Procurement is an essential operation of every organization regardless of its size or domain. As such, aggregating the demands could lead to better value-for-money due to: (1) lower bulk prices; (2) larger vendor tendering; (3) lower shipping and handling fees; and (4) reduced legal and administration overheads. This paper describes our experience in developing an AI solution for demand aggregation and deploying it in A*STAR, a large governmental research organization in Singapore with procurement expenditure to the scale of hundreds of millions of dollars annually. We formulate the demand aggregation problem using a bipartite graph model depicting the relationship between procured items and target vendors, and show that identifying maximal edge bicliques within that graph would reveal potential demand aggregation patterns. We propose an unsupervised learning methodology for efficiently mining such bicliques using a novel Monte Carlo subspace clustering approach. Based on this, a proof-of-concept prototype was developed and tested with the end users during 2017, and later trialed and iteratively refined, before being rolled out in 2019. The final performance was 71% of past cases transformed into bulk tenders correctly detected by the engine; for new opportunities pointed out by the engine 81% were deemed useful for potential bulk tender contracts in the future. Additionally, per each valid pattern identified, the engine achieved 100% precision (all aggregated purchase orders were correct), and 79% recall (the engine correctly identified 79% of orders that should have been put into the bulk tenders). Overall, the cost savings from the true positive contracts spotted so far are estimated to be S$7 million annually.

Many telecommunication companies today have actively started to transform the way they do business, going beyond communication infrastructure providers are repositioning themselves as data-driven service providers to create new revenue streams. In this paper, we present a novel industrial application where a scalable Big data approach combined with deep learning is used successfully to classify massive mobile web log data, to get new aggregated insights on customer web behaviors that could be applied to various industry verticals.

Procurement is a set of activities and processes related to acquisition of goods and services through purchase orders placed by organization employees,from external contractors. Thisarticle describes practical experiments with procurement dataset of a major governmental organization in Singapore. In particular, we highlight the problems that emerge when trying to implement analytics for prediction of future purchases. The goal of such analytics is to deliver beneficial information to procurement office that plans and manages relationships with external sellers. In the article we describe the characteristics of the procurement dataset specifics and its implications on the future purchase problem that we attempt to solve using Markov chains model. Our analysis showshigh diversity of purchase descriptionsresulting in low ability to detect sequential patterns of purchasing officers. The solution presented in the article is additional dataset preprocessing involving use of hierarchical clustering. Our experiments with various similarity measures show an improvement allowing a practical deployment within our procurement analytics system prepared for the case study governmental organization.

Service composition is the process of automatically constructing a workflow from individual services so as to satisfy user requirements. When composing service workflows, it is important that both functional and non-functional requirements need to be considered. The so-called QoS-aware service composition is typically formulated analogous to the classical MMMKP optmization problem, and does not account for service and QoS dependencies which are often indispensable in real life. In this paper, we consider QoS-aware service composition in the presence of service-dependent QoS and propose a novel method that dynamically refines the composed workflow in light of QoS dependencies and user-provided topological and QoS constraints. Through evaluations, we demonstrate that our approach is capable of offering significant improvements in performance on real life scenarios with complex service and QoS dependencies.

Return on investment is a critical decision factor for end-users going for cloud deployments. However, major cloud vendors typically provide a myriad of interdependent cloud service options in a variety of purchasing models, that severely complicates cost estimation and optimization. In this paper, we propose a novel Amazon EC2 cost optimization system, called EC2 Bargain Hunter, that innovatively combines services and cloud computing principles with ideas from semantic technologies. The system supports the entire-range of EC2 instance types, and can be used in real-time to perform live cost optimization. We demonstrate that unprecedented cost savings, by a factor of 30, on Amazon EC2 offerings can be found with this system in a few clicks. Furthermore, our approach can be adapted to other IaaS providers, which enables truly real-life cloud cost optimization and thus is a significant step towards making the cloud really cost-effective for the end-users.

