Indian Institute of Technology Mandi

Hate speech detection on social media faces challenges in both accuracy and explainability, especially for underexplored Indic languages. We propose a novel explainability-guided training framework, X-MuTeST (eXplainable Multilingual haTe Speech deTection), for hate speech detection that combines high-level semantic reasoning from large language models (LLMs) with traditional attention-enhancing techniques. We extend this research to Hindi and Telugu alongside English by providing benchmark human-annotated rationales for each word to justify the assigned class label. The X-MuTeST explainability method computes the difference between the prediction probabilities of the original text and those of unigrams, bigrams, and trigrams. Final explanations are computed as the union between LLM explanations and X-MuTeST explanations. We show that leveraging human rationales during training enhances both classification performance and the model’s explainability. Moreover, combining human rationales with our explainability method to refine the model’s attention yields further improvements. We evaluate explainability using Plausibility metrics such as Token-F1 and IOU-F1, and Faithfulness metrics such as Comprehensiveness and Sufficiency. By focusing on under-resourced languages, our work advances hate speech detection across diverse linguistic contexts. Our dataset includes token-level rationale annotations for 6,004 Hindi, 4,492 Telugu, and 6,334 English samples.

This study demonstrates the potential of cost-effective sonar integrated with uncrewed surface vehicles (USVs) for bathymetric surveys of high-altitude mountain lakes in the northwestern Himalaya, India. Three lakes were selected as test sites, Deepak Tal (3,770 m), Suraj Tal (4,777 m) near Baralacha La in Lahaul & Spiti, Himachal Pradesh, and a smaller lake near Parashar Lake (hereafter referred to as NP Lake; 2,592 m) in Mandi, Himachal Pradesh, India. The surveys were conducted using a Deeper Pro+ sonar mounted on a USV. The sonar recorded depth measurements at individual points, which were later interpolated using Kriging to generate continuous bathymetric surfaces for each lake.For Deepak Tal, over 25,000 depth points were collected, yielding a maximum depth of 7.44 m and an estimated lake volume of 41,929.05 m³. For Suraj Tal, more than 5,000 points were acquired, with a maximum depth of 2.56 m and a corresponding volume of 9,400.95 m³. For NP Lake, approximately 1,000 points were recorded, with a maximum depth of ~1.3 m and a volume of 2,947.53 m³. Several empirical formulas were applied to calculate volume of the lakes and compared with the field-derived volumes, revealing extremely large deviations (up to ~129,908 times), highlighting the limitations of the empirical formulas used for volume estimation of high-altitude lakes.Overall, this study demonstrates that the Deeper Pro+ sonar mounted on a USV provides an effective, low-cost, and low-risk approach for generating high-resolution bathymetry, accurate volume estimates, and detailed morphometric information for high-mountain lakes.Keywords: Cost-Effective Bathymetry; High-Altitude Lakes; USV; Sonar Mapping; Lake Volume; Northwestern Himalaya

Seasonal snowmelt strongly influences hydrological and ecological processes by controlling the timing of peak soil moisture and subsequent vegetation growth; yet, this relationship is less studied in the Indian Himalayan region, especially in Himachal Pradesh’s snow region. This study investigates the linkage between snow disappearance timing and peak soil moisture using station data across various elevation ranges from 1571 m to 3325 m and the FLDAS dataset (daily, 0.01° resolution), and then their effect on vegetation growth using the MODIS NDVI product (8-day, 250-m resolution). We quantified the spatial and temporal variability in snow recession by fitting an exponential decay model. The recession rate varies between 0.03 and 0.27 across various elevation ranges and temporal periods. The recession rate also exhibits a strong elevation dependency, being low at higher elevations (e.g., Lari, 3325 m m.s.l., k = 0.044 m/day) and high at lower elevations (e.g., Dodra Kawar, 2522 m m.s.l., k = 0.169 m/day). Based on analysis from 2001 to 2020 across nine stations, results show that snow onset occurs in mid-December, followed by snow recession in late February and complete snow disappearance by late March across Himachal Pradesh. We observed an average lag of 3-4 days between the timing of peak soil moisture and snow disappearance, and a correlation of 0.94 (p < 0.05) was observed across various stations. Early melt does contribute to greening, as evidenced by the weaker but still positive correlation (0.54, p < 0.05) between the timing of snow disappearance and the rise in NDVI. The results show that the timing of snowmelt primarily influences soil moisture dynamics and controls vegetation activity in Himalayan catchments.

