Attaching any kind of clue – event, location, person, tag or keyword – to a photo eases the process of searching. Often the problem is that the user feels that it is difficult to think of good tags or that the tagging process is too tedious or cumbersome. At the same time, users use social media daily, and write about topics they feel are important and that they are actively interested in. This paper presents a method for extracting metadata (tag suggestions) from social media profiles and illustrates the use of the tags for photo tagging by means of a web-based photo application.
In this paper, we propose evaluation of customer journey for contents of Owned Media. In recent years, many companies publish the Owned Media in order to brand its products and services. The Owned Media is useful for provision of novel information correctly and rapidly. On these backgrounds, there is the demand of evaluation for effectiveness of the Owned Media. It is necessary to recognize which detailed content of Owned Media has importance. Our proposed method is able to produce attractive contents through appreciating which contents make approaches to customers. We demonstrate the evaluation using a certain Web-site as Owned Media and show the effectiveness of our methods.
Organizations in the publishing field accumulate large amounts of data and store it in different forms of knowledge databases. In the publishing field specifically, various multimedia content presentation is involved, requiring support for contemporary e-book standard formats (e-Pub) as well as traditional standard formats such as HTML, PDF, and others. Publishing organizations must support the same content presentation in various formats in order to meet the needs of their increasingly demanding clients. However, business (and organizational) processes in publishing in the presented case study have little IT support and almost no automation. Successful business flow and a competitive advantage for publishing organizations, especially for SMEs, requires a holistic solution for the optimization of publishing processes, including data/knowledge acquisition, data/knowledge aggregation and the creation of new knowledge based on existing data. Although partial IT solutions are available and can be used to support individual steps in the publishing process, they do not come in an out-of-the box solutions, suitable for adequate publishing processes automation and optimization. Two basic challenges regarding existing solutions available on the market are: (1) they do not provide support for automation or semi automation in the publishing processes, and (2) they are expensive and consequently not acceptable for SMEs. This paper is focused on analyzing publishing work processes and its quality aspects, addressing the automation of all possible process steps: especially the optimization of acquisition, aggregation and building of multimedia content for any device. In addition, a solution built on open source components is proposed as well, supporting the presentation of content in different published formats. Based on a case study, we present the solution's architectural design and analyze standards and technologies, providing an acceptable solution for SME's from an economical point of view, optimizing its work processes to the largest possible extent. A renewed publishing process is proposed and future research directions are also discussed.
This paper presents results of classification on imbalanced data with ensemble allocation method. The result of the allocation method were compared to traditional techniques for dealing with imbalaced datasets – the sampling methods. The allocation method is a two level ensemble that combines unsupervised and supervised learning. In this research the first level of allocation the unsupervised anomaly detection is used as an allocator which is combined with several traditional classification method on second level of ensemble. The allocation method is tested on imbalanced datasets and the results are compared to two well used sampling methods – under-sampling of majority instances, and over-sampling with SMOTE which introduces new artificial instances of minority class to the dataset. Results of all of the methods were compared on overall accuracy and average F-score metrics. The results show that allocation method produces the best classification model, which is also supported by statistical analysis.
The aim of this research study is to explore the forecasting technique to fit to the existing waste data of a tape converting production factory using a time series method. Optimal type of time series techniques to minimise the error between actual and forecasted data was chosen after comparing three types of time series techniques. The data analysis, based on the accuracy of their outputs, resulted in the “Double Exponential Smoothing” being the preferable choice since the error was less than that for other techniques. The projected forecasted amounts were 13.09, 11.08, 11.77 and 10.25 (Unit of measurement is 10,000 kilograms) for January, February, March and April 2015. After benchmarking error percentages (MAPE) that against similar techniques and problems from other references,the error in this study (18%) was less than the benchmarking source (21%  and 33%  respectively). Therefore this technique is more accurate than the benchmarking techniques by 17% and 83% respectively. After rechecking the actual data with forecast time series data, the average MAPE was found to be around 15% which is still lower that the errors quoted in other reference papers.