Hadoop is a popular open source implementation of MapReduce, that has a number of prominent users including Yahoo!, Facebook, and Twitter. Though several works have focused on deploying algorithms on Hadoop MapReduce, research efforts into applying formal methods to prove the correctness of hadoop systems are limited. In this paper we propose a holistic approach to verify the correctness of hadoop systems using model checking techniques. We model Hadoop's parallel architecture to constraint it to valid start up ordering and identify and prove the benefits of data locality, deadlock-freeness and non-termination among others.

Analogous to software-as-a-service (SaaS), platform-as-a-service (PaaS) and infrastructure-as-a-service (IaaS), data-as-a-service (DaaS) is used to provide data on demand to users over the Internet and is gaining popularity in the current cloud computing era. In particular, several Open Data initiatives have led to a number of data services in various formats becoming available online, and it remains a challenge to make full use of the data by transferring between and answering queries from different sources in a automatic, dynamic and meaningful manner. In this paper we propose a service-oriented, composition-based approach towards tackling the integration of open data services. We provide a formal, semantic-based modeling of data services and convert the integration problem into the service composition problem. Then the composition graph can be used to create the executable queries to the various data services. We illustrate our idea by using a case study in the real estate domain.

Cloud computing has attracted a lot of interest from both academia and industry because of the promise of cost-effectiveness. As a result, increasingly large numbers of cloud services have become available. This implies challenges in brokering between the cloud consumers and providers. To meet this demand, several cloud brokers have been introduced. Unfortunately, most of the cloud brokers cannot solve real life scenarios and cannot be evaluated in an unbiased manner because their evaluations are often academic and independent. This paper proposes a new benchmark to evaluate and compare cloud brokers. Our benchmark called Cloud Broker Challenge (CBC) describes the cloud providers, cloud consumers, and goals with five variety levels of complexities. CBC Benchmark is useful to the Cloud community for an unbiased evaluation and comparison of brokers and can potentially enable development of real-life cloud brokers.

We propose a novel approach based on model checking for automated non-linear service composition. Modeling services as interleaved processes, we formulate the service composition problem as verifying a safety property and show that multiple non-linear compositions can be constructed from the counter-example. The state explosion problem is tackled by using service clustering and computing service closures.

Cloud broker is an entity that manages the use, performance and delivery of cloud services and negotiates relationships between cloud providers and cloud consumers. In real life scenarios, automated cloud service brokering is often challenging because the service descriptions may involve complex constraints and require flexible semantic matching. Furthermore, cloud providers often use non-standard formats leading to semantic interoperability issues. In this paper, we formulate cloud service brokering under a service oriented framework, and propose a novel OWL-S based semantic cloud service discovery and selection system. The proposed system supports dynamic semantic matching of cloud services described with complex constraints. We consider a practical cloud service brokering scenario, and show with detailed illustration that our system is promising for real-life applications.

IPTV deals with multimedia services delivered over IP based networks managed to provide the required level of quality of service and experience, security, interactivity and reliability. Service discovery is an important component of IPTV system, but it is challenging due to the unique domain specifics of IPTV and complexities of IPTV ecosystem. In this paper, we formulate the IPTV service discovery problem under an SOA framework, and propose a novel standards-based semantic IPTV service discovery system. The proposed system is based on unifying the traditionally step-wise iterative process of IPTV service discovery, and a semantic integration of service provider and IPTV service descriptions. For demonstration, we consider a specific discovery scenario involving service providers that offer a combination of standard TV services such as linear TV and VoD content, and also advanced interactive multimedia-based IPTV services. With detailed illustration, we demonstrate that our proposed system provides simple but effective discovery of IPTV services, and possesses several benefits that can potentially enhance end-user experience.

Rich information on mutations and their impacts is scattered across scientific texts and literature. Reuse of mutation impact annotations requires grounding mutations to the correct positions on sequences extracted from protein databases as a critical step. This paper presents a generic method for grounding textual mentions of mutation entities to protein sequences, that is based on an OWL-DL ontology driven workflow that integrates text and sequence information in a semantically consistent way. Mutation mentions mined from texts are iteratively mapped onto candidate proteins, and an ontology mining algorithm facilitates their correct grounding to a protein sequence. Using a gold standard corpus of full text articles and corresponding protein sequences we show the proposed method is promising compared to existing approaches.

