Sumit Soman

Sujoy Roychowdhury, Sumit Soman, H G Ranjani et al.

Retrieval Augmented Generation (RAG) is widely used to enable Large Language Models (LLMs) perform Question Answering (QA) tasks in various domains. However, RAG based on open-source LLM for specialized domains has challenges of evaluating generated responses. A popular framework in the literature is the RAG Assessment (RAGAS), a publicly available library which uses LLMs for evaluation. One disadvantage of RAGAS is the lack of details of derivation of numerical value of the evaluation metrics. One of the outcomes of this work is a modified version of this package for few metrics (faithfulness, context relevance, answer relevance, answer correctness, answer similarity and factual correctness) through which we provide the intermediate outputs of the prompts by using any LLMs. Next, we analyse the expert evaluations of the output of the modified RAGAS package and observe the challenges of using it in the telecom domain. We also study the effect of the metrics under correct vs. wrong retrieval and observe that few of the metrics have higher values for correct retrieval. We also study for differences in metrics between base embeddings and those domain adapted via pre-training and fine-tuning. Finally, we comment on the suitability and challenges of using these metrics for in-the-wild telecom QA task.

Giriprasad Sridhara, Sujoy Roychowdhury, Sumit Soman et al.

This paper presents our findings on the automatic summarization of Java methods within Ericsson, a global telecommunications company. We evaluate the performance of an approach called Automatic Semantic Augmentation of Prompts (ASAP), which uses a Large Language Model (LLM) to generate leading summary comments for Java methods. ASAP enhances the $LLM's$ prompt context by integrating static program analysis and information retrieval techniques to identify similar exemplar methods along with their developer-written Javadocs, and serves as the baseline in our study. In contrast, we explore and compare the performance of four simpler approaches that do not require static program analysis, information retrieval, or the presence of exemplars as in the ASAP method. Our methods rely solely on the Java method body as input, making them lightweight and more suitable for rapid deployment in commercial software development environments. We conducted experiments on an Ericsson software project and replicated the study using two widely-used open-source Java projects, Guava and Elasticsearch, to ensure the reliability of our results. Performance was measured across eight metrics that capture various aspects of similarity. Notably, one of our simpler approaches performed as well as or better than the ASAP method on both the Ericsson project and the open-source projects. Additionally, we performed an ablation study to examine the impact of method names on Javadoc summary generation across our four proposed approaches and the ASAP method. By masking the method names and observing the generated summaries, we found that our approaches were statistically significantly less influenced by the absence of method names compared to the baseline. This suggests that our methods are more robust to variations in method names and may derive summaries more comprehensively from the method body than the ASAP approach.

Sujoy Roychowdhury, Sumit Soman, HG Ranjani et al.

With the ubiquitous use of document corpora for question answering, one important aspect which is especially relevant for technical documents is the ability to extract information from tables which are interspersed with text. The major challenge in this is that unlike free-flow text or isolated set of tables, the representation of a table in terms of what is a relevant chunk is not obvious. We conduct a series of experiments examining various representations of tabular data interspersed with text to understand the relative benefits of different representations. We choose a corpus of $3^{rd}$ Generation Partnership Project (3GPP) documents since they are heavily interspersed with tables. We create expert curated dataset of question answers to evaluate our approach. We conclude that row level representations with corresponding table header information being included in every cell improves the performance of the retrieval, thus leveraging the structural information present in the tabular data.

Sumit Soman, Sujoy Roychowdhury

Retrieval augmented generation (RAG) for technical documents creates challenges as embeddings do not often capture domain information. We review prior art for important factors affecting RAG and perform experiments to highlight best practices and potential challenges to build RAG systems for technical documents.

Sujoy Roychowdhury, H. G. Ranjani, Sumit Soman et al.

There has been limited evaluation of advanced Text-to-Speech (TTS) models with Mathematical eXpressions (MX) as inputs. In this work, we design experiments to evaluate quality and intelligibility of five TTS models through listening and transcribing tests for various categories of MX. We use two Large Language Models (LLMs) to generate English pronunciation from LaTeX MX as TTS models cannot process LaTeX directly. We use Mean Opinion Score from user ratings and quantify intelligibility through transcription correctness using three metrics. We also compare listener preference of TTS outputs with respect to human expert rendition of same MX. Results establish that output of TTS models for MX is not necessarily intelligible, the gap in intelligibility varies across TTS models and MX category. For most categories, performance of TTS models is significantly worse than that of expert rendition. The effect of choice of LLM is limited. This establishes the need to improve TTS models for MX.

