Kalpa Gunaratna

A statistical and schema independent approach to identify equivalent properties on linked data

Linked Open Data (LOD) cloud has gained significant attention in the Semantic Web community recently. Currently it consists of approximately 295 interlinked datasets with over 50 billion triples including 500 million links, and continues to expand in size. This vast source of structured information has the potential to have a significant impact on knowledge-based applications. However, a key impediment to the use of LOD cloud is limited support for data integration tasks over concepts, instances, and properties. Efforts to address this limitation over properties have focused on matching data-type properties across datasets; however, matching of object-type properties has not received similar attention. We present an approach that can automatically match object-type properties across linked datasets, primarily exploiting and bootstrapping from entity co-reference links such as owl:sameAs. Our evaluation, using sample instance sets taken from Freebase, DBpedia, LinkedMDB, and DBLP datasets covering multiple domains shows that our approach matches properties with high precision and recall (on average, F measure gain of 57% - 78%).

Gleaning Types for Literals in RDF Triples with Application to Entity Summarization

Associating meaning with data in a machine-readable format is at the core of the Semantic Web vision, and typing is one such process. Typing assigning a class selected from schema information can be attached to URI resources in RDF/S knowledge graphs and datasets to improve quality, reliability, and analysis. There are two types of properties: object properties, and datatype properties. Type information can be made available for object properties as their object values are URIs. Typed object properties allow richer semantic analysis compared to datatype properties, whose object values are literals. In fact, many datatype properties can be analyzed to suggest types selected from a schema similar to object properties, enabling their wider use in applications. In this paper, we propose an approach to glean types for datatype properties by processing their object values. We show the usefulness of generated types by utilizing them to group facts on the basis of their semantics in computing diversified entity summaries by extending a state-of-the-art summarization algorithm.

eAssistant: Cognitive Assistance for Identification and Auto-Triage of Actionable Conversations

The browser and screen have been the main user interfaces of the Web and mobile apps. The notification mechanism is an evolution in the user interaction paradigm by keeping users updated without checking applications. Conversational agents are posed to be the next revolution in user interaction paradigms. However, without intelligence on the triage of content served by the interaction and content differentiation in applications, interaction paradigms may still place the burden of information overload on users. In this paper, we focus on the problem of intelligent identification of actionable information in the content served by applications, and in particular in productivity applications (such as email, chat, messaging, social collaboration tools, etc.). We present eAssistant, which offers a novel fine-grained action identification method in an adaptive, personalizable, and online-trainable manner, and a cognitive agent/API that uses action information and user-centric conversation characteristics to auto-triage user conversations. The introduced method identifies individual actions and associated metadata; it is extensible in terms of the number of action classes; it learns in an online and continuous manner via user interactions and feedback, and it is personalizable to different users. We have evaluated the proposed method using real-world datasets. The results show that the method achieves higher accuracy compared to traditional ways of formulating the problem, while exhibiting additional desired properties of online, personalized, and adaptive learning. In eAssistant, we introduce a multi-dimensional learning model of conversations auto-triage, defined based on a user study and NLP-based information extraction techniques, to auto-triage user conversations on social collaboration and productivity tools.

Alignment and dataset identification of linked data in Semantic Web

The Linked Open Data (LOD) cloud has gained significant attention in the Semantic Web community over the past few years. With rapid expansion in size and diversity, it consists of over 800 interlinked datasets with over 60 billion triples. These datasets encapsulate structured data and knowledge spanning over varied domains such as entertainment, life sciences, publications, geography, and government. Applications can take advantage of this by using the knowledge distributed over the interconnected datasets, which is not realistic to find in a single place elsewhere. However, two of the key obstacles in using the LOD cloud are the limited support for data integration tasks over concepts, instances, and properties, and relevant data source selection for querying over multiple datasets. We review, in brief, some of the important and interesting technical approaches found in the literature that address these two issues. We observe that the general purpose alignment techniques developed outside the LOD context fall short in meeting the heterogeneous data representation of LOD. Therefore, an LOD‐specific review of these techniques (especially for alignment) is important to the community. The topics covered and discussed in this article fall under two broad categories, namely alignment techniques for LOD datasets and relevant data source selection in the context of query processing over LOD datasets.

Relatedness-based Multi-Entity Summarization

Representing world knowledge in a machine processable format is important as entities and their descriptions have fueled tremendous growth in knowledge-rich information processing platforms, services, and systems. Prominent applications of knowledge graphs include search engines (e.g., Google Search and Microsoft Bing), email clients (e.g., Gmail), and intelligent personal assistants (e.g., Google Now, Amazon Echo, and Apples Siri). In this paper, we present an approach that can summarize facts about a collection of entities by analyzing their relatedness in preference to summarizing each entity in isolation. Specifically, we generate informative entity summaries by selecting: (i) inter-entity facts that are similar and (ii) intra-entity facts that are important and diverse. We employ a constrained knapsack problem solving approach to efficiently compute entity summaries. We perform both qualitative and quantitative experiments and demonstrate that our approach yields promising results compared to two other stand-alone state-of-the-art entity summarization approaches.

A Study in Hadoop Streaming with Matlab for NMR Data Processing

Applying Cloud computing techniques for analyzing large data sets has shown promise in many data-driven scientific applications. Our approach presented here is to use Cloud computing for Nuclear Magnetic Resonance (NMR)data analysis which normally consists of large amounts of data. Biologists often use third party or commercial software for ease of use. Enabling the capability to use this kind of software in a Cloud will be highly advantageous in many ways. Scripting languages especially designed for clouds may not have the flexibility biologists need for their purposes. Although this is true, they are familiar with special software packages that allow them to write complex calculations with minimum effort, but are often not compatible with a Cloud environment. Therefore, biologists who are trying to perform analysis on NMR data, acquire many advantages due to our proposed solution. Our solution gives them the flexibility to Cloud-enable their familiar software and it also enables them to perform calculations on a significant amount of data that was not previously possible. Our study is also applicable to any other environment in need of similar flexibility. We are currently in the initial stage of developing a framework for NMR data analysis.

Demo: approximate semantic matching in the collider event processing engine

This demo presents a use case from the energy management domain. It builds upon previous work on approximate semantic matching of heterogeneous events and compares two semantic matching scenarios: exact and approximate. It illustrates how a large number of exact matching event subscriptions are needed to match heterogeneous power consumption events. It then demonstrates how a small number of approximate semantic matching subscriptions are needed but possibly with a lower true positives/negatives performance. The demo is delivered via the COLLIDER approximate event processing engine currently under development in DERI.