Web service composition is the process of constructing a set of Web services which, when invoked with some user input in a particular order, can produce the output to the user's requirements. This paper proposes a novel model checking based approach for automated service composition. Modeling services as a set of interleaved processes in a class of process algebra, we formulate service composition as model checking asserted on a specific type of property on the model. We show that, under this formulation, correct composition workflows can be constructed from the counter-examples provided by model checking. With a case study on online hotel booking services, we demonstrate that the proposed approach can support directed a cyclic composition graphs and the generated composition graphs are automatically verified.

Semantic web services (SWS) provide rich and formal representations of services by capturing the semantics of requests and service descriptions as well as the context of service interactions. Recently, multiple SWS standards such OWL-S, WSMO, and SAWSDL have emerged making service interoperability increasingly a challenge. In order to leverage the true potential of Semantic Web services, this service heterogeneity problem must be overcome. This paper introduces an OWL-S based mediator for service level and data level mediation across heterogeneous semantic service descriptions. We consider WSMO and SAWSDL service description formalisms and present mediation strategies via OWL-S, by converting service descriptions and mapping the underlying core ontology from WSML to OWL. Through experiments, the proposed mediator is found to be promising.

SNOMED CT is a comprehensive medical terminology used for standardizing storage, retrieval and exchange of electronic health data. Towards more expressive knowledge representation, recent efforts have attempted to model SNOMED CT using OWL, but have often been stymied due to its size and complexity. This paper proposes a new approach to model and realize SNOMED1as an OWL ontology on large-scale. Rather than constructing raw owl files, we exploit a framework based on a triplestore-powered semantic backend that is interfaced through web services. With this framework, we show that almost the whole of the core tables of SNOMED including concepts, descriptions and relationships can be loaded into an OWL ontology, which can be queried in real-time. As part of the OWL modeling, we also propose a new strategy to model role groups via n-ary relationship ontology patterns. Through experiments on several use cases, we demonstrate that ours is a promising approach for modeling and realizing SNOMED in OWL.

Discovering web services is the process of finding web services that satisfy specific requirements. Rich and formal representations of services and interactions are required for principled selection of services, context-aware analysis and satisfaction of requests, as well as dynamic interaction and negotiation with the service providers. Semantic technologies facilitate specialization and generalization of service needs as well as service composition. Thus, a higher degree of automation and more precise results can be achieved. In this paper, we provide a review of current Semantic Web Service (SWS) discovery approaches.

Public relations is an important management staff function in this computer age, and accurate media information retrieval and appropriate measurement for publicity tracking in a campaign management system is of utmost importance. This paper proposes an intelligent campaign management system, called KnowleTracker, based on a semantic service-oriented approach. KnowleTracker has powerful deep mining functions to pull out news and other information that may lie several layers below the front page based on a semantic search. The returned results are presented at different levels of detail and viewed from different angles, and show customers, detailed information of where their news releases are distributed. We demonstrate the utility of KnowleTracker in a real-world scenario through a case study on "A*STAR Event Campaign Management", and argue that our system is an effective tool to tackle the complex task of campaign management and tracking.

Creation of pedagogical learning models to handle the specificity of mobile learning and the inherent constraints of mobile devices is a fundamental challenge in mobile learning. In this paper, we propose a generic semantics-based service-oriented architecture (SOA) and show how ontologies, when used together with Web services, can provide mobile users with a fresh learning experience. The proposed SOA is flexible and loosely coupled. It facilitates pluggable context-aware learner services that are multi-modal, as well as content services that support seamless integration of legacy content from possibly multiple vendors for customizable delivery. As illustration, we consider a mobile learning scenario in a zoo, and demonstrate how a mobile device user can interact with the proposed system for m-learning.

