2024

1. AAAI ’24 Conference

   Knowledge-Powered Recommendation for an Improved Diet Water Footprint

   Saurav Joshi, Filip Ilievski, and Jay Pujara

   Proceedings of the AAAI Conference on Artificial Intelligence Mar 2024

   Abs URL

   According to WWF, 1.1 billion people lack access to water, and 2.7 billion experience water scarcity at least one month a year. By 2025, two-thirds of the world’s population may be facing water shortages. This highlights the urgency of managing water usage efficiently, especially in water-intensive sectors like food. This paper proposes a recommendation engine, powered by knowledge graphs, aiming to facilitate sustainable and healthy food consumption. The engine recommends ingredient substitutes in user recipes that improve nutritional value and reduce environmental impact, particularly water footprint. The system architecture includes source identification, information extraction, schema alignment, knowledge graph construction, and user interface development. The research offers a promising tool for promoting healthier eating habits and contributing to water conservation efforts.

2023

1. Arxiv Preprint

   Contextualizing Internet Memes Across Social Media Platforms

   Saurav Joshi, Filip Ilievski, and Luca Luceri

   arXiv preprint arXiv:2311.11157 2023

   Abs URL

   Internet memes have emerged as a novel format for communication and expressing ideas on the web. Their fluidity and creative nature are reflected in their widespread use, often across platforms and occasionally for unethical or harmful purposes. While computational work has already analyzed their high-level virality over time and developed specialized classifiers for hate speech detection, there have been no efforts to date that aim to holistically track, identify, and map internet memes posted on social media. To bridge this gap, we investigate whether internet memes across social media platforms can be contextualized by using a semantic repository of knowledge, namely, a knowledge graph. We collect thousands of potential internet meme posts from two social media platforms, namely Reddit and Discord, and perform an extract-transform-load procedure to create a data lake with candidate meme posts. By using vision transformer-based similarity, we match these candidates against the memes cataloged in a recently released knowledge graph of internet memes, IMKG. We provide evidence that memes published online can be identified by mapping them to IMKG. We leverage this grounding to study the prevalence of memes on different platforms, discover popular memes, and select common meme channels and subreddits. Finally, we illustrate how the grounding can enable users to get context about memes on social media thanks to their link to the knowledge graph.

2. CACM ’23 Journal

   Identifying and consolidating knowledge engineering requirements

   Bradley P Allen, Filip Ilievski, and Saurav Joshi

   arXiv preprint arXiv:2306.15124 2023

   Abs URL

   Knowledge engineering is the process of creating and maintaining knowledge-producing systems. Throughout the history of computer science and AI, knowledge engineering workflows have been widely used because high-quality knowledge is assumed to be crucial for reliable intelligent agents. However, the landscape of knowledge engineering has changed, presenting four challenges: unaddressed stakeholder requirements, mismatched technologies, adoption barriers for new organizations, and misalignment with software engineering practices. In this paper, we propose to address these challenges by developing a reference architecture using a mainstream software methodology. By studying the requirements of different stakeholders and eras, we identify 23 essential quality attributes for evaluating reference architectures. We assess three candidate architectures from recent literature based on these attributes. Finally, we discuss the next steps towards a comprehensive reference architecture, including prioritizing quality attributes, integrating components with complementary strengths, and supporting missing socio-technical requirements. As this endeavor requires a collaborative effort, we invite all knowledge engineering researchers and practitioners to join us.

2022

1. INCET ’22 Conference

   A Study of Various Word Embeddings in Deep Learning

   Parth Shah, Smit Shah, and Saurav Joshi

   In 2022 3rd International Conference for Emerging Technology (INCET) 2022

   Abs URL

   With the number of users and reviews rising on the internet, and the liberty was given to post anything on it, it is still a major issue to identify the sentiment of the reviews. The traditional machine learning models are taking a lot of computational time to get desired results. The advancement in the field of deep learning has left people wondering whether it would provide better results for such applications. Word Embeddings play a huge role in the classification of text, as they represent similar meaning documents with equivalent representation. The result of the deep learning model trained on a particular set of data depends upon the sort of word embedding used. In this work, we intend to present a study of various embeddings like GloVe, BERT, Word2vec, and Fast-text using a proposed model. The research was conducted on the IMDB movie review dataset and the result displayed that GloVe embeddings produced the best result with an accuracy of 87.936 % on our proposed model.

2021

1. ICCCNT ’21 Conference

   Emotion Based Music Recommendation System Using LSTM - CNN Architecture

   Saurav Joshi, Tanuj Jain, and Nidhi Nair

   In 2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT) 2021

   Abs URL

   In the current era of media and technology, music information retrieval techniques have made progress in the past years but the music recommendation system is still at a rudimentary stage. However personalized music recommendation is quotidian, but recommending songs based on emotions is still an uphill battle. Music has greatly influenced the human brain and helps dispense an exhilarating and frivolous state of mind because it helps us work more effectively. Recommending songs based on emotions will comfort the listener by suggesting music in keeping with the listeners’ pervading mental and physical state. Hence, Natural Language Processing and Deep Learning technologies made it possible for machines to read and interpret emotions through texts by recognizing patterns and finding correlations. In this paper, various deep learning models such as Long Short-Term Memory (LSTM), Convolution Neural network (CNN), CNN-LSTM, and LSTM-CNN Architectures were collated for detecting emotions such as angry, happy, love, and sad, the best model was integrated into the application. To enhance the application, a CNN model was used to detect emotions through facial expressions. The application takes text input from the user or a facial expression input. Depending upon the emotion detected, it recommends the user songs and playlists.

2. ICIRCA ’21 Conference

   Study of Various Dimensionality Reduction and Classification Algorithms on High Dimensional Dataset

   Smit Shah, and Saurav Joshi

   In 2021 Third International Conference on Inventive Research in Computing Applications (ICIRCA) 2021

   Abs URL

   A potential drawback of huge data is that it makes analysis of the data hard and also computationally infeasible. Health care, finance, retail, and education are a few of the data mining applications that involve very high-dimensional data. A large number of dimensions introduce a popular problem of “Curse of Dimensionality”. This problem makes it difficult to perform classification and engenders lower accuracy of machine learning classifiers. This paper computes a threshold value (35%) to which if the data is reduced, the best accuracy can be obtained. Further, this research work considers an image dataset of very high dimensions on which different dimensionality reduction techniques such as PCA, LDA, and SVD are performed to find out the best dimension fit for an image dataset. Also, various ML classification algorithms, such as Logistic Regression, Random Forest Classifier, Naive Bayes, and SVM are applied to find out the best classifier for the dimensionally reduced dataset. Finally, this research work has concluded that, PCA+SVM, LDA+Random Forest, and SVD+SVM have produced the best results out of all the possible combinations from the comparative study.