Shaily Gandhi

Dr. Shaily Gandhi


Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
Verified
Journal Article
Open access
PDF file
CZChristina Zorenböhmer
Christina Zorenböhmer
Shaily GandhiShaily Gandhi
Shaily Gandhi
SSSebastian Schmidt
Sebastian Schmidt
Bernd ReschBernd Resch
Bernd Resch
Natural disasters like wildfires pose significant threats to communities, which necessitates timely and effective disaster response strategies. While Aspect-based Sentiment Analysis (ABSA) has been widely used to extract sentiment-related information at the sub-sentence level, the corresponding field of Aspect-based Emotion Analysis (ABEA) remains underexplored due to dataset limitations and the increased complexity of emotion classification. In this study, we used EmoGRACE, a fine-tuned BERT-based model for ABEA, which we applied to georeferenced tweets of the 2020 California wildfires. The results for this case study reveal distinct spatio-temporal emotion patterns for wildfire-related aspect terms, with fear and sadness increasing near wildfire perimeters. This study demonstrates the feasibility of tracking emotion dynamics across disaster-affected regions and highlights the potential of ABEA in real-time disaster monitoring. The results suggest that ABEA can provide a nuanced understanding of public sentiment during crises for policymakers.

11 May 2016 in International Journal of Built Environment and Sustainability
Verified
Journal Article
Open access
PDF file
Shaily GandhiShaily Gandhi
Shaily Gandhi
SSShashikant A Sharma
Shashikant A Sharma
AVAnjana Vyas
Anjana Vyas
Urban sprawl is one of the avidly urban issues today. “Sprawling” means the spreading of urban area towards rural area surrounding it. There are different social and economic factors which led to urban sprawl. The study of urban sprawl in this paper attempts to measure the sprawl of urban areas into rural areas and the type of sprawl and the pre-dominant causes responsible for sprawl. It first reviews and analyses past research on the definitions of urban form, compactness and sprawl, and corresponding quantitative variables. Measures of urban sprawl will be identified and procedures will be developed to compute these indicators in GIS environment. The indicators of sprawl defined by (Galster et al, 2001) are considered as major characteristics of sprawl which have been adopted in this paper to develop a set of variables for quantification, characterization and dimension of sprawl.The study has been done on grid level (Grid Size = 500m x 500 m). The density index shows that in 2001 Rajkot had 9292 persons per grid of 500 meters x 500 meters which has decreased to 7394 persons per grid which shows that the population is moving from center to outer fringes. The value of concentration index has decreased from 0.56 in 2001 to 0.23 in 2011 which shows the sprawl development pattern. Centrality has been one of the major indicators of urban sprawl the value of centrality has been observed as 3.7 in 2001 and it has increased to 7.6 in 2011 which indicates sprawl.

2 July 2025
Verified
Conference Paper
Closed access
DHDavid Hanny
David Hanny
SSSebastian Schmidt
Sebastian Schmidt
Shaily GandhiShaily Gandhi
Shaily Gandhi
Bernd ReschBernd Resch
Bernd Resch
Social media can provide real-time insights during natural disasters, yet efficiently identifying relevant content remains a challenge due to the reliance on large labelled datasets and high computational costs. This study therefore investigates the potential of Few-Shot Learning (FSL) for relevance classification of textual social media posts during disasters. We compare few-shot prompting using eight Small Language Models (SLMs) and a contrastive learning approach (SetFit) with data from five disasters across the world: the 2020 California wildfires, 2021 Ahr Valley floods, 2023 Chile wildfires, 2023 Emilia-Romagna floods, and 2023 Turkey/Syria earthquake. GPT-4o-mini achieves the highest average macro F1 score (0.77) using just five labelled examples per class, while the multilingual-e5-base model fine-tuned with SetFit offers a strong alternative (avg. macro F1 = 0.65) without reliance on prompt engineering. Our findings highlight the potential of SLMs and FSL for scalable and resource-efficient data analytics in disaster management and broader social science research.

11 February 2026 in Urban Informatics
Verified
Journal Article
Open access
PDF file
MMMartin Karl Moser
Martin Karl Moser
DGDavid Ruben Max Graf
David Ruben Max Graf
Shaily GandhiShaily Gandhi
Shaily Gandhi
Bernd ReschBernd Resch
Bernd Resch
The spatiotemporal nature of urban environments — encompassing both, interactions among stationary features of the built and natural surroundings, and dynamic elements such as road users — pose significant challenges to the adoption of active mobility. Previously, influences of the urban environment on active mobility users have been typically investigated in unimodal approaches, thus greatly neglecting the diverse range of urban stressors that adversely affect (sustainable) mobility experiences, which can be inferred from multimodal data. For this reason, we propose a multimodal approach to investigate urban stress factors. Our methodology integrates data from wearable sensors and visual urban media to gain a more comprehensive understanding of spatiotemporal stressors in urban environments and active mobility. Spatially clustered stress measurements, i.e., hotspots and coldspots, derived from physiological reactions of the body, are used as labels for the classification of high-stress and low-stress urban areas. Semantic segmentation-based visual features, describing the immediate surroundings, are derived from real-time videos and snapshots of street scenes, captured through Street View Imagery (SVI). By comparing isovist features, i.e., visual impressions of dynamically changing urban scenes from a cyclists’ point of view (POV), with scenes captured through SVI, we show that SVI provides a valuable data source for urban visual intelligence and relating high-stress cycling experiences to the surrounding environmental characteristics. While our Random Forest (RF) model trained on SVI-based features outperformed a POV video-based model by 3.9 percentage points in accuracy and 5.3 percentage points in recall (accuracy: 72.6% vs. 68.7%, recall: 72.2% vs. 66.9%), we encourage future studies to validate our findings in regions with higher environmental diversity, lower coverage of SVI data, and with additional data sources to account for confounding factors.

13 May 2022 in Journal of Education, Society and Behavioural Science
Verified
Journal Article
Closed access
PDF file
Shaily GandhiShaily Gandhi
Shaily Gandhi
Felix Emeka AnyiamFelix Emeka Anyiam
Felix Emeka Anyiam
This paper explains the benefits of Urban Data Science Education in improving data driving policy-making and highlights the impact of vital use of available open data using a practical demonstration from the Summer-Winter School (SWS) at CEPT University Ahmedabad India. Good Urban data policies can be an important driver for smart city research and the implementation of good data management practices. Many authors are also becoming interested in the data revolution that is enhancing the way we study and understand cities. It now becomes appropriate to respond to this need in an evidence-informed manner by working with stakeholders to encourage a more resilient approach via education for the younger enthusiast, leading to more transparent, improved, sustainable and safer urban cities. However, the major challenge we foresee is the inability to utilize the large amount of urban data generated daily. With the drive to mobilize more education towards urban data science and analytics, generated data can be effectively utilized to make informed and evidence-based decisions. This paper points to a practical methodology for reproducibility, the key steps taken, as it views the inclusion of an Urban Data Science curriculum as priceless, and a decision that will bring into notice the importance of engaging the younger generation in scholastic learning on data principles and analytics. Data along with spatial enhancement can be the driving factor towards improving data-driven policy-making for solving urban problems in smart cities.