In this paper we introduce an approach, called LODQA, for open domain Question Answering over Linked Open Data. We confine ourselves to three kinds of questions: factoid, confirmation, and definition questions. By using LODQA it is feasible to answer questions over 400 millions of entities of any domain without using any training data, since we exploit simultaneously 400 Linked datasets. In particular, we exploit the services of LODsyndesis, a suite of services (based on semantics-aware indexes) which supports cross-dataset reasoning over hundreds of Linked datasets and 2 billion triples. The proposed Question Answering process follows an information extraction approach and comprises several steps including question cleaning, heuristic based question type identification, entity recognition, linking and disambiguation using Linked Data-based methods and pure NLP methods (specifically DBpedia Spotlight and Stanford CoreNLP), WordNet-based question expansion for tackling the lexical gap (between the input question and the underlying sources), and triple scoring for producing the final answer. We discuss the benefits of this approach in terms of answerable questions and answer verification, and we investigate, through experimental results, how the aforementioned steps of the process affect the effectiveness and the efficiency of question answering.
s of Keynote Talks Market Design for Constrained Matching
A huge amount of data is generated each day from various sources. Analysis of these massive data is difficult, and requires new forms of processing to enable enhanced decision making, insight discovery and process optimization. In addition, besides their ever increasing volume, datasets change frequently, and as such, results to continuous queries have to be updated at short intervals. In this paper, we address the problem of evaluating continuous queries over big data streams that are frequently updated, adopting HIFUN, a high-level query language introduced recently. HIFUN offers a clear separation between the conceptual layer, where analytic queries are defined independently of the nature and location of data, and the physical layer where queries are evaluated, by encoding them as map-reduce jobs or as SQL group-by queries. Using HIFUN, we devise an algorithm for incremental processing of continuous queries, processing only the most recent data partition, and exploiting already computed information, without requiring evaluating the query over the complete dataset. Subsequently, we translate the generic algorithm to both SQL and MapReduce using SPARK, exploiting the query rewriting method provided by HIFUN. The experiments performed show the advantages of our solution in terms of query answering efficiency.
This paper focuses first on the formal modeling of complex application scenarios using autonomic proximity-based federation among smart objects with wireless network connectivity, and then on a new framework for complex secure federation scenarios. Our modeling consists of three different levels. In the first-level modeling, each smart object is modeled as a set of ports, each of which represents an I/O interface for a function of this smart object to interoperate with some function of another smart object. The federation between a pair of smart objects having a pair of ports of the same type with opposite polarities is modeled as the port matching between these two ports. The second-level modeling describes the dynamic change of the federation structure among smart objects as a graph rewriting system, where each node and each directed link respectively represent a smart object and a connection between two smart objects. The third-level modeling uses a binary autocatalytic-reaction network to describe each complex federation scenario in which more than one federation are involved, and an output federation of a reaction may work either as an input federation of another reaction and/or a catalyst to activate another composition or decomposition reaction. Based on these models previously proposed by the current author, this paper proposes a new simplified application framework for implementing any complex application scenario describable as a binary autocatalytic-reaction network as a graph rewriting system of smart objects, and then proposes a new framework-level solution to the secure federation of smart objects, which is independent from the encryption-based technologies for secure communication between two smart objects.
A recommender system can be considered as an information filtering system that seeks to predict the preference a user would have for a data item. It is commonly utilized in digital stores to recommend products to their users according to the users' previous purchases. This applies to Steam as well, a widely used digital distribution platform for games. The existing recommender system mainly suggests new games to a given user by calculating similarities between games they own and those that they do not. These similarities are based on predefined attributes (game genres). Additionally, the system is able to recommend games based on the game preferences of the user's friends. In this work, we target at creating an enhanced recommender system for Steam. The goal is to design a hybrid approach for producing suggestions that will utilize data, such as playing time, game price and game release date, in addition to the genres and the preferences of friends.