Sumit Soman, H. G. Ranjani, Sujoy Roychowdhury et al.

Question-Answering (QA) from technical documents often involves questions whose answers are present in figures, such as flowcharts or flow diagrams. Text-based Retrieval Augmented Generation (RAG) systems may fail to answer such questions. We leverage graph representations of flowcharts obtained from Visual large Language Models (VLMs) and incorporate them in a text-based RAG system to show that this approach can enable image retrieval for QA in the telecom domain. We present the end-to-end approach from processing technical documents, classifying image types, building graph representations, and incorporating them with the text embedding pipeline for efficient retrieval. We benchmark the same on a QA dataset created based on proprietary telecom product information documents. Results show that the graph representations obtained using a fine-tuned VLM model have lower edit distance with respect to the ground truth, which illustrate the robustness of these representations for flowchart images. Further, the approach for QA using these representations gives good retrieval performance using text-based embedding models, including a telecom-domain adapted one. Our approach also alleviates the need for a VLM in inference, which is an important cost benefit for deployed QA systems.

Rishika Sen, Sujoy Roychowdhury, Sumit Soman et al.

Knowledge Distillation (KD) is one of the approaches to reduce the size of Large Language Models (LLMs). A LLM with smaller number of model parameters (student) is trained to mimic the performance of a LLM of a larger size (teacher model) on a specific task. For domain-specific tasks, it is not clear if teacher or student model, or both, must be considered for domain adaptation. In this work, we study this problem from perspective of telecom domain Question-Answering (QA) task. We systematically experiment with Supervised Fine-tuning (SFT) of teacher only, SFT of student only and SFT of both prior to KD. We design experiments to study the impact of vocabulary (same and different) and KD algorithms (vanilla KD and Dual Space KD, DSKD) on the distilled model. Multi-faceted evaluation of the distillation using 14 different metrics (N-gram, embedding and LLM-based metrics) is considered. Experimental results show that SFT of teacher improves performance of distilled model when both models have same vocabulary, irrespective of algorithm and metrics. Overall, SFT of both teacher and student results in better performance across all metrics, although the statistical significance of the same depends on the vocabulary of the teacher models.

Sujoy Roychowdhury, Sumit Soman, Ranjani Hosakere Gireesha et al.

A plethora of sentence embedding models makes it challenging to choose one, especially for technical domains rich with specialized vocabulary. In this work, we domain adapt embeddings using telecom data for question answering. We evaluate embeddings obtained from publicly available models and their domain-adapted variants, on both point retrieval accuracies, as well as their (95%) confidence intervals. We establish a systematic method to obtain thresholds for similarity scores for different embeddings. As expected, we observe that fine-tuning improves mean bootstrapped accuracies. We also observe that it results in tighter confidence intervals, which further improve when pre-training is preceded by fine-tuning. We introduce metrics which measure the distributional overlaps of top-$K$, correct and random document similarities with the question. Further, we show that these metrics are correlated with retrieval accuracy and similarity thresholds. Recent literature shows conflicting effects of isotropy on retrieval accuracies. Our experiments establish that the isotropy of embeddings (as measured by two independent state-of-the-art isotropy metric definitions) is poorly correlated with retrieval performance. We show that embeddings for domain-specific sentences have little overlap with those for domain-agnostic ones, and fine-tuning moves them further apart. Based on our results, we provide recommendations for use of our methodology and metrics by researchers and practitioners.

Sumit Soman, Ranjani H G

The landscape for building conversational interfaces (chatbots) has witnessed a paradigm shift with recent developments in generative Artificial Intelligence (AI) based Large Language Models (LLMs), such as ChatGPT by OpenAI (GPT3.5 and GPT4), Google's Bard, Large Language Model Meta AI (LLaMA), among others. In this paper, we analyze capabilities and limitations of incorporating such models in conversational interfaces for the telecommunication domain, specifically for enterprise wireless products and services. Using Cradlepoint's publicly available data for our experiments, we present a comparative analysis of the responses from such models for multiple use-cases including domain adaptation for terminology and product taxonomy, context continuity, robustness to input perturbations and errors. We believe this evaluation would provide useful insights to data scientists engaged in building customized conversational interfaces for domain-specific requirements.

Kartikeya Badola, Sameer Ambekar, Himanshu Pant et al.

