Ozer Ozdikis

Semantic Expansion of Tweet Contents for Enhanced Event Detection in Twitter

This paper aims to enhance event detection methods in a micro-blogging platform, namely Twitter. The enhancement technique we propose is based on lexico-semantic expansion of tweet contents while applying document similarity and clustering algorithms. Considering the length limitations and idiosyncratic spelling in Twitter environment, it is possible to take advantage of word similarities and to enrich texts with similar words. The semantic expansion technique we implement is based on syntagmatic and paradigmatic relationships between words, extracted from their co-occurrence statistics. As our technique does not depend on an existing ontology or a lexicon database such as Word Net, it should be applicable for any language. The proposed technique is applied on a tweet set collected for three days from the users in Turkey. The results indicate earlier detection of events and improvements in accuracy.

A platform for agent behavior design and multi agent orchestration

Agents show considerable promise as a new paradigm for software development. However for wider adoption and deployment of agent technology, powerful design and development tools are needed. Such tools should empower software developers to cater agent solutions more efficiently and at a lower cost for their customers with rapidly changing requirements and differing application specifications. In this paper, an agent orchestration platform that allows the developers to design a complete agent-based scenario through graphical user interfaces is presented. The scenario produced by the platform is a rule based system in contrast to the existing systems where agents are coded through a programming language. In this way, the platform provides a higher level of abstraction to agent development making it easier to adapt to rapidly changing user requirements or differing software specifications. The system is highly transportable and interoperable. The platform helps to design a multi-agent system either from scratch, or by adapting existing distributed systems to multi agent systems. It contains tools that handle the agent system design both at the macro level, that is, defining the interaction between agents and at the micro level which deals with internal design of agents. Agent behaviour is modeled as a workflow of basic agent behaviour building blocks (such as receiving a message, invoking an application, making a decision or sending a message) by considering the data and control dependencies among them, and a graphical user interface is provided to construct agent behaviours. The platform allows agent templates to be constructed from previously defined behaviours. Finally through a Scenario Design Tool, a multi-agent system is designed by specifying associations among agents. The scenario is stored in a knowledge base by using the Agent Behaviour Representation Language (ABRL) which is developed for this purpose. Finally to be able to demonstrate the execution of the system on a concrete agent platform, we mapped the ABRL rules to JESS and executed the system on JADE.

Evidential location estimation for events detected in Twitter

Event detection from microblogs and social networks, especially from Twitter, is an active and rich research topic. By grouping similar tweets in clusters, people can extract events and follow the happenings in a community. In this work, we focus on estimating the geographical locations of events that are detected in Twitter. An important novelty of our work is the application of evidential reasoning techniques, namely the Demspter-Shafer Theory (DST), for this problem. By utilizing several features of tweets, we aim to produce belief intervals for a set of possible discrete locations. DST helps us deal with uncertainties, assign belief values to subsets of solutions, and combine pieces of evidence obtained from different tweet features. The initial results on several real cases suggest the applicability and usefulness of DST for the problem.

A survey on location estimation techniques for events detected in Twitter

Detection of events using voluntarily generated content in microblogs has been the objective of numerous recent studies. One essential challenge tackled in these studies is estimating the locations of events. In this paper, we review the state-of-the-art location estimation techniques used in the localization of events detected in microblogs, particularly in Twitter, which is one of the most popular microblogging platforms worldwide. We analyze these techniques with respect to the targeted event type, granularity of estimated locations, location-related features selected as sources of spatial evidence, and the method used to make aggregate decisions based on the extracted evidence. We discuss the strengths and advantages of alternative solutions to various problems related to location estimation, as well as their preconditions and limitations. We examine the most widely used evaluation methods to analyze the accuracy of estimations and present the results reported in the literature. We also discuss our findings and highlight important research challenges that may need further attention.

