Human-centricity is the core value behind the evolution of manufacturing towards Industry 5.0. Nevertheless, there is a lack of architecture that considers safety, trustworthiness, and human-centricity at its core. Therefore, we propose an architecture that integrates Artificial Intelligence (Active Learning, Forecasting, Explainable Artificial Intelligence), simulated reality, decision-making, and users' feedback, focussing on synergies between humans and machines. Furthermore, we align the proposed architecture with the Big Data Value Association Reference Architecture Model. Finally, we validate it on three use cases from real-world case studies.

Industry 4.0 aims to optimize the manufacturing environment by leveraging new technological advances, such as new sensing capabilities and artificial intelligence. The DRAEM technique has shown state-of-the-art performance for unsupervised classification. The ability to create anomaly maps highlighting areas where defects probably lie can be leveraged to provide cues to supervised classification models and enhance their performance. Our research shows that the best performance is achieved when training a defect detection model by providing an image and the corresponding anomaly map as input. Furthermore, such a setting provides consistent performance when framing the defect detection as a binary or multiclass classification problem and is not affected by class balancing policies. We performed the experiments on three datasets with real-world data provided by Philips Consumer Lifestyle BV.

Quality control is a crucial activity performed by manufacturing companies to ensure their products conform to the requirements and specifications. The introduction of artificial intelligence models enables to automate the visual quality inspection, speeding up the inspection process and ensuring all products are evaluated under the same criteria. In this research, we compare supervised and unsupervised defect detection techniques and explore data augmentation techniques to mitigate the data imbalance in the context of automated visual inspection. Furthermore, we use Generative Adversarial Networks for data augmentation to enhance the classifiers' discriminative performance. Our results show that state-of-the-art unsupervised defect detection does not match the performance of supervised models but can be used to reduce the labeling workload by more than 50%. Furthermore, the best classification performance was achieved considering GAN-based data generation with AUC ROC scores equal to or higher than 0,9898, even when increasing the dataset imbalance by leaving only 25\% of the images denoting defective products. We performed the research with real-world data provided by Philips Consumer Lifestyle BV.

Active learning is a subfield of machine learning that studies how to identify data instances that contribute most to the learning of a given learner and requests some oracle to provide complementary information to enhance the learner's learning process towards a specific goal. This paper proposes a novel active learning strategy for image classification tasks. Furthermore, we obtain insights that lay the ground for developing a novel early-stopping criteria to maximize the model's learning while reducing the required instances. The proposed active learning technique uses insights obtained via GradCAM activation maps to understand the cognition process triggered by each image in the learner and select those that trigger the most diverse patterns, presupposing that diverse GradCAM patterns point to new learning opportunities. To our knowledge, the approach is among the first to use insights from explainable artificial intelligence to drive data selection and annotation. Nevertheless, our results show that it does not achieve the same quality of results as the random and uncertainty sampling techniques we compared to. More research is required to understand how to enhance the models' performance with such a strategy.

The Workshop Program of the Association for the Advancement of Artificial Intelligence's Thirtieth AAAI Conference on Artificial Intelligence (AAAI‐16) was held at the beginning of the conference, February 12–13, 2016. Workshop participants met and discussed issues with a selected focus, and the workshop provided an informal setting for active exchange among researchers, developers, and users on topics of current interest. The AAAI‐16 workshops were an excellent forum for exploring emerging approaches and task areas, for bridging the gaps between AI and other fields or between subfields of AI, for elucidating the results of exploratory research, or for critiquing existing approaches. The 15 workshops held at AAAI‐16 were Artificial Intelligence Applied to Assistive Technologies and Smart Environments (WS‐16‐01), AI, Ethics, and Society (WS‐16‐02), Artificial Intelligence for Cyber Security (WS‐16‐03), Artificial Intelligence for Smart Grids and Smart Buildings (WS‐16‐04), Beyond NP (WS‐16‐05), Computer Poker and Imperfect Information Games (WS‐16‐06), Declarative Learning Based Programming (WS‐16‐07), Expanding the Boundaries of Health Informatics Using AI (WS‐16‐08), Incentives and Trust in Electronic Communities (WS‐16‐09), Knowledge Extraction from Text (WS‐16‐10), Multiagent Interaction Without Prior Coordination (WS‐16‐11), Planning for Hybrid Systems (WS‐16‐12), Scholarly Big Data: AI Perspectives, Challenges, and Ideas (WS‐16‐13), Symbiotic Cognitive Systems (WS‐16‐14), and World Wide Web and Population Health Intelligence (WS‐16‐15).

