Paras Mehta

Continuous Summarization of Streaming Spatio-Textual Posts

In this paper, we address the problem of continuously maintaining a concise, diversified summary of the contents of a sliding window over a stream of geotagged posts. Selecting posts to include in the summary takes into account both the criteria of coverage and diversity, and the summary is updated dynamically when the window slides. Our proposed strategy provides a trade-off between information quality and performance. An experimental evaluation of our method is presented using two real-world datasets containing spatio-textual posts from Twitter and Flickr.

Coverage and diversity aware top-k query for spatio-temporal posts

Large amounts of user-generated content are posted daily on the Web, including textual, spatial and temporal information. Exploiting this content to detect, analyze and monitor events and topics that have a potentially large span in space and time requires efficient retrieval and ranking based on criteria including all three dimensions. In this paper, we introduce a novel type of spatial-temporal-keyword query that combines keyword search with the task of maximizing the spatio-temporal coverage and diversity of the returned top-f results. We first describe a baseline algorithm based on related search results diversification problems. Then, we develop an efficient approach which exploits a hybrid spatial-temporal-keyword index to drastically reduce query execution time. To that end, we extend two state-of-the- art indices for top-f spatio-textual queries and describe how our proposed approach can be applied on top of them. We evaluate the efficiency of our algorithms by conducting experiments on two large, real-world datasets containing geotagged tweets and photos.

DOOR: A Data Model for Crowdsourcing with Application to Emergency Response

Crowdsourcing allows us to employ collective human intelligence and resources in completing tasks in a wide variety of domains, such as mapping, translation, emergency response, and even fund raising. It first involves identification of a problem that can be solved using crowdsourcing and then its decomposition into tasks that workers can finish in a timely manner. Worker engagement analysis and data quality analysis are done afterwards. Such analysis activities are not supported by current platforms and are done in an ad-hoc fashion leading to duplicate efforts. As a first step towards realizing such analysis mechanisms, we propose a Data mOdel for crOwdsouRcing (DOOR), which is based on a fuzzy Entity-Relationship model in order to capture the uncertainty that is inherent in any crowdsourcing process. To illustrate its application, we have chosen the problem of collection of data about incidents for emergency response.

Clustering spatial data streams for targeted alerting in disaster response

Natural calamities and man-made hazards can occur in an unexpected and unanticipated manner. They cause large-scale damage, create disruptions, and need instant reaction. In the event of sudden onset of a crisis, rapid formulation of a notification strategy, timely dispatch of alerts, and action on those alerts are important elements of early warning systems that can save lives. However, current methods of disaster alerting lack in the area of targeted communication of hazard information. Location data of the population available as a spatial data stream can allow dynamic identification of homogeneous clusters of people. Crisis notifications can then be targeted by personalizing information and instructions for each cluster. In this paper, we present an approach for dynamically partitioning a region into areas around a hazard using clustering of real-time streaming data to aid emergency response management. We lay down important requirements for the clustering technique from the perspective of our scenario and select an algorithm for our implementation after comparison with others. We employ a weighted distance measure and demonstrate the performance of our model in different settings through a series of experiments using a dataset of cell tower locations of users in Ivory Coast in Africa.

MoveSafe: a framework for transportation mode-based targeted alerting in disaster response

Disasters, whether natural or man-made, can occur in an unexpected and unanticipated manner causing damage and disruptions. In the event of sudden onset of a hazard, private and public transport users and pedestrians need to be informed and guided to safety. Targeted alerting in early warning systems involves the communication of personalized information to a variety of communities based on their different needs and situations to improve alert usability and compliance. In this paper, we present MoveSafe, a generic and extensible framework for transportation mode-based dynamic partitioning of a population for targeted alerting and for better transport management in hazard occurrence scenarios. We infer the transportation mode of the users dynamically using their location traces through continuous feature extraction and maintenance. In combination with the hazard location, we use the transportation mode information to find clusters of people at potentially different levels of risk and with different information needs. The framework also supports a variety of classification features, classifiers, clustering dimensions, and clustering algorithms. We evaluate its performance in different settings and present the results.

Selecting representative and diverse spatio-textual posts over sliding windows

Thousands of posts are generated constantly by millions of users in social media, with an increasing portion of this content being geotagged. Keeping track of the whole stream of this spatio-textual content can easily become overwhelming for the user. In this paper, we address the problem of selecting a small, representative and diversified subset of posts, which is continuously updated over a sliding window. Each such subset can be considered as a concise summary of the streams contents within the respective time interval, being dynamically updated every time the window slides to reflect newly arrived and expired posts. We define the criteria for selecting the contents of each summary, and we present several alternative strategies for summary construction and maintenance that provide different trade-offs between information quality and performance. Furthermore, we optimize the performance of our methods by partitioning the newly arriving posts spatio-textually and computing bounds for the coverage and diversity of the posts in each partition. The proposed methods are evaluated experimentally using real-world datasets containing geotagged tweets and photos.

Keyword-based retrieval of frequent location sets in geotagged photo trails

We propose and study a novel type of keyword search for locations. Sets of locations are selected and ranked based on their co-occurrence in user trails in addition to satisfying a set of query keywords. We formally define the problem, outline our approach, and present experimental results.

Spatio-temporal keyword queries for moving objects

Many applications involve queries that combine spatial, temporal and textual filters. In this paper, we address the problem of efficient evaluation of queries that perform spatial, temporal and keyword-based filtering on historical movement data of objects which are additionally associated with textual information in the form of keywords. Our work combines and builds upon concepts and techniques for spatio-temporal and spatio-textual queries, proposing two hybrid indexes for this purpose. An experimental evaluation of the proposed approaches is presented, using real-world datasets from two different types of sources.

Trajectory Aggregation for a Routable Map

In this paper, we compare different approaches to merge trajectory data for later use in a map construction process. Merging trajectory data reduces storage space and can be of great help as far as data privacy is concerned. We consider different distance measures and different merge strategies, taking into account the cost of calculation, the connectivity of the results, and the storage space of the result. Finally, we give a hint on a possible information loss for each approach.