Search

We study efficient full implementation via transfers in unique rationalizable strategies in environments that are symmetric in two senses: first, agents display the same total level of preference interdependence; second, types are commonly known to be drawn from distributions with identical (but unknown) means. We characterize the conditions under which full efficient implementation is possible via direct mechanisms, as well as the transfer schemes that achieve it whenever possible. We discuss a further robustness property--robustness to mistaken play--and show that it uniquely selects the transfer scheme that induces an even redistribution of strategic externalities.

Payments in online ad auctions are typically derived from click-through rates, so that advertisers do not pay for ineffective ads. But advertisers often care about more than just clicks. That is, for example, if they aim to raise brand awareness or visibility. There is thus an opportunity to devise a more effective ad pricing paradigm, in which ads are paid only if they are actually noticed. This article contributes a novel auction format based on a pay-per-attention (PPA) scheme. We show that the PPA auction inherits the desirable properties (strategy-proofness and efficiency) as its pay-per-impression and pay-per-click counterparts, and that it also compares favourably in terms of revenues. To make the PPA format feasible, we also contribute a scalable diagnostic technology to predict user attention to ads in sponsored search using raw mouse cursor coordinates only, regardless of the page content and structure. We use the user attention predictions in numerical simulations to evaluate the PPA auction scheme. Our results show that, in relevant economic settings, the PPA revenues would be strictly higher than the existing auction payment schemes.

Document clustering is a powerful method with numerous applications, where the core idea is to group text into smaller and more manageable pieces of semantically related information. While there has been progress in the research community on improving the quality of clusters, the extraction of information that explains the semantic content of the clusters is still mostly a manual activity, as it requires the inspection of sample documents. In this work, we propose three main cluster explanation approaches that extend the current state of the art, which is mostly based on predicting a label for each cluster (i.e. cluster labelling). The first approach is based on scores extracted from the word distributions; the second is based on augmenting the score-based explanations using external knowledge-bases, and the third one combines the first two methods to exploit the knowledge coming from a labelled dataset. We also discuss the limitations of the current metric presented in the literature and extend it to evaluate our approaches using ground truth. We also provided an extensive set of evaluations to verify the effectiveness of the proposed algorithms, by running multiple experiments to compare our approaches with different baselines. Results indicate that improved explanations are possible and that linking external knowledge can provide more general cluster descriptions. Semi-supervised approaches also demonstrated meaningful insight. User studies were also conducted highlighting that users prefer specialized explanations.

We introduce a framework to study the problem of full implementation under general restrictions on agents’beliefs, which we call �-Implementation. First we provide a characterization of the properties of the mechanisms that achieve �-Implementation. These conditions relate the possibility of achieving full implementation to the strength of the strategic externalities induced by the mechanism. We then study how to achieve full implementation via the design of simple transfer schemes. In these context, the general conditions for �-Implementation suggest a simple design principle, in which belief restrictions are used to weaken the strategic externalities induced by the baseline transfers of the belief-free benchmark. Importantly, our results require minimal information on agents’beliefs, which regard simple moments of the distribution of types. These moment conditions arise naturally in applications and in practical problems of mechanism design.

Digital forensics departments usually have to analyse vast amounts of audio-visual content, such as videos collected from street CCTV footage, hard drives or online resources. The framework presented in this article, which has been developed in the context of the EU-funded project LASIE, aims to assist investigators in their everyday tasks, through the provision of innovative tools for image and video analysis, object detection and tracking and event detection. These tools exploit the latest advances in machine learning, including deep neural networks, to handle the challenges in processing content from real-world data sources. The framework is enhanced with advanced inference and recommendation capabilities, which filter-out inconsistencies and recommend additional evidence. An intuitive user interface allows exploiting the capabilities of the available tools in a user-friendly manner. The framework supports distributed processing, with easy deployment of the services in clusters of multiple workstations, making the proposed solution appropriate for big data analytics tasks.