This paper presents an approach for recommending news articles on a large news portal. Focus is given to interpretability of the developed models, analysis of their performance, and deriving understanding of short and long-term user behavior on a news portal.

This demo presents a web application which implements a pipeline for searching and browsing through newspaper archives. It uses a combination of information extraction, enrichment and visualization algorithms to help users to grasp a large amount of articles normally collected in archives. Illustrative results show appropriateness of the proposed pipeline for searching and browsing news archives.

We describe the development of a predictive model for vehicle journey time on highways. Accurate travel time prediction is an important problem since it enables planning of cost effective vehicle routes and departure times, with the aim of saving time and fuel while reducing pollution. The main information source used is data from roadside double inductive loop sensors which measure vehicle speed, flow and density at specific locations. We model the spatiotemporal distribution of travel times by using local linear regression. The use of real-time data is very accurate for shorter journeys starting now and less reliable as journey times increase. Local linear regression can be used to optimally balance the use of historical and real time data. The main contribution of the paper is the extension of local linear models with higher order autoregressive travel time variables, namely vehicle flow data, and density data. Using two years of UK Highways Agency (HA) loop sensor data we found that the extended model significantly improves predictive performance while retaining the main benefits of earlier work: interpretability of linear models as well as computationally simple predictions.

QMiner is an open source analytics platform for performing large scale data analysis written in C++ and exposed via a Javascript API. The paper presents five main design elements which focus on focus on storage, online and real-time processing as well as fast prototyping and give QMiner unique advantages as a data analytics platform for processing streams of structured and unstructured data. These design elements are incorporated in a five layer architecture represented by 1) storage and indexing, 2) stream aggregate, 3) feature extractor, 4) linear algebra and 5) analytics layers. The functionality of the platform is demonstrated by three representative usage examples containing code fragments: text classification, time series prediction and community detection in graphs.

The amount of structured data is growing rapidly. Given a structured query that asks for some entities, the number of matching candidate results is often very high. The problem of ranking these results has gained attention. Because results in this setting equally and perfectly match the query, existing ranking approaches often use features that are independent of the query. A popular one is based on the notion of centrality that is derived via PageRank. In this paper, we adopt learning to rank approach to this structured query setting, provide a systematic categorization of query-independent features that can be used for that, and finally, discuss how to leverage information in access logs to automatically derive the training data needed for learning. In experiments using real-world datasets and human evaluation based on crowd sourcing, we show the superior performance of our approach over two relevant baselines.

We built a pipeline for translating text into logical representation, which falls in the category of Machine Reading (MR). The essential component of the pipeline and our main contribution is the non-probabilistic rule-based framework. Other components are: syntatic parser, XLE, to extract gramamtical features from text, Enrycher, for additional natural language processing, and Cyc for semantic resources, reasoning, and its language CycL to respresent the translated knowledge. We defined and implemented several rules on the framework. We evaluated them on business news. In the discussion we identified several challenges in MR. Prevajanje novic v jezik CycL s pomočjo razčlenjevalnika XLE Zgradili smo cevovod za prevajanje besdila v logično predstavitev. Naše delo spada v področje strojnega branja. Glavna komponenta cevovoda in naš glavni prispevek je neprobabilistično programsko ogrodje, ki temelji na pravilih. Ostale komponente so: XLE – sintaktični razčlenjevalnik, Enrycher – storitev za procesiranje naravenega jezika in Cyc – semantični vir ter avtomatski pojasnjevalnik. Za predstavitev prevedenega znanja smo uporabili jezik CycL. Na programskem ogrodju smo definirali in implementirali nekaj pravil. Ocenili smo, kako prevajajo besedila iz poslovnih novic. V zaključku smo odkrili številne izzive v strojnem branju.

