In the evolving realm of social media, music significantly enhances post appeal and viewer engagement. However, challenges such as copyright and royalties complicate its usage. Artificial intelligence (AI) might be used to generate royalty- free music that complements visual content. This paper presents an AI-based melody generator designed to create music suitable for various applications, particularly for enhancing social media posts. Unlike full songs, which involve complex AI models, our focus on melodies addresses a more specific and manageable aspect of music generation. We developed an algorithm to differentiate between main and background melodies, leveraging an LSTM and Transformer architecture to capture musical dependencies. Training on the Lakn MIDI dataset, which includes 178,000 files, our model achieved 64% accuracy in predicting main melodies and 78% in background melodies. Evaluation by 23 volunteers revealed that AI-generated melodies were as pleasant as human-composed ones and revealed that participants struggled to distinguish whether the melody they heard was human-composed or AI-generated. This indicates that our AI model might offer significant benefits in scenarios where melodies play an important role.

Sequence modeling neural networks are being deployed over a growing number of applications, a phenomenon partially motivated by the advent of Large Language Models (LLMs), Neural Networks often characterized by billions of parameters. Such a size poses an obstacle to their deployment in constrained devices, which motivates the development of compression methods. In this work, we introduce a new parameter-sharing method that leverages the embedding matrix to learn the model's hidden layers. To demonstrate its effectiveness, we present a new architecture family called ShareBERT, which can preserve up to 95.5% of BERT accuracy performances, using only 5M parameters (21.9× fewer parameters) without the help of Knowledge Distillation. The evaluation of multiple linguistic benchmarks showcases that our compression method does not negatively affect the model's learning capabilities, instead, it can be beneficial for representation learning. The method is robust and flexible across different neural architecture types (such as Recurrent, Convolution, and Transformers), layers (e.g., encoder, decoder, autoregressive, and non-autoregressive modules), and tasks (e.g., translation, captioning, and language modeling). Our proposal pushes the model compression to a new level by enabling the design of near-zero architectures, and on top of that, it is orthogonal to most existing approaches, which can be further applied to ease the deployment in low-powered and embedded devices. Code is available at https://github.com/jchenghu/sharebert.

The deployment of Pre-trained Language Models in memory-limited devices is hindered by their massive number of parameters, which motivated the interest in developing smaller architectures. Established works in the model compression literature showcased that small models often present a noticeable performance degradation and need to be paired with transfer learning methods, such as Knowledge Distillation. In this work, we propose a parameter-sharing method that consists of sharing parameters between embeddings and the hidden layers, enabling the design of near-zero parameter encoders. To demonstrate its effectiveness, we present an architecture design called ShareBERT, which can preserve up to 95.5% of BERT Base performances, using only 5M parameters (21.9× fewer parameters) without the help of Knowledge Distillation. We demonstrate empirically that our proposal does not negatively affect the model learning capabilities and that it is even beneficial for representation learning. Code will be available at https://github.com/jchenghu/sharebert.

The emergence of digital music platforms has fundamentally transformed the way we engage with and organize music. As playlist creation has gained widespread popularity, there is an increasing desire among music aficionados and industry experts to comprehend the factors that drive playlist success. This paper presents a machine learning-based approach designed to predict the success of music playlists. By analyzing various musical characteristics of songs, our model achieves an impressive accuracy of 89.6% in predicting playlist success. Notably, it exhibits a remarkable 92.0% accuracy in forecasting the success of popular playlists, while also effectively identifying unpopular playlists with an accuracy of 89.4%. These findings provide invaluable insights into playlist creation, ultimately enhancing the overall music-listening experience. By harnessing the power of machine learning, our proposed approach unlocks new prospects for optimizing playlist design strategies and delivering personalized music recommendations. This has significant ramifications for music enthusiasts and industry professionals seeking to elevate playlist creation and enrich the music consumption experience.

Resource scarcity in Neural Machine Translation is a challenging problem in both industry applications and in the support of less-spoken languages represented, in the worst case, by endangered and low-resource languages. Many Data Augmentation methods rely on additional linguistic sources and software tools but these are often not available in less favoured language. For this reason, we present USKI (Unaligned Sentences Keytokens pre-traIning), a pre-training strategy that leverages the relationships and similarities that exist between unaligned sentences. By doing so, we increase the dataset size of endangered and low-resource languages by the square of the initial quantity, matching the typical size of high-resource language datasets such as WMT14 En-Fr. Results showcase the effectiveness of our approach with an increase on average of 0.9 BLEU across the benchmarks using a small fraction of the entire unaligned corpus, suggesting the importance of the research topic and the potential of a currently under-utilized resource and under-explored approach.

We introduce a method called the Expansion mechanism that processes the input unconstrained by the number of elements in the sequence. By doing so, the model can learn more effectively compared to traditional attention-based approaches. To support this claim, we design a novel architecture ExpansionNet v2 that achieved strong results on the MS COCO 2014 Image Captioning challenge and the State of the Art in its respective category, with a score of 143.7 CIDErD in the offline test split, 140.8 CIDErD in the online evaluation server and 72.9 AllCIDEr on the nocaps validation set. Additionally, we introduce an End to End training algorithm up to 2.8 times faster than established alternatives.

Increased network bandwidth is making desktop video conferencing an attractive application for an increasing number of computer users. Unfortunately, two competing standards for video conferencing signaling are in use, H.323 and SIP. In this paper we look at the interoperability between these two standards by developing a conferencing gateway that supports conferences involving both SIP and H.323 clients. By appropriately translating between H.323 and SIP operations, our prototype gateway supports basic multi-party video conferencing between NetMeeting (an H.323 client) and VIC (a SIP client) without modifications to the clients. However, our experiments also show that seamless interoperation would require changes to the client implementations and the standards.