The gold standard for COVID-19 is RT-PCR, testing facilities for which are limited and not always optimally distributed. Test results are delayed, which impacts treatment. Expert radiologists, one of whom is a co-author, are able to diagnose COVID-19 positivity from Chest X-Rays (CXR) and CT scans, that can facilitate timely treatment. Such diagnosis is particularly valuable in locations lacking radiologists with sufficient expertise and familiarity with COVID-19 patients. This paper has two contributions. One, we analyse literature on CXR based COVID-19 diagnosis. We show that popular choices of dataset selection suffer from data homogeneity, leading to misleading results. We compile and analyse a viable benchmark dataset from multiple existing heterogeneous sources. Such a benchmark is important for realistically testing models. Our second contribution relates to learning from imbalanced data. Datasets for COVID X-Ray classification face severe class imbalance, since most subjects are COVID -ve. Twin Support Vector Machines (Twin SVM) and Twin Neural Networks (Twin NN) have, in recent years, emerged as effective ways of handling skewed data. We introduce a state-of-the-art technique, termed as Twin Augmentation, for modifying popular pre-trained deep learning models. Twin Augmentation boosts the performance of a pre-trained deep neural network without requiring re-training. Experiments show, that across a multitude of classifiers, Twin Augmentation is very effective in boosting the performance of given pre-trained model for classification in imbalanced settings.

Himanshu Pant, Jayadeva, Sumit Soman

One of the issues faced in training Generative Adversarial Nets (GANs) and their variants is the problem of mode collapse, wherein the training stability in terms of the generative loss increases as more training data is used. In this paper, we propose an alternative architecture via the Low-Complexity Neural Network (LCNN), which attempts to learn models with low complexity. The motivation is that controlling model complexity leads to models that do not overfit the training data. We incorporate the LCNN loss function for GANs, Deep Convolutional GANs (DCGANs) and Spectral Normalized GANs (SNGANs), in order to develop hybrid architectures called the LCNN-GAN, LCNN-DCGAN and LCNN-SNGAN respectively. On various large benchmark image datasets, we show that the use of our proposed models results in stable training while avoiding the problem of mode collapse, resulting in better training stability. We also show how the learning behavior can be controlled by a hyperparameter in the LCNN functional, which also provides an improved inception score.

Siddharth Srivastava, Sumit Soman, Astha Rai

This paper introduces a novel approach for dengue fever classification based on online learning paradigms. The proposed approach is suitable for practical implementation as it enables learning using only a few training samples. With time, the proposed approach is capable of learning incrementally from the data collected without need for retraining the model or redeployment of the prediction engine. Additionally, we also provide a comprehensive evaluation of machine learning methods for prediction of dengue fever. The input to the proposed pipeline comprises of recorded patient symptoms and diagnostic investigations. Offline classifier models have been employed to obtain baseline scores to establish that the feature set is optimal for classification of dengue. The primary benefit of the online detection model presented in the paper is that it has been established to effectively identify patients with high likelihood of dengue disease, and experiments on scalability in terms of number of training and test samples validate the use of the proposed model.

Mayank Sharma, Aayush Yadav, Sumit Soman et al.

Deep neural networks are over-parameterized, which implies that the number of parameters are much larger than the number of samples used to train the network. Even in such a regime deep architectures do not overfit. This phenomenon is an active area of research and many theories have been proposed trying to understand this peculiar observation. These include the Vapnik Chervonenkis (VC) dimension bounds and Rademacher complexity bounds which show that the capacity of the network is characterized by the norm of weights rather than the number of parameters. However, the effect of input noise on these measures for shallow and deep architectures has not been studied. In this paper, we analyze the effects of various regularization schemes on the complexity of a neural network which we characterize with the loss, $L_2$ norm of the weights, Rademacher complexities (Directly Approximately Regularizing Complexity-DARC1), VC dimension based Low Complexity Neural Network (LCNN) when subject to varying degrees of Gaussian input noise. We show that $L_2$ regularization leads to a simpler hypothesis class and better generalization followed by DARC1 regularizer, both for shallow as well as deeper architectures. Jacobian regularizer works well for shallow architectures with high level of input noises. Spectral normalization attains highest test set accuracies both for shallow and deeper architectures. We also show that Dropout alone does not perform well in presence of input noise. Finally, we show that deeper architectures are robust to input noise as opposed to their shallow counterparts.

Mayank Sharma, Jayadeva, Sumit Soman

Explaining the unreasonable effectiveness of deep learning has eluded researchers around the globe. Various authors have described multiple metrics to evaluate the capacity of deep architectures. In this paper, we allude to the radius margin bounds described for a support vector machine (SVM) with hinge loss, apply the same to the deep feed-forward architectures and derive the Vapnik-Chervonenkis (VC) bounds which are different from the earlier bounds proposed in terms of number of weights of the network. In doing so, we also relate the effectiveness of techniques like Dropout and Dropconnect in bringing down the capacity of the network. Finally, we describe the effect of maximizing the input as well as the output margin to achieve an input noise-robust deep architecture.