We introduce a model of strategic thinking in games of initial response. Unlike standard models of strategic thinking, in this framework the player’s ‘depth of reasoning’ is endogenously determined, and it can be disentangled from his beliefs over his opponent’s cognitive bound. In our approach, individuals act as if they follow a cost-benet analysis. The depth of reasoning is a function of the player’s cognitive abilities and his payos. The costs are exogenous and represent the game theoretical sophistication of the player; the benet instead is related to the game payos. Behavior is in turn determined by the individual’s depth of reasoning and his beliefs about the reasoning process of the opponent. Thus, in our framework, payos not only aect individual choices in the traditional sense, but they also shape the cognitive process itself. Our model delivers testable implications on players’ chosen actions as incentives and opponents change. We then test the model’s predictions with an experiment. We administer dierent treatments that vary beliefs over payos and opponents, as well as beliefs over opponents’ beliefs. The results of this experiment, which are not accounted for by current models of reasoning in games, strongly support our theory. We also show that the predictions of our model are highly consistent, both qualitatively and quantitatively, with well-known unresolved empirical puzzles. Our approach therefore serves as a novel, unifying framework of strategic thinking that allows for predictions across games.

Italy’s Cultural Heritage is the world’s most diverse and rich patrimony and attracts millions of visitors every year to monuments, archaeological sites and museums. The valorization of cultural heritage represents nowadays one of the most important research challenges in the Italian scenario. In this paper, we present a general multimedia recommender system able to uniformly manage heterogeneous multimedia data and to provide context-aware recommendation techniques supporting intelligent multimedia services for the users. A specific application of our system within the cultural heritage domain is proposed by means of a real case study in the mobile environment related to an outdoor scenario, together with preliminary results on user’s satisfaction.

Weinstein and Yildiz (2007) show that only very weak predictions are robust to misspecifications of higher order beliefs. Whenever a type has multiple rationalizable actions, any of these actions is uniquely rationalizable for some arbitrarily close type. Hence, refinements of rationalizability are not robust. This negative result is obtained under a richness condition, which essentially means that all common knowledge assumptions on payoffs are relaxed. In many settings, this condition entails an unnecessarily demanding robustness test. It is, therefore, natural to explore the structure of rationalizability when arbitrary common knowledge assumptions are relaxed (i.e., without assuming richness). For arbitrary spaces of uncertainty and for every player i, I construct a set A ∞ of actions that are uniquely rationalizable for some hierarchy of beliefs. The main result shows that for any type ti and any action ai rationalizable for ti ,i fai belongs to A ∞ and is justified by conjectures concentrated on A ∞ , then there exists a sequence of types converging to ti for which ai is uniquely rationalizable. This result significantly generalizes Weinstein and Yildiz’s. Some of its implications are discussed in the context of auctions and equilibrium refinements, and in connection with the literature on global games.

The valorization and promotion of worldwide Cultural Heritage by the adoption of Information and Communication Technologies represent nowadays some of the most important research issues with a large variety of potential applications. This challenge is particularly perceived in the Italian scenario, where the artistic patrimony is one of the most diverse and rich of the world, able to attract millions of visitors every year to monuments, archaeological sites and museums. In this paper, we present a general recommendation framework able to uniformly manage heterogeneous multimedia data coming from several web repositories and to provide context-aware recommendation techniques supporting intelligent multimedia services for the users--i.e. dynamic visiting paths for a given environment. Specific applications of our system within the cultural heritage domain are proposed by means of real case studies in the mobile environment related both to an outdoor and indoor scenario, together with some results on user's satisfaction and system accuracy.

Weinstein and Yildiz (2007) have shown that in static games, only very weak predictions are robust to perturbations of higher order beliefs. These predictions are precisely those provided by interim correlated rationalizability (ICR). This negative result is obtained under the assumption that agents have no information on payoffs. This assumption is unnatural in many settings. It is therefore natural to ask whether Weinstein and Yildiz's results remain true under more general information structures. This paper characterizes the “robust predictions” in static and dynamic games, under arbitrary information structures. This characterization is provided by an extensive form solution concept: interim sequential rationalizability (ISR). In static games, ISR coincides with ICR and does not depend on the assumptions on agents' information. Hence the “no information” assumption entails no loss of generality in these settings. This is not the case in dynamic games, where ISR refines ICR and depends on the details of the information structure. In these settings, the robust predictions depend on the assumptions on agents' information. This reveals a hitherto neglected interaction between information and higher order uncertainty, raising novel questions of robustness.

