Yosi Mass

Distributed Trust in Open Multi-agent Systems

Facilitated by the rapid growth of the Internet, electronic commerce is growing exponentially. As a result, millions of players participate in electronic trade, yet many of these players are strangers to each other. This implies mistrust, which may bring about manipulative and malicious trade behaviors among the parties. This problem intensifies in electronic environments where agents act on behalf of humans. There, self-interested, utility-maximizing agents, have a strong motivation, and no moral means against, malicious action. Attempts to prevent such misbehavior usually concentrate on designing non-manipulable mechanisms. Yet, these tend to be either computationally intractable or sub-optimal. We suggest a new approach: a mechanism that allows agents in an open system to establish trust among themselves and to dynamically update this trust. Although we rely on certificates for our solution, we do not require (in contrast to previous solutions) any centralized certificate authority system, nor do we require some well known, trusted parties. Our solution is fully distributed, it is computationally feasible, and can be easily added to any agent architecture.

Extracting user profiles from large scale data

In this work we present the details of a large scale user profiling framework that we developed here in IBM on top of Apache Hadoop. We address the problem of extracting and maintaining a very large number of user profiles from large scale data. We first describe an efficient user profiling framework with high user profiling quality guarantees. We then describe a scalable implementation of the proposed framework in Apache Hadoop and discuss its challenges.

Best-Effort Top-k Query Processing Under Budgetary Constraints

We consider a novel problem of top-k query processing under budget constraints. We provide both a framework and a set of algorithms to address this problem. Existing algorithms for top-k processing are budget-oblivious, i.e., they do not take budget constraints into account when making scheduling decisions, but focus on the performance to compute the final top-k results. Under budget constraints, these algorithms therefore often return results that are a lot worse than the results that can be achieved with a clever, budget-aware scheduling algorithm. This paper introduces novel algorithms for budget-aware top-k processing that produce results that have a significantly higher quality than those of state-of-the-art budget-oblivious solutions.

VRCommerce — electronic commerce in virtual reality

Existing technology and standards allow the creation of threedimensional, virtual-reality browsing experience. Such an interface may be more attractive and natural (at least to some). In particular, with the increase in electronic commerce, the creation of virtual reality stores and shopping malls seems of potential value. However, there are substantial challenges in the use of existing virtual reality tools to create any large space, in particular a store or a shopping mall. The main challenges are the substantial size of the representation of the space (communication and processing overhead), difficulties of navigation using typical UI devices, and support for interconnecting separately designed spaces (stores) into one continuous virtual space (mall). We present an approach to address these challenges, based on limiting the spaces to a modular collection of basic architectural elements such as rooms and hallways. We describe our implementation of virtual-reality, three-dimensional e-commerce, and the VR Commerce toolkit, implemented using standard VRML and Java. VRCommerce is an integrated solution for creation, online operation and navigation in three dimensional malls and stores. It enables continuous navigation between separately designed and managed stores and incremental loading of spaces and objects as needed. Furthermore, VRCommerce offers simplified navigation with less decision making via a two-dimensional `mall directory map` and automated walk modes.

Semantic Documents Relatedness using Concept Graph Representation

We deal with the problem of document representation for the task of measuring semantic relatedness between documents. A document is represented as a compact concept graph where nodes represent concepts extracted from the document through references to entities in a knowledge base such as DBpedia. Edges represent the semantic and structural relationships among the concepts. Several methods are presented to measure the strength of those relationships. Concepts are weighted through the concept graph using closeness centrality measure which reflects their relevance to the aspects of the document. A novel similarity measure between two concept graphs is presented. The similarity measure first represents concepts as continuous vectors by means of neural networks. Second, the continuous vectors are used to accumulate pairwise similarity between pairs of concepts while considering their assigned weights. We evaluate our method on a standard benchmark for document similarity. Our method outperforms state-of-the-art methods including ESA (Explicit Semantic Annotation) while our concept graphs are much smaller than the concept vectors generated by ESA. Moreover, we show that by combining our concept graph with ESA, we obtain an even further improvement.

Just in time indexing for up to the second search

E-commerce and intranet search systems require newly arriving content to be indexed and made available for search within minutes or hours of arrival. Applications such as file system and email search demand even faster turnaround from search systems, requiring new content to become available for search almost instantaneously. However, incrementally updating inverted indices, which are the predominant datastructure used in search engines, is an expensive operation that most systems avoid performing at high rates. We present JiTI, a Just-in-Time Indexing component that allows searching over incoming content (nearly) as soon as that content reaches the system. JiTIs main idea is to invest less in the preprocessing of arriving data, at the expense of a tolerable latency in query response time. It is designed for deployment in search systems that maintain a large main index and that rebuild smaller stop-press indices once or twice an hour. JiTI augments such systems with instant retrieval capabilities over content arriving in between the stop-press builds. A main design point is for JiTI to demand few computational resources, in particular RAM and I/O. Our experiments consisted of injecting several documents and queries per second concurrently into the system over half-hour long periods. We believe that there are search applications for which the combination of the workloads we experimented with and the response times we measured present a viable solution to a pressing problem.

Folksonomy-Based Term Extraction for Word Cloud Generation

In this work we study the task of term extraction for word cloud generation in sparsely tagged domains, in which manual tags are scarce. We present a folksonomy-based term extraction method, called tag-boost, which boosts terms that are frequently used by the public to tag content. Our experiments with tag-boost based term extraction over different domains demonstrate tremendous improvement in word cloud quality, as reflected by the agreement between manual tags of the testing items and the cloud’s terms extracted from the items’ content. Moreover, our results demonstrate the high robustness of this approach, as compared to alternative cloud generation methods that exhibit a high sensitivity to data sparseness. Additionally, we show that tag-boost can be effectively applied even in nontagged domains, by using an external rich folksonomy borrowed from a well-tagged domain.

Folksonomy-based term extraction for word cloud generation

In this work we study the task of term extraction for word cloud generation. We present a folksonomy-based term extraction method, called tag-boost, which boosts terms that are frequently used by the public to tag content. Our experiments with tag-boost-based term extraction over different domains demonstrate tremendous improvement in word cloud quality, as reflected by the agreement between extracted terms and manually assigned tags of the testing items. Additionally, we show that tag-boost can be effectively applied even in non-tagged domains, by using an external rich folksonomy borrowed from a well-tagged domain.

A statistical approach to mining customers' conversational data from social media

In this paper, we present one possible way of analyzing social media conversional data in order to better understand customers. Ultimately, our goal is to analyze customer behavior as it is expressed in free-form conversations and extract from it commercially valuable information about the customer. In this study, we concentrate on using statistical techniques for analyzing this unstructured data at two levels: 1) at the level of the words used in the conversation and 2) by mapping those words to abstract concepts. The goal of such a statistical analysis is twofold. First, the statistically significant terms used by the users and the concepts associated with them provide insight on a users interests that commercial services can use, for example, in order to target advertisements. In addition, knowing the evolution of a customers interests and hobbies can be exploited commercially by retailers, media and entertainment companies, telecommunications companies, and more. In this paper, we describe a general framework for the analysis of social media data and, in turn, the application of the framework to the statistical analysis of the language of tweets.

Multimedia information retrieval: "new challenges in audio visual search"

With the rising popularity of rich media services such as Flickr, YouTube, and Jumpcut, new challenges in large scale multimedia information retrieval have emerged that not only rely on meta-data but on content-based information retrieval combined with the collective knowledge of users and geo-referenced meta-data that is captured during the creation process. For the future, it is envisioned that multimedia search in mobile environments or on P2P networks will take off on a large scale.