Abstract Motivation: Locating transcription factor binding sites (motifs) is a key step in understanding gene regulation. Based on Tompa's benchmark study, the performance of current de novo motif finders is far from satisfactory (with sensitivity ≤0.222 and precision ≤0.307). The same study also shows that no motif finder performs consistently well over all datasets. Hence, it is not clear which finder one should use for a given dataset. To address this issue, a class of algorithms called ensemble methods have been proposed. Though the existing ensemble methods overall perform better than stand-alone motif finders, the improvement gained is not substantial. Our study reveals that these methods do not fully exploit the information obtained from the results of individual finders, resulting in minor improvement in sensitivity and poor precision. Results: In this article, we identify several key observations on how to utilize the results from individual finders and design a novel ensemble method, MotifVoter, to predict the motifs and binding sites. Evaluations on 186 datasets show that MotifVoter can locate more than 95% of the binding sites found by its component motif finders. In terms of sensitivity and precision, MotifVoter outperforms stand-alone motif finders and ensemble methods significantly on Tompa's benchmark, Escherichia coli, and ChIP-Chip datasets. MotifVoter is available online via a web server with several biologist-friendly features. Availability: http://www.comp.nus.edu.sg/~bioinfo/MotifVoter Contact: [email protected] supplementary information: Supplementary data are available at Bioinformatics online.

BACKGROUND: The indexing of scientific literature and content is a relevant and contemporary requirement within life science information systems. Navigating information available in legacy formats continues to be a challenge both in enterprise and academic domains. The emergence of semantic web technologies and their fusion with artificial intelligence techniques has provided a new toolkit with which to address these data integration challenges. In the emerging field of lipidomics such navigation challenges are barriers to the translation of scientific results into actionable knowledge, critical to the treatment of diseases such as Alzheimer's syndrome, Mycobacterium infections and cancer. RESULTS: We present a literature-driven workflow involving document delivery and natural language processing steps generating tagged sentences containing lipid, protein and disease names, which are instantiated to custom designed lipid ontology. We describe the design challenges in capturing lipid nomenclature, the mandate of the ontology and its role as query model in the navigation of the lipid bibliosphere. We illustrate the extent of the description logic-based A-box query capability provided by the instantiated ontology using a graphical query composer to query sentences describing lipid-protein and lipid-disease correlations. CONCLUSION: As scientists accept the need to readjust the manner in which we search for information and derive knowledge we illustrate a system that can constrain the literature explosion and knowledge navigation problems. Specifically we have focussed on solving this challenge for lipidomics researchers who have to deal with the lack of standardized vocabulary, differing classification schemes, and a wide array of synonyms before being able to derive scientific insights. The use of the OWL-DL variant of the Web Ontology Language (OWL) and description logic reasoning is pivotal in this regard, providing the lipid scientist with advanced query access to the results of text mining algorithms instantiated into the ontology. The visual query paradigm assists in the adoption of this technology.

Uninhibited access to the unstructured information distributed across the web and in scientific literature databases continues to be beyond the reach of scientists and health professionals. To address this challenge we have developed a literature driven, ontology-centric navigation infrastructure consisting of a content acquisition engine, a domain-specific ontology (in OWL-DL) and an ontology instantiation pipeline delivering sentences derived by domain-specific text mining. A visual query tool for reasoning over A-box instances in the populated ontology is presented and used to build conceptual queries that can be issued to the knowledgebase. We have deployed this generic infrastructure to facilitate data integration and knowledge sharing in the domain of dengue, which is one of the most prevalent viral diseases that continue to infect millions of people in the tropical and subtropical regions annually. Using our unique methodology we illustrate simplified search and discovery on dengue information derived from distributed resources and aggregated according to dengue ontology. Furthermore we apply data mining to the instantiated ontology to elucidate trends in the mentions of dengue serotypes in scientific abstracts since 1974.