A class of bargaining games in which agents bargain over prices and maximum trading constraints is considered: It is proved that all the Stationary Subgame Perfect Equilibria of these games implement Walrasian allocations as the bargaining frictions vanish. The result holds for any number of agents and is robust to different specifications of the bargaining process.

We present a multimedia summarizer system for retrieving relevant information from some web repositories based on the extraction of semantic descriptors of documents. In particular, semantics attached to each document textual sentences is expressed as a set of assertions in the \(\langle subject,verb,object \rangle \) shape as in the RDF data model. While, images’ semantics is captured using a set of keywords derived from high level information such as the related title, description and tags. We leverage an unsupervised clustering algorithm exploiting the notion of semantic similarity and use the centroids of clusters to determine the most significant summary sentences. At the same time, several images are attached to each cluster on the base of keywords’ term frequency. Finally, several experiments are presented and discussed.

In the last decade, the spread of broadband Internet connections even for mobile devices has contributed to an increased availability of multimedia information on the Web. At the same time, due to the decrease of storage cost and the increasing popularity of storage services in the cloud, the problem of information overload has become extremely serious even in personal/company archives. The need to manage, retrieve, and present all these data has promoted the development of advanced multimedia information systems, which include recommendation modules to account for the requests of personalized data selection and presentation. Recommendation systems estimate ratings, or utilities, which quantify users’ degree of interest for the different available data, so that the data can be offered to the users in a personalized way, in decreasing order of interest. Multiple approaches have been proposed in the literature to estimate such degrees of interest. In content-based filtering, the utility (for a user) of a given item is estimated as a function of the ratings given by the same user to other similar items. A dual approach is collaborative filtering, in which filtering (i.e., estimating the object’s utilities) for a given user is computed by referring to the opinions of other similar users. Thus, a major challenge faced by collaborative filtering is the need to associate each user to a set of other users having similar profiles. In this chapter, we first present the co-clustering-based recommendation techniques, which allow to combine heterogeneous multimedia content information and data about the users’ preferences and rankings, thus overcoming some of the content-based filtering drawbacks, as well as some collaborative filtering weaknesses. Then, we briefly discuss the challenges in multimedia delivery and the most common strategies adopted in the context of cultural heritage media delivery.

On the grounds, ontologies have been shown to be a powerful resource for the interpretation and translation of the terminological and semantic relationships within domains of interest but it is still unclear how they can be applied in the context of multimedia data. In this paper, we describe a framework which can capture and manage semantic information related to the multimedia data by modeling in the ontology their features. In particular, the proposed ontology-based framework is organized in the following way: at the lower levels, spatial objects, colors, shapes are represented, and semantic relationships can be established among them; at the higher levels, objects with semantic properties are put into relationship among themselves as well as with the corresponding low-level objects. On this basis, we have designed an ontological system particularly suitable for image retrieval. We have also taken into account the inherent uncertainty related to the representation and detection of multimedia properties in this complex domain. Along this work, we have provided examples from the image domain; moreover, since ontologies provide a semantic means for the semantic comparison of objects and relationships across different formats, the system is easily extensible to other, heterogeneous data sources.

In this paper, we present a general framework for retrieving relevant information from news papers that exploits a novel summarization algorithm based on a deep semantic analysis of texts. In particular, we extract from each Web document a set of triples (subject, predicate, object) that are then used to build a summary through an unsupervised clustering algorithm exploiting the notion of semantic similarity. Finally, we leverage the centroids of clusters to determine the most significant summary sentences using some heuristics. Several experiments are carried out using the standard DUC methodology and ROUGE software and show how the proposed method outperforms several summarizer systems in terms of recall and readability.

