Abstract

In the natural course of spoken language, individuals often engage in thinking and self-correction during speech production. These instances of interruption or correction are commonly referred to as disfluencies. When preparing data for subsequent downstream NLP tasks, these linguistic elements can be systematically removed, or handled as required, to enhance data quality. In this study, we present a comprehensive research on disfluencies in Indian languages. Our approach involves not only annotating real-world conversation transcripts but also conducting a detailed analysis of linguistic nuances inherent to Indian languages that are necessary to consider during annotation. Additionally, we introduce a robust algorithm for the synthetic generation of disfluent data. This algorithm aims to facilitate more effective model training for the identification of disfluencies in real-world conversations, thereby contributing to the advancement of disfluency research in Indian languages

Abstract

India's education sector faces a significant challenge due to its linguistic diversity, hindering inclusivity. The dominance of English on the internet underscores the critical need for translating educational content into Indian languages to enhance accessibility. Although Speech-to-Speech Machine Translation (SSMT) technologies exist, their deficiency in reproducing intonation results in monotonous translations, diminishing audience engagement and interest in the content. To address this issue, this paper demonstrates an SSMT application with a Text-to-Speech (TTS) architecture capable of incorporating stress into synthesized speech to give a more engaging experience. The SSMT pipeline also has components like a stress classifier that captures the stress in the source speech and allows it to be utilized during speech generation. The application takes in a speech file and gives a translated speech file with stress transferred from the source.

Abstract

No Results Found

Abstract

The language diversity in India's education sector poses a significant challenge, hindering inclusivity. Despite the democratization of knowledge through online educational content, the dominance of English, as the internet's lingua franca, limits accessibility, emphasizing the crucial need for translation into Indian languages. Despite existing Speech-to-Speech Machine Translation (SSMT) technologies, the lack of intonation in these systems gives monotonous translations, leading to a loss of audience interest and disengagement from the content. To address this, our paper introduces a dataset with stress annotations in Indian English and also a Text-to-Speech (TTS) architecture capable of incorporating stress into synthesized speech. This dataset is used for training a stress detection model, which is then used in the SSMT system for detecting stress in the source speech and transferring it into the target language speech. The TTS architecture is based on FastPitch and can modify the variances based on stressed words given. We present an Indian English-to-Hindi SSMT system that can transfer stress and aim to enhance the overall quality and engagement of educational content.

Abstract

Monitoring water flow helps to identify leaks and wastage, leading to better management of water resources and conservation of this precious resource. To address this challenge, there is a need for an efficient and sustainable water management system. This paper presents an Internet of Things (IoT) based solution that involves retrofitting existing analog water meters using readily available off-the-shelf electronic components. Real-time data collection and analysis are performed through edge computation, which locally processes water meter images captured by the camera and extracts water meter readings. These readings are transmitted to the cloud for storage and further analysis. Various strategies have been implemented to optimize supply-current usage, preserving charge-discharge cycles of solar-powered batteries even in adverse environmental conditions. To streamline the firmware update process for multiple connected devices, a broadcasting technique is employed, offering the benefits of reduced manual labor and time savings. To assess the reliability and performance of developed solution, field deployment is conducted over several months, enabling the characterization of water usage patterns across different locations. Integrating energy harvesting capabilities into system reduces maintenance costs and promotes eco-friendly energy practices. Overall, this solution offers an effective and comprehensive approach towards achieving efficient and sustainable water management.

Abstract

Accurate multi-organ segmentation in abdominal CT is essential for diagnosis, treatment planning, and disease monitoring. However, it suffers from inaccurate segmentation due to intricate spatial relationships and varying organ shapes. Existing deep learning methods often struggle with imbalanced classes, size bias, and ambiguous boundaries, hindering segmentation accuracy. To address these challenges, this paper proposes Guided-nnUNet, a two-stage segmentation framework that decomposes abdominal multi-organ segmentation into organ localization and then localization-guided fine segmentation. In the first stage, a ResNet-50 model generates a low-dimensional localization map guiding organ locations. This spatial guidance is then fed into the second stage, where a 3D U-Net with dynamic affine feature-map transform performs the fine-grained segmentation by integrating spatial context from the localization map. Our evaluation on the publicly available AMOS and BTCV datasets demonstrates the effectiveness of the model. Guided-nnUNet achieves an average improvement of 7% and 9% on the AMOS and BTCV datasets, respectively, compared to the baseline nnUNet model. Additionally, our model outperforms the state-of-the-art MedNeXt by 3.6% and 5.3% on the AMOS and BTCV datasets, respectively. These results suggest that our two-stage solution offers a promising approach for accurate abdominal organ segmentation, particularly for overcoming the challenges associated with complex organ structures.

Abstract

We present a novel solution for accurate segmentation of pneumothorax from chest radiographs utilizing free-text radiology reports. Our solution employs text-guided attention to leverage the findings in the report to initially produce a low-dimensional region-localization map. These prior region maps are integrated at multiple scales in an encoder-decoder segmentation framework via dynamic affine feature map transform (DAFT). Extensive experiments on a public dataset CANDID-PTX, show that the integration of free-text reports significantly reduces the false positive predictions, while the DAFT-based fusion of localization maps improves the positive cases. Our method achieves a DSC of 0.60 for positive and 0.052 FPR for positive and negative cases, respectively, and 0.70 to 0.85 DSC for medium and large pneumothoraces.

Abstract

No Results Found

Abstract

We study a family of subcodes of the $m$-dimensional product code $mathscr{C}^{otimes m}$ (`subproduct codes') that have a recursive Plotkin-like structure, and which include Reed-Muller (RM) codes and Dual Berman codes as special cases. We denote the codes in this family as $mathscr{C}^{otimes [r,m]}$, where mbox{$0 leq r leq m$} is the `order' of the code. These codes allow a `projection' operation that can be exploited in iterative decoding, viz., the sum of two carefully chosen subvectors of any codeword in $mathscr{C}^{otimes [r,m]}$ belongs to $mathscr{C}^{otimes [r-1,m-1]}$. Recursive subproduct codes provide a wide range of rates and block lengths compared to RM codes while possessing several of their structural properties, such as the Plotkin-like design, the projection property, and fast ML decoding of first-order codes. Our simulation results for first-order and second-order codes, that are based on a belief propagation decoder and a local graph search algorithm, show instances of subproduct codes that perform either better than or within $0.5$~dB of comparable RM codes and CRC-aided Polar codes.

Abstract

Conversation is the most natural form of human communication, where each utterance can range over a variety of possible emotions. While significant work has been done towards the detection of emotions in text, relatively little work has been done towards finding the cause of the said emotions, especially in multimodal settings. SemEval 2024 introduces the task of Multimodal Emotion Cause Analysis in Conversations, which aims to extract emotions reflected in individual utterances in a conversation involving multiple modalities (textual, audio, and visual modalities) along with the corresponding utterances that were the cause for the emotion. In this paper, we propose models that tackle this task as an utterance labeling and a sequence labeling problem and perform a comparative study of these models, involving baselines using different encoders, using BiLSTM for adding contextual information of the conversation, and finally adding a CRF layer to try to model the inter-dependencies between adjacent utterances more effectively. In the official leaderboard for the task, our architecture was ranked 8th, achieving an F1-score of 0.1759 on the leaderboard.