Recommender systems have become pervasive on the web, shaping way users see information and thus decisions they make. As these get more complex, there is a growing need for transparency. In this paper, we study problem of generating visualizing personalized explanations hybrid recommender systems, which incorporate many different data sources. We build upon probabilistic graphical model develop an approach to generate real-time recommendations along with explanations. To benefits conduct...

As the amount of recorded digital information increases, there is a growing need for flexible recommender systems which can incorporate richly structured data sources to improve recommendations. In this paper, we show how recently introduced statistical relational learning framework be used develop generic and extensible hybrid system. Our approach, HyPER (HYbrid Probabilistic Extensible Recommender), incorporates reasons over wide range sources. Such include multiple user-user item-item...

Hybrid recommender systems combine several different sources of information to generate recommendations. These demonstrate improved accuracy compared single-source recommendation strategies. However, hybrid strategies are inherently more complex than those that use a single source information, and thus the process explaining recommendations users becomes challenging. In this paper we describe system built on probabilistic programming language, discuss benefits challenges its users. We...

Recommender systems are ubiquitous and shape the way users access information make decisions. As these become more complex, there is a growing need for transparency interpretability. In this article, we study problem of generating visualizing personalized explanations recommender that incorporate signals from many different data sources. We use flexible, extendable probabilistic programming approach show how can generate real-time recommendations. then turn recommendations into explanations....

Entity resolution in settings with rich relational structure often introduces complex dependencies between co-references. Exploiting these is challenging - it requires seamlessly combining statistical, relational, and logical dependencies. One task of particular interest entity familial networks. In this setting, multiple partial representations a family tree are provided, from the perspective different members, challenge to reconstruct multiple, noisy, views. This reconstruction crucial for...

Recommender systems are used in variety of domains affecting people's lives. This has raised concerns about possible biases and discrimination that such might exacerbate. There two primary kinds inherent recommender systems: observation bias stemming from imbalanced data. Observation exists due to a feedback loop which causes the model learn only predict recommendations similar previous ones. Imbalance data occurs when systematic societal, historical, or other ambient is present In this...

E-commerce applications rely heavily on session-based recommendation algorithms to improve the shopping experience of their customers. Recent progress in shows great promise. However, translating that promise real-world outcomes is a challenging task for several reasons, but mostly due large number and varying characteristics available models. In this paper, we discuss approach lessons learned from process identifying deploying successful algorithm leading e-commerce application...

Recommender systems are an integral part of eCommerce services, helping to optimize revenue and user satisfaction. Bundle recommendation has recently gained attention by the research community since behavioral data supports that users often buy more than one product in a single transaction. In most cases, bundle recommendations form "users who bought A also products B, C, D". Although such can be useful, there is no guarantee A, D may actually related each other. this paper, we address...

Named entity recognition (NER) is a fundamental task for several important applications such as knowledge base construction and semantic search. So far, the focus has been on building machine learning models, which identify generic named entities (e.g., person, date). Such models can be used off-the-shelf without requiring ground truth labels training. However, cannot generalize to specialized domains that have domain-specific legal domain). In these cases, it inevitable generate data...

Motivated from the Context Aware Computing, and more particularly Data-Driven Process Adaptation approach, we propose Semantic Space (SCS) Engine which aims to facilitate provision of adaptable business processes. The SCS provides a space stores semantically annotated data it is open other processes, systems, external sources for information exchange. specified implementation inspired TupleSpace uses JavaSpace Service Jini Framework (changed Apache River lately) as an underlying basis....