Social Networking Sites, such as Facebook and Linkedin, are clear examples of the impact that the Web 2.0 has on people around the world, because they target an aspect of life that is extremely important to anyone: social relationships. The key to building a social network is the ability of finding people that we know in real life, which, in turn, requires those people to make publicly available some personal information, such as their names, family names, locations and birth dates, just to name a few. However, it is not uncommon that individuals create multiple profiles in several social networks, each containing partially overlapping sets of personal information. Matching those different profiles allows to create a global profile that gives a holistic view of the information of an individual. In this paper, we present an algorithm that uses the network topology and the publicly available personal information to iteratively match profiles across n social networks, based on those individuals who disclose the links to their multiple profiles. The evaluation results, obtained on a real dataset composed of around 2 million profiles, show that our algorithm achieves a high accuracy.

The massively distributed publication of linked data has brought to the attention of scientific community the limitations of classic methods for achieving data integration and the opportunities of pushing the boundaries of the field by experimenting this collective enterprise that is the linking open data cloud. While reusing existing ontologies is the choice of preference, the exploitation of ontology alignments still is a required step for easing the burden of integrating heterogeneous data sets. Alignments, even between the most used vocabularies, is still poorly supported in systems nowadays whereas links between instances are the most widely used means for bridging the gap between different data sets. We provide in this paper an account of our statistical and qualitative analysis of the network of instance level equivalences in the Linking Open Data Cloud (i.e. the sameAs network) in order to automatically compute alignments at the conceptual level. Moreover, we explore the effect of ontological information when adopting classical Jaccard methods to the ontology alignment task. Automating such task will allow in fact to achieve a clearer conceptual description of the data at the cloud level, while improving the level of integration between datasets.

We study the problem of efficient auction design in environments with interdependent values, under arbitrary common knowledge assumptions. We propose a simple mechanism and show that, under a rather mild condition, it "robustly" achieves efficiency. Our mechanism consists in a standard Vickrey auction, preceded by one round of communication, where agents report their private signals and receive transfers from the designer. We interpret the transfers as the cost for the designer to robustly achieve efficiency. We introduce a notion of robust informational size and show that the transfers are small if agents are informationally small in our sense. Furthermore, the transfers are decreasing in the amount of information available to the designer and in the strength of the common knowledge assumptions. In other words, the more robust the efficient implementation result, the higher the cost of achieving efficiency. We thus formalize the intuitive idea of a trade-off between robustness and efficient implementation and analyze the determinants of the "cost of robustness".

Seeking bits of useful information from a large amount of data on the Web still remains a difficult and time consuming task for a wide range of people such as students, reporters, and many other types of professionals. This problem requires to investigate new ways to handle and process information, that has to be delivered in a rather small space, retrieved in a short time, and represented as accurately as possible. This is surely one of the most important reasons for searching suitable and efficient summarization techniques capable of "distilling" the most important information from a variety of logically related sources, as the one returned from classic search engines, in order to produce a short, concise and grammatically meaningful version of information spread out in pages and pages of texts. In this paper we present a summarizer system, named iWIN (information on the Web In a Nutshell), that is able to perform an automatic summarization of multiple documents through: a semantic analysis of the text, a ranking method used to evaluate the relevance of the information for the specific user, a clustering method based on the document representation in terms of set of triplets (subject, verb, object) and a sentences’ selection/ordering process to make the final summary as much readable as possible. Some preliminary results about system performances obtained using the ROUGE evaluation software are presented and discussed.

The World Wide Web serves as a valuable source of culture-relevant information, which can be used to support cultural modeling and analysis activities. Part of the challenge in exploiting the Web as a source of culture-relevant information relates to the need to detect and extract information about beliefs, attitudes, and values from a variety of different resources. The Web thus features a rich variety of information resources, and these are seldom categorized with respect to the dimensions in which cultural analysts are interested. Exploiting the Web as a source of culture-relevant information therefore requires techniques and approaches that enable cultural analysts to extract relevant information and organize extracted content in various ways. In this paper, we outline an approach to assist cultural analysts in the extraction and organization of relevant information. We show techniques that can be used to extract information about the attitudes, beliefs, and values of individuals, and how this data can, in turn, be used to support cultural modeling and analysis.