Convolutional Neural Networks (CNNs) have been recently introduced in the domain of session-based next item recommendation. An ordered collection past items user has interacted with a session (or sequence) are embedded into 2-dimensional latent matrix, and treated as an image. The convolution pooling operations then applied to mapped embeddings. In this paper, we first examine typical CNN recommender show that both generative model network architecture suboptimal when modeling long-range...

In session-based or sequential recommendation, it is important to consider a number of factors like long-term user engagement, multiple types user-item interactions such as clicks, purchases etc. The current state-of-the-art supervised approaches fail model them appropriately. Casting recommendation task reinforcement learning (RL) problem promising direction. A major component RL train the agent through with environment. However, often problematic recommender in an on-line fashion due...

Traditionally, the efficiency and effectiveness of search systems have both been great interest to information retrieval community. However, an in-depth analysis on interplay between response latency web users' experience has missing so far. In order fill this gap, we conduct two separate studies aiming reveal how affects user behavior in search. First, a controlled study trying understand users perceive system sensitive they are increasing delays response. This reveals that, when artificial...

Online content providers, such as news portals and social media platforms, constantly seek new ways to attract large shares of online attention by keeping their users engaged. A common challenge is identify which aspects interaction influence user engagement the most. In this article, through an analysis a article collection obtained from Y ahoo N ews US , we demonstrate that articles exhibit considerable variation in terms sentimentality polarity content, depending on factors provider...

User feedback is considered to be a critical element in the information seeking process. An important aspect of cycle relevance assessment that has progressively become popular practice web searching activities and interactive retrieval (IR). The value lies disambiguation user's need, which achieved by applying various techniques. Such techniques vary from explicit implicit help determine retrieved documents.The former type usually obtained through intended indication documents as relevant...

Since the inception of Recommender Systems (RS), accuracy recommendations in terms relevance has been golden criterion for evaluating quality RS algorithms. However, by focusing on item relevance, one pays a significant price other important metrics: users get stuck "filter bubble" and their array options is significantly reduced, hence degrading user experience leading to churn. Recommendation, particular session-based/sequential recommendation, complex task with multiple - often...

Multimedia search systems face a number of challenges, emanating mainly from the semantic gap problem. Implicit feedback is considered useful technique in addressing many semantic-related issues. By analysing implicit information can tailor criteria to address more effectively users' needs. In this paper we examine whether could employ affective as an source evidence, through aggregation various sensory channels. These channels range between facial expressions neuro-physiological signals and...

The availability of large volumes interaction data and scalable mining techniques have made possible to study the online behaviour for millions Web users. Part efforts focused on understanding how users interact engage with web content. However, measurement within-content engagement remains a difficult unsolved task. This is because lack standardised, well-validated methods measuring engagement, especially in an context. To address this gap, we perform controlled user where observe respond...

Not all smartphone owners use their device in the same way. In this work, we uncover broad, latent patterns of mobile phone behavior. We conducted a study where, via dedicated logging app, collected daily activity data from sample 340 participants for period four weeks. Through an unsupervised learning approach and methodologically rigorous analysis, reveal five generic profiles which describe at least 10% each: limited use, business power personality- & externally induced problematic use....

Casting session-based or sequential recommendation as reinforcement learning (RL) through reward signals is a promising research direction towards recommender systems (RS) that maximize cumulative profits. However, the direct use of RL algorithms in RS setting impractical due to challenges like off-policy training, huge action spaces and lack sufficient signals. Recent approaches for attempt tackle these by combining (self-)supervised learning, but still suffer from certain limitations. For...

Multimodal foundation models are transformative in sequential recommender systems, leveraging powerful representation learning capabilities. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt for recommendation tasks, most research prioritizes parameter efficiency, often overlooking critical factors like GPU memory efficiency and training speed. Addressing this gap, our paper introduces IISAN (Intra- Inter-modal Side Adapted Network Representation), a simple plug-and-play...

Recent advancements in Recommender Systems (RS) have incorporated Reinforcement Learning (RL), framing the recommendation as a Markov Decision Process (MDP). However, offline RL policies trained on static user data are vulnerable to distribution shift when deployed dynamic online environments. Additionally, excessive focus exploiting short-term relevant items can hinder exploration, leading suboptimal recommendations and negatively impacting long-term gains. Online RL-based RS also face...

Popular Micro-videos, dominant on platforms like TikTok and YouTube, hold significant commercial value. The rise of high-quality AI-generated content has spurred interest in AI-driven micro-video creation. However, despite the advanced capabilities large language models (LLMs) ChatGPT DeepSeek text generation reasoning, their potential to assist creation popular micro-videos remains largely unexplored. In this paper, we conduct an empirical study LLM-assisted (LLMPopcorn). Specifically,...

Over the years, recommender systems have been systematically applied in both industry and academia to assist users dealing with information overload. One of factors that determine performance a system is user feedback, which has traditionally communicated through application explicit implicit feedback techniques. In this paper, we propose novel video search interface predicts topical relevance by analysing affective aspects behaviour. We, furthermore, present method for incorporating such...

Recommender systems have been systematically applied in industry and academia to help users cope with information uncertainty. However, given the multiplicity of preferences needs it has shown that no approach is suitable for all situations. Thus, believed an effective recommender system should incorporate a variety techniques features offer valuable recommendations enhance search experience. In this paper we propose novel video interface employs multimodal system, which can predict topical...

Information retrieval systems face a number of challenges, originating mainly from the semantic gap problem. Implicit feedback techniques have been employed in past to address many these issues. Although this was step towards right direction, need personalise and tailor search experience user-specific needs has become evident. In study we examine ways personalising affective models trained on facial expression data. Using personalised data adapt individual users compare their performance...

Predicting user engagement with direct displays (DD) is of paramount importance to commercial search engines, as well performance evaluation. However, understanding within-content on a web page not trivial task mainly because two reasons: (1) subjective and different users may exhibit behavioural patterns; (2) existing proxies (e.g., clicks, dwell time) suffer from certain caveats, such the well-known position bias, are effective in discriminating between useful non-useful components. In...

Data imputation and data generation have important applications for many domains, like healthcare finance, where incomplete or missing can hinder accurate analysis decision-making. Diffusion models emerged as powerful generative capable of capturing complex distributions across various modalities such image, audio, time series data. Recently, they been also adapted to generate tabular In this paper, we propose a diffusion model that introduces three key enhancements: (1) conditioning...

Objective measurements of engagement are increasingly sought after by both the media industry and scholar communities to explain what drives people consume audiovisual contents. However, is a complex construct that, at psychological level, has been mainly operationalised through indicators attentional emotional processes, often overlooking motivational factors. We claim that in context news consumption, motivation, as intrinsic interest for consuming given content, needs be factored together...

Understanding the impact of a search system's response latency on its users' searching behaviour has been recently an active research topic in information retrieval and human-computer interaction areas. Along same line, this paper focuses user makes following two contributions. First, through controlled experiment, we reveal physiological effects users show that these are present even at small increases latency. We compare with gathered from self-reports they capture nuanced attentional...

In recent years, the search engine results pages (SERP's) have been augmented with new markup elements that introduce seamlessly additional semantic information. Examples of such are aggregated disseminated by vertical portals, and enriched snippets display meta-information from landing pages. this paper, we investigate gaze behaviour web users who inter- act SERP's contain plain rich snippets, observe impact both types on experience. For our study, consider a wide range snippet types, as...