Jayadeva, Himanshu Pant, Mayank Sharma et al.

Modern neural network architectures for large-scale learning tasks have substantially higher model complexities, which makes understanding, visualizing and training these architectures difficult. Recent contributions to deep learning techniques have focused on architectural modifications to improve parameter efficiency and performance. In this paper, we derive a continuous and differentiable error functional for a neural network that minimizes its empirical error as well as a measure of the model complexity. The latter measure is obtained by deriving a differentiable upper bound on the Vapnik-Chervonenkis (VC) dimension of the classifier layer of a class of deep networks. Using standard backpropagation, we realize a training rule that tries to minimize the error on training samples, while improving generalization by keeping the model complexity low. We demonstrate the effectiveness of our formulation (the Low Complexity Neural Network - LCNN) across several deep learning algorithms, and a variety of large benchmark datasets. We show that hidden layer neurons in the resultant networks learn features that are crisp, and in the case of image datasets, quantitatively sharper. Our proposed approach yields benefits across a wide range of architectures, in comparison to and in conjunction with methods such as Dropout and Batch Normalization, and our results strongly suggest that deep learning techniques can benefit from model complexity control methods such as the LCNN learning rule.

Jayadeva, Himanshu Pant, Sumit Soman et al.

Twin Support Vector Machines (TWSVMs) have emerged an efficient alternative to Support Vector Machines (SVM) for learning from imbalanced datasets. The TWSVM learns two non-parallel classifying hyperplanes by solving a couple of smaller sized problems. However, it is unsuitable for large datasets, as it involves matrix operations. In this paper, we discuss a Twin Neural Network (Twin NN) architecture for learning from large unbalanced datasets. The Twin NN also learns an optimal feature map, allowing for better discrimination between classes. We also present an extension of this network architecture for multiclass datasets. Results presented in the paper demonstrate that the Twin NN generalizes well and scales well on large unbalanced datasets.

Sumit Soman, Siddharth Srivastava, Saurabh Srivastava et al.

In recent times, there have been significant advancements in utilizing the sensing capabilities of mobile devices for developing applications. The primary objective has been to enhance the way a user interacts with the application by making it effortless and convenient. This paper explores the capabilities of using Brain Computer Interfaces (BCI), an evolving subset of Human Computer Interaction (HCI) paradigms, to control mobile devices. We present a comprehensive survey of the state-of-the-art in this area, discussing the challenges and limitations in using BCI for mobile applications. Further we propose possible modalities that in future can benefit with BCI applications. This paper consolidates research directions being pursued in this domain, and draws conclusions on feasibility and benefits of using BCI systems effectively augmented to the mobile application development domain.

Udit Kumar, Sumit Soman, Jayadeva

This paper presents a comparative analysis of the performance of the Incremental Ant Colony algorithm for continuous optimization ($IACO_\mathbb{R}$), with different algorithms provided in the NLopt library. The key objective is to understand how the various algorithms in the NLopt library perform in combination with the Multi Trajectory Local Search (Mtsls1) technique. A hybrid approach has been introduced in the local search strategy by the use of a parameter which allows for probabilistic selection between Mtsls1 and a NLopt algorithm. In case of stagnation, the algorithm switch is made based on the algorithm being used in the previous iteration. The paper presents an exhaustive comparison on the performance of these approaches on Soft Computing (SOCO) and Congress on Evolutionary Computation (CEC) 2014 benchmarks. For both benchmarks, we conclude that the best performing algorithm is a hybrid variant of Mtsls1 with BFGS for local search.

Jayadeva, Sumit Soman, Amit Bhaya

The recently proposed Minimal Complexity Machine (MCM) finds a hyperplane classifier by minimizing an exact bound on the Vapnik-Chervonenkis (VC) dimension. The VC dimension measures the capacity of a learning machine, and a smaller VC dimension leads to improved generalization. On many benchmark datasets, the MCM generalizes better than SVMs and uses far fewer support vectors than the number used by SVMs. In this paper, we describe a neural network based on a linear dynamical system, that converges to the MCM solution. The proposed MCM dynamical system is conducive to an analogue circuit implementation on a chip or simulation using Ordinary Differential Equation (ODE) solvers. Numerical experiments on benchmark datasets from the UCI repository show that the proposed approach is scalable and accurate, as we obtain improved accuracies and fewer number of support vectors (upto 74.3% reduction) with the MCM dynamical system.