Evidential estimation of event locations in microblogs using the Dempster-Shafer theory

We estimate locations of events detected in Twitter using Dempster-Shafer theory.Our method combines evidence from multiple tweet features using combination rules.Our method is applicable to any event type and does not require training.Comparisons were made with the Bayesian methods under different settings.Estimations are made for two levels of location granularity with enhanced accuracy. Detecting real-world events by following posts in microblogs has been the motivation of numerous recent studies. In this work, we focus on the spatio-temporal characteristics of events detected in microblogs, and propose a method to estimate their locations using the Dempster-Shafer theory. We utilize three basic location-related features of the posts, namely the latitude-longitude metadata provided by the GPS sensor of the users device, the textual content of the post, and the location attribute in the user profile, as three independent sources of evidence. Considering this evidence in a complementary way, we apply combination rules in the Dempster-Shafer theory to fuse them into a single model, and estimate the whereabouts of a detected event. Locations are treated at two levels of granularity, namely, city and town. Using the Dempster-Shafer theory to solve this problem allows uncertainty and missing data to be tolerated, and estimations to be made for sets of locations in terms of upper and lower probabilities. We demonstrate our solution using public tweets on Twitter posted in Turkey. The experimental evaluations conducted on a wide range of events including earthquakes, sports, weather, and street protests indicate higher success rates than the existing state of the art methods.

Tool support for transformation from an OWL ontology to an HLA Object Model

Designing simulation architectures based on domain models is a promising approach. Tools to support transformation of formalized domain models to design models are essential. Ontology languages offer a way of formally specifying the domain knowledge. We adopt a user-guided approach to model transformation, where the source is an OWL ontology and the target is an HLA Object Model, in particular, a federation object model (FOM). This paper presents a flexible transformation tool that enables the user to define transformations in terms of mappings from OWL constructs to HLA Object Model Template (OMT) constructs. The overall objective is to facilitate ontology-based model-driven development in distributed simulation.

Context Based Semantic Relations in Tweets

Twitter, a popular social networking platform, provides a medium for people to share information and opinions with their followers. In such a medium, a flash event finds an immediate response. However, one concept may be expressed in many different ways. Because of users’ different writing conventions, acronym usages, language differences, and spelling mistakes, there may be variations in the content of postings even if they are about the same event. Analyzing semantic relationships and detecting these variations have several use cases, such as event detection, and making recommendations to users while they are posting tweets. In this work, we apply semantic relationship analysis methods based on term co-occurrences in tweets, and evaluate their effect on detection of daily events from Twitter. The results indicate higher accuracy in clustering, earlier event detection and more refined event clusters.

Towards interoperable and composable trajectory simulations: an ontology-based approach

Trajectory simulation is a software module that computes the flight path and flight parameters of munitions. It is used throughout the engineering process, including simulations for studying the design trade-offs, to mission simulations for defended area analysis. In this wide application domain, reuse has always been one of the challenges of the trajectory simulation community. We apply an ontology-based simulation development methodology to fulfil the functional requirements of a trajectory simulation while targeting reuse through interoperability and composability. Trajectory Simulation ONTology (TSONT) has been constructed as a simulation conceptual model for trajectory simulations. Based on the knowledge captured in TSONT, a domain-oriented reuse methodology has been leveraged to develop HLA-compliant trajectory simulations. A trajectory simulation federate was developed by conforming to the simulation object model based on TSONT. This paper demonstrates our approach to achieve composable and interoperable simulations over a case study in which a trajectory simulation federate serves in a variety of federations that have been constructed.

Confidence-Based incremental classification for objects with limited attributes in vertical search

With vertical search engines, it is possible to search the web pages on a specific domain such as products, restaurants or academic papers and present the users only the interested information. Gathering and integrating such objects from multiple web pages into a single system provides a useful facility for users. Placing the extracted objects from multiple data sources into a single hierarchical structure is a challenging classification problem, especially if there are limited object attributes. In this work, we propose a confidence-based incremental Naive Bayesian approach for categorization, focusing on the product domain. Incremental approach is based on extending the training set and retraining the classifier as new objects are assigned to a category with high confidence. The ordering of product data is taken into account as well. The proposed approach is applied on a vertical search engine that collects product data from several online stores.

Incremental clustering with vector expansion for online event detection in microblogs

Identifying similarities in microblog posts for event detection poses challenges due to short texts with idiosyncratic spellings, irregular writing styles, abbreviations and synonyms. In order to overcome these challenges, we present an enhancement to the incremental clustering techniques by detecting similar terms in microblog posts in a temporal context. We devise an unsupervised method to measure the similarities online using co-occurrence-based techniques and use them in a vector expansion process. The results of our evaluation performed on a tweet set indicate that the proposed vector expansion method helps identify similarities in tweets despite differences in their content. This facilitates the clustering of tweets and detection of events with higher accuracy without incurring a high execution cost.