This paper describes our Align-and-Differentiate approach to the SemEval 2015 Task 2 competition for English Semantic Textual Similarity (STS) systems.Our submission achieved the top place on two of the five evaluation datasets.Our team placed 3rd among 28 participating teams, and our three runs ranked 4th, 6th and 7th among the 73 runs submitted by the 28 teams.Our approach improves upon the UMBC PairingWords system by semantically differentiating distributionally similar terms.This novel addition improves results by 2.5 points on the Pearson correlation measure.

Given a set A of activities expressed via temporal stochastic automata, and a set O of observations (detections of low level events), we study the problem of identifying instances of activities from A in O. While past work has developed algorithms to solve this problem, in this paper, we develop methods to significantly scale these algorithms. Our PASS architecture consists of three parts: (i) leveraging past work to represent all activities in A via a single “merged” graph, (ii) partitioning the graph into a set of C subgraphs, where (C + 1) is the number of compute nodes in a cluster, and (iii) developing a parallel activity detection algorithm that uses a different compute node in the cluster to intensively process each subgraph. We propose three possible partitioning methods and a parallel activity-search detection (PASS_Detect) algorithm that coordinates computations across nodes in the cluster. We report on experiments showing that our algorithms enable us to handle both large numbers of observations per second as well as large merged graphs. In particular, on a cluster with 9 compute nodes, PASS can reliably handle between 400K and 569K observations per second and merged graphs with as many as 50K vertices.

We present a hybrid method for automated on-demand creation of conceptual models of domain-specific knowledge. Models are thereby created using a two-step process of Domain Definition and Domain Description. Domain Definition creates a conceptual base whereas in the Domain Description relationships are added to the conceptual model using a pattern-based relational-targeting Information Extraction algorithm. The two-step process has the advantage over traditional approaches to ontology learning that it provides conceptual grounding through a top-down extraction and over information extraction that the extraction operates on a conceptual level so that concept integrity and reference are guaranteed. At the core of the extraction algorithm is a novel measure for semantic overlap of relationships that allows the extraction of multiple intensionally similar relationships while disambiguating merely extensionally similar relationships. The envisioned use of the created models is primarily in Information Retrieval applications, but the models can also serve as starting points for formal ontologies in Knowledge Representation

In domain-specific search systems, knowledge of a domain of interest is embedded as a backbone that guides the search process. But the knowledge used in most such systems 1. exists only for few well known broad domains; 2. is of a basic nature: either purely hierarchical or involves only few relationship types; and 3. is not always kept up-to-date missing insights from recently published results. In this paper we present a framework and implementation of a focused and up-to-date knowledge-based search system, called Scooner, that utilizes domain-specific knowledge extracted from recent bioscience abstracts. To our knowledge, this is the first attempt in the field to address all three shortcomings mentioned above. Since recent introduction for operational use at Applied Biotechnology Branch of AFRL, some biologists are using Scooner on a regular basis, while it is being made available for use by many more. Initial evaluations point to the promise of the approach in addressing the challenge we set out to address.

Admittedly this is a presumptuous title that should never be used when reporting on individual research advances. Wisdom is just not a scientific concept. In this case, though, we are reporting on recent developments on the web that lead us to believe that the web is on the way to providing a platform for not only information acquisition and business transactions but also for large scale knowledge development and decision support. It is likely that by now every web user has participated in some sort of social function or knowledge accumulating function on the web, many times without even being aware of it, simply by searching and browsing, other times deliberately by e.g. adding a piece of information to a Wikipedia article or by voting on a movie on IMDB.com. In this paper we will give some examples of how Web Wisdom is already emerging, some ideas of how we can create platforms that foster Web Wisdom and a critical evaluation of types of problems that can be subjected to Web Wisdom.

This paper describes a minimally guided approach to automatic domain model creation. The first step is to carve an area of interest out of the Wikipedia hierarchy based on a simple query or other starting point. The second step is to connect the concepts in this domain hierarchy with named relationships. A starting point is provided by Linked Open Data, such as DBPedia. Based on these community-generated facts we train a pattern-based fact-extraction algorithm to augment a domain hierarchy with previously unknown relationship occurrences. Pattern vectors are learned that represent occurrences of relationships between concepts. The process described can be fully automated and the number of relationships that can be learned grows as the community adds more information. Unlike approaches that are aimed at finding single, highly indicative patterns, we use the cumulative score of many pattern occurrences to increase extraction recall. The relationship identification process itself is based on positiveonly classification of training facts.

The traditional semantic approach has difficulties dealing with the dynamic domains involved in social, mobile, and sensor webs. Continuous semantics, which captures changing conceptualizations and relevant knowledge, can help us model those domains and analyze the related real-time data.

We present a method for growing the amount of knowledge available on the Web using a hermeneutic method that involves background knowledge, Information Extraction techniques and validation through discourse and use of the extracted information. We exemplify this using Linked Data as background knowledge, automatic Model/Ontology creation for the IE part and a Semantic Browser for evaluation. The hermeneutic approach, however, is open to be used with other IE techniques and other evaluation methods. We will present results from the model creation and anecdotal evidence for the feasibility of Validation through Use.

Many research studies adopt manually selected patterns for semantic relation extraction. However, manually identifying and discovering patterns is time consuming and it is difficult to discover all potential candidates. Instead, we propose an automatic pattern construction approach to extract verb synonyms and antonyms from English newspapers. Instead of relying on a single pattern, we combine results indicated by multiple patterns to maximize the recall.

