We discussed a prevalent method of representing text data to capture semantics, emphasizing how different techniques can be effectively applied to interpret text data. Unlike traditional word representations, such as Word to Vector (word2vec), newer contextual embeddings, like Bidirectional Encoder Representations from Transformers (BERT) and Embeddings from Language Models (ELMo), have pushed the boundaries by capturing the use of words through diverse contexts and encoding information transfer across different languages. These embeddings leverage context to represent words, leading to innovative applications in various NLP tasks.

Spatio-Temporal Monitoring and Risk Mapping of Glacial Lake Outburst Flood in Hunza Valley, Pakistan

Nausheen Mazhar — 2024-09-23

Glacial lake outburst flood (GLOF) disasters are serious and potentially increase huge risks to livelihoods and infrastructure in the mountain regions of the world. The northern highland regions of Pakistan are home to some of the biggest alpine glaciers. In this investigation, the Hunza Valley of Pakistan has undergone remote sensing-based risk assessment for Glacial Lake Outburst Floods. Borith and Passu Lakes were chosen to identify flood risk in the downstream areas. For such, Landsat images were used from 1990-2020. Different spectral indices such as the Normalized Difference Snow Index (NDSI), Normalized Difference Glacier Index (NDGI) and Normalized Difference Water Index (NDWI) were applied to evaluate snow cover changes. Furthermore, these Lakes were digitized to evaluate and check the total increase in Lake Areas. Also, built-up areas close to lakes were digitized to identify total risk. The Land Surface Temperature (LST), NDSI, NDGI, NDWI, Digital Elevation Model (DEM) and Slope variables were given weights using the Analytic Hierarchy Process (AHP) method. In the analysis of flood risk mapping, maximum weight was assigned to Land Surface Temperature, and minimum weight was assigned to the slope. The result revealed that the settlements located in the Ghulkin, Gulmit, Husseini, Passu, Zarabad, and Khorramabad are at moderate risk while settlements located near Hunza River such as Karimabad, Khanna Abad, and Aliabad are at high risk. The outcome also showed that Borith Lake's area expanded, going from 0.059 km2 in 1985 to 0.074 km² in 2020 and Passu Lake's area also grew, going from 0.074 km² in 2005 to 0.077 km² in 2020. In last, Buffer analysis was performed to identify areas that are likely to be affected by the flood. The result of the study can help carry out a downstream risk assessment and better preparedness for future flood hazards.

Selecting the Optimal Classifier for Wrapper-Based Feature Selection Methods

Farzad Zandi — 2024-09-24

Dimensionality reduction, the elimination of irrelevant features, and the selection of an optimal subset of features are critical components in the construction of an efficacious machine learning model. Among the various feature selection methodologies, wrapper-based methods yield superior results due to their evaluation of candidate subsets. Numerous meta-heuristic methods have been employed in this feature selection process. A significant and complex issue in feature selection utilizing these methods is the selection of the most suitable classifier. In this study, we propose a novel method for selecting the optimal classifier during the feature selection process. We employ ten distinct classifiers for two swarm intelligence methods, namely Bat and Gray Wolf, and compute their results on four cancer datasets: Leukemia, SRBCT, Prostate, and Colon. Our findings demonstrate that the proposed method identifies the optimal classifiers for all four datasets. Consequently, when employing wrapper-based methods to select features for each dataset, the optimal classifier is identified.

Kinetics and Thermodynamic Study of Cellulase Embedded Metal Organic Frameworks

Kainat Zahra — 2024-09-23

Cellulase is an important enzyme used for many purposes in different industrial sectors. Even though cellulase has so many applications, it easily denatures with a little change in pH and temperature, which causes low stability, usability, and activity. To enhance its activity and stability, immobilization of porous materials is the best way to enhance its activity, stability, and life span. For immobilization, a Metal-organic framework (MOF) is considered as a potential candidate. Cellulase@Zn- benzene 1-4 di-carboxylic acid (BDC) by hydrothermal method and Zn-cellulase-benzene 1-4 dicarboxylic acid (BDC) by de novo approach were prepared, and their activities were analyzed. Zn-cellulase-benzene 1-4 dicarboxylic acid (BDC) produced by the de novo approach, shows higher activity, stability, catalytic performance, and life span than the free enzyme, and cellulase@Zn- benzene 1-4 di-carboxylic acid (BDC) produced by the hydrothermal method.

Extraction of Natural Dye using Peels of Citrus Fruits for Enhancing Color Fastness of Fabrics

Farea Noor — 2024-09-23

The current study investigates the extraction of dyestuff using citrus fruits and assesses its color, light, rubbing, and perspiration fastness. For the experimental study, three types of citrus fruits—orange, lemon, and grapefruit—were chosen to extract dye from their peels. The dye was applied to two types of fabrics (100% cotton and a blended fabric made of polyester-cotton (PC) with a ratio of 50:50) using the conventional aqueous method. The results revealed the remarkable efficacy of orange dye on both fabric types, demonstrating excellent color fastness attributes, with a minor preference for PC fabric in washing fastness. In contrast, lemon dye displayed better washing fastness properties on the tested materials as well as considerable staining potential. Grapefruit dye performed exceptionally well in terms of water and perspiration fastness. Future research could focus on improving dye extraction techniques for citrus fruits to increase color absorption and penetration. Determining various solvents, time duration for dyeing, and temperature settings could enhance the dyeing performance and effectiveness of natural dyes in different industrial applications. Further studies could also conduct life cycle assessments to measure the environmental impact of using citrus peels as a natural dye source compared to synthetic dyes in the textile industry.