In this paper, we describe Videk - a physical mashup which uses artificial intelligence technology. We make an analogy between human senses and sensors; and between human brain and artificial intelligence technology respectively. This analogy leads to the concept of Global Oracle. We introduce a mashup system which automatically collects data from sensors. The data is processed and stored by SenseStream while the meta-data is fed into ResearchCyc. SenseStream indexes aggregates, performs clustering and learns rules which then it exports as RuleML. ResearchCyc performs logical inference on the meta-data and transliterates logical sentences. The GUI mashes up sensor data with SenseStream output, ResearchCyc output and other external data sources: GoogleMaps, Geonames, Wikipedia and Panoramio.

The aim of this tutorial is to present an overview of text stream processing starting with a description and properties of text streams, and continuing with a series of text processing techniques and their applicability to text streams. Among the text processing techniques we are going to describe entity extraction and resolution, event and fact extraction, word sense disambiguation, sentiment analysis, summarization, social network analysis, all in the context of text streams. The goal is to present the list of problems and challenges arising when processing text streams and to show how they can be approached using text mining, natural language processing and semantic analysis techniques and tools. The tutorial will describe available approaches and show some demos on text data streams, using publicly available tools.

The paper proposes a platform for integration, reasoning and planning over the Personal Information Management data. It aims at simplifying day-to-day tasks by means of automating some of the steps related to planning activities depending on different data sources that need to be integrated. Usage of semantic technologies enables automatically providing suggestions by means of reasoning. We present the underlying architecture and explain an application scenario for complex travel planning supported by the proposed platform.

Service oriented access in a multi-application, multi-access network environment is faced with the problem of cross-layer interoperability among technologies. In this demo, we present a knowledge base (KB) which contains local (user terminal specific) knowledge that enables pro-active network selection by translating technology specific parameters to higher-level, more abstract parameters. We implemented a prototype which makes use of semantic technology (namely ResearchCyc) for creating the elements of the KB and uses reasoning to determine the best access network. The system implements technology-specific parameter mapping according to the IEEE 802.21 draft standard recommendation.

In this paper, we present a technique for visual analysis of documents based on the semantic representation of text in the form of a directed graph, referred to as semantic graph. This approach can aid data mining tasks, such as exploratory data analysis, data description and summarization. In order to derive the semantic graph, we take advantage of natural language processing, and carry out a series of operations comprising a pipeline, as follows. Firstly, named entities are identified and co-reference resolution is performed; moreover, pronominal anaphors are resolved for a subset of pronouns. Secondly, subject -- predicate -- object triplets are automatically extracted from the Penn Treebank parse tree obtained for each sentence in the document. The triplets are further enhanced by linking them to their corresponding co-referenced named entity, as well as attaching the associated WordNet synset, where available. Thus we obtain a semantic directed graph composed of connected triplets. The document's semantic graph is a starting point for automatically generating the document summary. The model for summary generation is obtained by machine learning, where the features are extracted from the semantic graph structure and content. The summary also has an associated semantic representation. The size of the semantic graph, as well as the summary length can be manually adjusted for an enhanced visual analysis. We also show how to employ the proposed technique for the Visual Analytics challenge.

In this paper we present a semi-automatic ontology editor as implemented in a new version of OntoGen system. The system integrates machine learning and text mining algorithms into an efficient user interface lowering the entry barrier for users who are not professional ontology engineers. The main features of the systems include unsupervised and supervised methods for concept suggestion and concept naming, as well as ontology and concept visualization. The system was tested in extensive user trails and in several real-world scenarios with very positive results.