We present a new method for mapping ontology schemas that address similar domains. The problem of ontology matching is crucial since we are witnessing a decentralized development and publication of ontological data. We formulate the problem of inferring a match between two ontologies as a maximum likelihood problem, and solve it using the technique of expectation-maximization (EM). Specifically, we adopt directed graphs as our model for ontology schemas and use a generalized version of EM to arrive at a map between the nodes of the graphs. We exploit the structural, lexical and instance similarity between the graphs, and differ from the previous approaches in the way we utilize them to arrive at, a possibly inexact, match. Inexact matching is the process of finding a best possible match between the two graphs when exact matching is not possible or is computationally difficult. In order to scale the method to large ontologies, we identify the computational bottlenecks and adapt the generalized EM by using a memory bounded partitioning scheme. We provide comparative experimental results in support of our method on two well-known ontology alignment benchmarks and discuss their implications.

We present a new method for mapping ontology schemas that address similar domains. The problem of ontology matching is crucial since we are witnessing a decentralized development and publication of ontological data. We formulate the problem of inferring a match between two ontologies as a maximum likelihood problem, and solve it using the technique of expectation-maximization (EM). Specifically, we adopt directed graphs as our model for ontology schemas and use a generalized version of EM to arrive at a map between the nodes of the graphs. We exploit the structural, lexical and instance similarity between the graphs, and differ from the previous approaches in the way we utilize them to arrive at, a possibly inexact, match. Inexact matching is the process of finding a best possible match between the two graphs when exact matching is not possible or is computationally difficult. In order to scale the method to large ontologies, we identify the computational bottlenecks and adapt the generalized EM by using a memory bounded partitioning scheme. We provide comparative experimental results in support of our method on two well-known ontology alignment benchmarks and discuss their implications.

Domain hierarchies are widely used as models underlying information retrieval tasks. Formal ontologies and taxonomies enrich such hierarchies further with properties and relationships but require manual effort; therefore they are costly to maintain, and often stale. Folksonomies and vocabularies lack rich category structure. Classification and extraction require the coverage of vocabularies and the alterability of folksonomies and can largely benefit from category relationships and other properties. With Doozer, a program for building conceptual models of information domains, we want to bridge the gap between the vocabularies and Folksonomies on the one side and the rich, expert-designed ontologies and taxonomies on the other. Doozer mines Wikipedia to produce tight domain hierarchies, starting with simple domain descriptions. It also adds relevancy scores for use in automated classification of information. The output model is described as a hierarchy of domain terms that can be used immediately for classifiers and IR systems or as a basis for manual or semi-automatic creation of formal ontologies.

Social networking, distributed problem solving and human computation have gained high visibility. Wikipedia is a well established service that incorporates aspects of these three fields of research. For this reason it is a good object of study for determining quality of solutions in a social setting that is open, completely distributed, bottom up and not peer reviewed by certified experts. In particular, this paper aims at identifying semantic convergence of Wikipedia articles; the notion that the content of an article stays stable regardless of continuing edits. This could lead to an automatic recommendation of good article tags but also add to the usability of Wikipedia as a Web Service and to its reliability for information extraction. The methods used and the results obtained in this research can be generalized to other communities that iteratively produce textual content.

Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark. A Cataloguing in Publication record for this book is available from the British Library. All work contributed to this encyclopedia set is new, previously-unpublished material. The views expressed in this encyclopedia set are those of the authors, but not necessarily of the publisher. ABSTRACT Enabling applications that exploit heterogeneous data in the Semantic Web will require us to harness a broad variety of semantics. Considering the role of semantics in a number of research areas in computer science, we organize semantics in three forms—implicit, formal, and powerful—and explore their roles in enabling some of the key capabilities related to the Semantic Web. The central message of this chapter is that building the Semantic Web purely on description logics will artificially limit its potential, and that we will need to both exploit well-known techniques that support implicit semantics, and develop more powerful semantic techniques.

High throughput glycoproteomics, similar to genomics and proteomics, involves extremely large volumes of distributed, heterogeneous data as a basis for identification and quantification of a structurally diverse collection of biomolecules. The ability to share, compare, query for and most critically correlate datasets using the native biological relationships are some of the challenges being faced by glycobiology researchers. As a solution for these challenges, we are building a semantic structure, using a suite of ontologies, which supports management of data and information at each step of the experimental lifecycle. This framework will enable researchers to leverage the large scale of glycoproteomics data to their benefit.In this paper, we focus on the design of these biological ontology schemas with an emphasis on relationships between biological concepts, on the use of novel approaches to populate these complex ontologies including integrating extremely large datasets ( 500MB) as part of the instance base and on the evaluation of ontologies using OntoQA [38] metrics. The application of these ontologies in providing informatics solutions, for high throughput glycoproteomics experimental domain, is also discussed. We present our experience as a use case of developing two ontologies in one domain, to be part of a set of use cases, which are used in the development of an emergent framework for building and deploying biological ontologies.

The field of BioInformatics has become a major venue for the development and application of computational ontologies. Ranging from controlled vocabularies to annotation of experimental data to reasoning tasks, BioOntologies are advancing to form a comprehensive knowledge foundation in this field. With the Glycomics Ontology (GlycO), we are aiming at providing both a sufficiently large knowledge base and a schema that allows classification of and reasoning about the concepts we expect to encounter in the glycoproteomics field. The schema exploits the expressiveness of OWL-DL to place restrictions on relationships, thus making it suitable to be used as a means to classify new instance data. On the instance level, the knowledge is modularized to address granularity issues regularly found in ontology design. Larger structures are semantically composed from smaller canonical building blocks. The information needed to populate the knowledge base is automatically extracted from several partially overlapping sources. In order to avoid multiple entries, transformation and disambiguation techniques are applied. An intelligent search is then used to identify the individual building blocks that model the larger chemical structures. To ensure ontological soundness, GlycO has been annotated with OntoClean properties and evaluated with respect to those. In order to facilitate its use in conjunction with other biomedical Ontologies, GlycO has been checked for NCBO compliance and has been submitted to the OBO website.