The intensification of extreme rainfall has resulted in widespread landslide hazards in mountainous regions of the world. The Indian Himalayan Region, one of the most densely urbanized, has been facing an alarming increase in landslides, the prediction of which is difficult using existing empirical rainfall thresholds. This study develops a novel machine learning-driven landslide nowcasting system by integrating the landslide susceptibility (LSM) and probability of rainfall-induced landslides (P-RIL). The LSM provides the spatial location of future landslides by analyzing the terrain characteristics, anthropogenic factors, hydrological presence, and geological formations using the random forest (RF) method based on landslides occurring between 2017-2024. The results indicated that 7% of the area was under high susceptibility, followed by 12% under high susceptibility. To calculate the effect of rainfall in triggering landslides, the P-RIL was calculated considering R1 (rainfall on 1st day of occurrence), R3 (rainfall on 3rd day), R7 (7th day rainfall), R15 (15th day rainfall), Wetdays, Max_72 Hours, and antecedent rainfall index (ARI) as variables to train in the RF model. Finally, each day nowcasting results were obtained by integrating the LSM and P-RIL within a probabilistic framework. The landslide occurring in 2025 was used to validate the nowcasting results. The results indicated that the landslides were ranked within the forecasted hazard distribution, with percentile values of 87%, 90%, 93%, and 99%, respectively, denoting the occurrence of landslides within the top 13%–1% of the most hazardous slope units at the time of prediction. One event lay in the extreme hazard class (>99th percentile), highlighting the model’s strong discriminatory capability. Finally, the forecast results for each day were updated in a Google Earth Engine application to aid policymakers and planners in developing better mitigation and preparedness strategies. This study represents the first of its kind landslide nowcasting system in Mandi district using the information obtained from landslide susceptibility and rainfall-derived triggering parameters, thus offering meaningful insight into a practical decision-support tool for policymakers and disaster management authorities.

The Himalayas are critical geographical regions, recognized for their remarkable beauty; however, plastic littering in the Himalayas is increasing exponentially due to ignorance in every matrix. Due to long-term nescience, the degradation of microplastics has been observed and found ubiquitous. Microplastics (MPs) belong to plastic particles less than 5 mm in size. The MPs are one of the critical environmental contaminant reported by various studies, such as oceans, rivers, lakes, and estuaries. Yet, their distribution in western Himalayan river systems are poorly understood. To understand this knowledge gap in research, this study provides a brief quantification, characterization, and its fate in selected western Himalayan rivers: Beas river, Parvati river, Uhl river, and Suketi river, originating from high altitude. Sediment samples were collected from 25 locations, while sampling, field images were taken to understand the source of contamination. The established protocol was performed for pre-treatment process involving sediment sieving: coarse sand (4.75 mm-2.36 mm), medium sand (2.36 mm-0.3 mm), fine sand (0.3mm-0.075mm), siltyclay ( fine sand> medium sand, including remote locations. All the samples resulted transparent MPs within the size range of 10-20 µm mainly belonging to PVDF (polyvinylidene fluoride), followed by PEG (polyethylene glycol), PE (polyester), and others prevailing types of MPs. The MPs was notably higher in siltyclay sediment fraction, which are easily transported from higher altitude to lower altitude. This study offers novel insights into the fate of MPs in fragile mountain ecosystems and emphasizes the role of sediments as an important reservoir influencing pollutant transport.Keywords- Himalayas, Microplastics, Sediment fraction, Raman Spectroscopy, Fe-SEM

RNA-binding protein (RBP)–RNA interactions are fundamental for gene regulation and cellular homeostasis. Ataxin-2 is an RBP that has been shown to play an instrumental role in pathophysiological processes by binding to mRNA. Methods such as RNA immunoprecipitation (RIP), cross-linking immunoprecipitation (CLIP), and their variants can be used to study the interactions between Ataxin-2 and its targets, although their high sample requirements and labor-intensive workflows can limit their widespread use. RNA editing-based approaches, such as targets of RBPs identified by editing (TRIBE), provide effective alternatives. TRIBE enables transcriptome-wide identification of RBP targets by inducing site-specific adenosine-to-inosine (A-to-I) editing, which is subsequently detected through high-throughput RNA sequencing in both in vivo and in vitro systems. Compared to in vivo models, cell lines offer a rapid and flexible experimental design. Drosophila S2 cells are a commonly used insect cell line to investigate RNA–protein dynamics and serve as a versatile platform for studying RBP function. Here, we describe a protocol used for identifying RNA targets of Ataxin-2, a versatile RBP involved in post-transcriptional and translational regulation, in S2 cells using TRIBE. This method allows rapid, efficient, and reliable identification of Ataxin-2-associated RNA targets and can be readily applied to other RBPs.Key features• Streamlined workflow for identifying RNA targets of RBPs using TRIBE in Drosophila S2 cells.• TRIBE highlights the sites and nature of interactions between RBPs and RNA.• TRIBE overcomes the limitations of conventional RBP–RNA interaction studies like RIP, CLIP, and their variants.