MOTIVATION: Identification of motifs is one of the critical stages in studying the regulatory interactions of genes. Motifs can have complicated patterns. In particular, spaced motifs, an important class of motifs, consist of several short segments separated by spacers of different lengths. Locating spaced motifs is not trivial. Existing motif-finding algorithms are either designed for monad motifs (short contiguous patterns with some mismatches) or have assumptions on the spacer lengths or can only handle at most two segments. An effective motif finder for generic spaced motifs is highly desirable. RESULTS: This article proposes a novel approach for identifying spaced motifs with any number of spacers of different lengths. We introduce the notion of submotifs to capture the segments in the spaced motif and formulate the motif-finding problem as a frequent submotif mining problem. We provide an algorithm called SPACE to solve the problem. Based on experiments on real biological datasets, synthetic datasets and the motif assessment benchmarks by Tompa et al., we show that our algorithm performs better than existing tools for spaced motifs with improvements in both sensitivity and specificity and for monads, SPACE performs as good as other tools. AVAILABILITY: The source code is available upon request from the authors.

In this paper we present a novel approach for retrieval of thematic video content from mixed media. Based on the principles of conceptual blending, information from different media is mined for emergent structures from mixed media. We have built a system, called OntoMedia, to test the efficacy of this approach over traditional methods for media retrieval. The system employs an ontology as a unified indexing scheme for associated text documents for the mixed media content. By applying graph theoretic path finding operations to the ontology to process queries for video content, we show that the system is able to synchronize information from multiple media sources and retrieve semantically related content across media types. A trial experiment was carried out using mixed media courseware from the Singapore-MIT Alliance (SMA) distance education course. Results showed that information retrieval along a path performs significantly better than keyword-based and ontology-focused searches.

This paper investigates wrapper induction from web sites whose layout may change over time. We formulate the reinduction as an incremental learning problem and identify that wrapper induction from an incomplete label is a key problem to be solved. We propose a novel algorithm for incrementally inducing LR wrappers and show that this algorithm asymptotically identifies the correct wrapper as the number of tuples is increased. This property is used to propose a LR wrapper reinduction algorithm. This algorithm requires examples to be provided exactly once and there-after the algorithm can detect the layout changes and reinduce wrappers automatically. In experimental studies, we observe that the reinduction algorithm is able to achieve near perfect performance.

We address the problem of mining a class of semantic networks, called concept frame graphs (CFG's), for knowledge discovery from text. This new representation is motivated by the need to capture richer text content so that nontrivial mining tasks can be performed. We first define the CFG representation and then describe a rule-based algorithm for constructing a CFG from text documents. Treating the CFG as a networked knowledge base, we propose new methods for text mining. On a specific task of discovering the top companies in an area, we observe that our approach leads to simpler content mining algorithms, once the CFG has been constructed. Moreover, exploiting the network structure of CFG results in significant improvements in precision and recall.

We address the text content mining problem through a concept based framework by constructing a conceptual knowledge base and discovering knowledge therefrom. Defining a novel representation called the Concept Frame Graph (CFG), we propose a learning algorithm for constructing a CFG knowledge base from text documents. An interactive concept map visualization technique is presented for user-guided knowledge discovery from the knowledge base. Through experimental studies on real life documents, we observe that the proposed approach is promising for mining deeper knowledge.

The problem of analyzing the finite time behavior of learning automata is considered. This problem involves the finite time analysis of the learning algorithm used by the learning automaton and is important in determining the rate of convergence of the automaton. In this paper, a general framework for analyzing the finite time behavior of the automaton learning algorithms is proposed. Using this framework, the finite time analysis of the Pursuit Algorithm is presented. We have considered both continuous and discretized forms of the pursuit algorithm. Based on the results of the analysis, we compare the rates of convergence of these two versions of the pursuit algorithm. At the end of the paper, we also compare our framework with that of Probably Approximately Correct (PAC) learning.

The problem of learning conjunctive concepts from a series of positive and negative examples of the concept is considered. Employing a probabilistic structure on the domain, the goal of such inductive learning is precisely characterized. A parallel distributed stochastic algorithm is presented. It is proved that the algorithm will converge to the concept description with maximum probability of correct classification in the presence of up to 50% unbiased noise. A novel neural network structure that implements the learning algorithm is proposed. Through empirical studies it is seen that the algorithm is quite efficient for learning conjunctive concepts.>