Intensional epistemic logics are not apt for handling properly the specification of communication and reasoning of resource-bounded agents in a multi-agent system. They oscillate between two unrealistic extremes: either the explicit knowledge of an ‘idiot’ agent, deprived of any inferential capabilities, or the implicit knowledge of an agent who is a logical/mathematical genius. The goal of this paper is to introduce the notion of inferable knowledge of a rational yet resource-bounded agent. The stock of inferable knowledge of such an agent a is the closure of a chain-of-knowledge sequence validly derivable from a's existing stock of explicit knowledge via one or more rules of inference that a masters. We are using Pavel Tichý's Transparent Intensional Logic as our framework. This logic models knowing as a relation-in-intension between an agent and a construction (a hyperintensional mode of presentation of a possible-world proposition) rather than a set of possible worlds or a piece of syntax. We motivate the restriction of the epistemic closure principle to inferable knowledge, present the theoretical framework, define the concept of inferable knowledge, and explain the technicalities of the so restricted closure principle.
Here are considered information and information growth in interactions of Information Processing Systems (IPS). Information does not exist ‘per se’ – it is always stored in some Information Processing System (living system, social system, business system, administrative/government system etc.). All IPS are finite and can be modelled as Finite State Machines (FSM) with memory. They are connected with each other and interact – exchange messages. Their messages (responses to input queries) reveal to others information about their functioning, thus IPS with more memory learn behaviour, i.e. infer information stored in other, smaller IPS. Thus information in network of connected IPS-s is accumulating in IPS with more memory and smaller IPS become parts of all greater and greater ‘super’ IPS – IPS on the next level of IPS development hierarchy. The whole human society is currently moving into new era – the era of networked Super Information Processing Systems, where main value is information, not material things. The material goods based exchange value – money is replaced by digital currencies, which are based on computer-generated informational structures. But the rules and practices to handle all the time growing informational values are still from era of material values, we do not yet have culture for information era, thus malware and information theft are growing exponentially and this may essentially change our ownership and market-based societal organization.
Each discipline, to some extent, has its own concise and precise vocabulary used to describe unambiguously the special concepts within the domain and the relationships bounding them. In the medical science for instance, doctors use specialized vocabulary and knowledge for an effective and efficient way of (1) communication, like filling in an EHR (Electronic Health Record), and (2) for problem solving, like the diagnosis process. For those who are unfamiliar with that vocabulary, it is hard for them to express relevant information like describing symptoms to a doctor in order to get diagnosed. In our work, we aim at bringing the common sense knowledge and the basic vocabulary closer to the expertise knowledge for an effective communication between what we call a layperson and an expert illustrated with a case of patient/doctor communication. In this paper, we define a communication process, pointing out the beneficial use of the expertise knowledge and the choice for the Ontology-Driven modelling that will enhance the notion of progressivity in the knowledge acquisition process. We also define our cyclic acquisition process step by step starting from the first and foremost step of processing the messages to the information extraction and the reasoning process until the last but not least step of outputting the ontological representation.
This paper proposes a scheme of detecting topic evolutions in bibliographic databases. There have been a lot of scientific bibliographies, such as DBLP, CiteSeerX, MEDLINE/PubMed, ADS, arXiv, etc., and hence it has been extremely important to extract useful information from these databases. It should be noticed that, in such databases, citations play crucial role to represent relationships among different publications. To make the best use of citation information as well as textual features for extracting topic evolutions in a bibliographic database, we propose a scheme based on non-negative matrix factorization (NMF). More precisely, we first partition the set of publications in a database according to their publication years, and apply NMF to extract clusters of publications. Notice that we take into account citation information to perform NMF for better clustering. Having obtained sets of publications for each time span, we associate similar clusters in consecutive time spans according to their similarity. Thus we can obtain time evolution of topics and clusters of publications. In the experiments we demonstrate the proposed scheme can successfully extract topic evolutions in real bibliographic databases, CiteSeerX and arXiv.