The role of context in image annotation and recommendation

With the rise of smart phones, lifelogging devices (e.g. Google Glass) and popularity of image sharing websites (e.g. Flickr), users are capturing and sharing every aspect of their life online producing a wealth of visual content. Of these uploaded images, the majority are poorly annotated or exist in complete semantic isolation making the process of building retrieval systems difficult as one must firstly understand the meaning of an image in order to retrieve it. To alleviate this problem, many image sharing websites offer manual annotation tools which allow the user to “tag” their photos, however, these techniques are laborious and as a result have been poorly adopted; Sigurbjörnsson and van Zwol (2008) showed that 64% of images uploaded to Flickr are annotated with < 4 tags. Due to this, an entire body of research has focused on the automatic annotation of images (Hanbury, 2008; Smeulders et al., 2000; Zhang et al., 2012a) where one attempts to bridge the semantic gap between an image’s appearance and meaning e.g. the objects present. Despite two decades of research the semantic gap still largely exists and as a result automatic annotation models often offer unsatisfactory performance for industrial implementation. Further, these techniques can only annotate what they see, thus ignoring the “bigger picture” surrounding an image (e.g. its location, the event, the people present etc). Much work has therefore focused on building photo tag recommendation (PTR) methods which aid the user in the annotation process by suggesting tags related to those already present. These works have mainly focused on computing relationships between tags based on historical images e.g. that NY and timessquare co-exist in many images and are therefore highly correlated. However, tags are inherently noisy, sparse and ill-defined often resulting in poor PTR accuracy e.g. does NY refer to New York or New Year? This thesis proposes the exploitation of an image’s context which, unlike textual evidences, is always present, in order to alleviate this ambiguity in the tag recommendation process. Specifically we exploit the “what, who, where, when and how” of the image capture process in order to complement textual evidences in various photo tag recommendation and retrieval scenarios. In part II, we combine text, content-based (e.g. # of faces present) and contextual (e.g. day-of-the-week taken) signals for tag recommendation purposes, achieving up to a 75% improvement to precision@5 in comparison to a text-only TF-IDF baseline. We then consider external knowledge sources (i.e. Wikipedia & Twitter) as an alternative to (slower moving) Flickr in order to build recommendation models on, showing that similar accuracy could be achieved on these faster moving, yet entirely textual, datasets. In part II, we also highlight the merits of diversifying tag recommendation lists before discussing at length various problems with existing automatic image annotation and photo tag recommendation evaluation collections. In part III, we propose three new image retrieval scenarios, namely “visual event summarisation”, “image popularity prediction” and “lifelog summarisation”. In the first scenario, we attempt to produce a rank of relevant and diverse images for various news events by (i) removing irrelevant images such memes and visual duplicates (ii) before semantically clustering images based on the tweets in which they were originally posted. Using this approach, we were able to achieve over 50% precision for images in the top 5 ranks. In the second retrieval scenario, we show that by combining contextual and content-based features from images, we are able to predict if it will become “popular” (or not) with 74% accuracy, using an SVM classifier. Finally, in chapter 9 we employ blur detection and perceptual-hash clustering in order to remove noisy images from lifelogs, before combining visual and geo-temporal signals in order to capture a user’s “key moments” within their day. We believe that the results of this thesis show an important step towards building effective image retrieval models when there lacks sufficient textual content (i.e. a cold start).

Interorganizational Innovation: The Role of Suppliers in Enhancing Buyer Innovation

This dissertation explores the effect of innovative knowledge transfer across supply chain partners. My research seeks to understand the manner by which a firm is able to benefit from the innovative capabilities of its supply chain partners and utilize the external knowledge they hold to increase its own levels of innovation. Specifically, I make use of patent data as a proxy for firm-level innovation and develop both independent and dependent variables from the data contained within the patent filings. I further examine the means by which key dyadic and portfolio supply chain relationship characteristics moderate the relationship between supplier innovation and buyer innovation. I investigate factors such as the degree of transactional reciprocity between the buyer and supplier, the similarity of the firms’ knowledge bases, and specific chain characteristics (e.g., geographic propinquity) to provide greater understanding of the means by which the transfer of innovative knowledge across firms in a supply chain can be enhanced or inhibited. This dissertation spans three essays to provide insights into the role that supply chain relationships play in affecting a focal firm’s level of innovation. While innovation has been at the core of a wide body of research, very little empirical work exists that considers the role of vertical buyer-supplier relationships on a firm’s ability to develop new and novel innovations. I begin by considering the fundamental unit of analysis within a supply chain, the buyer-supplier dyad. After developing initial insights based on the interactions between singular buyers and suppliers, essay two extends the analysis to consider the full spectrum of a buyer’s supply base by aggregating the individual buyer-supplier dyad level data into firm-supply network level data. Through this broader level of analysis, I am able to examine how the relational characteristics between a buyer firm and its supply base affect its ability to leverage the full portfolio of its suppliers’ innovative knowledge. Finally, in essay three I further extend the analysis to explore the means by which a buyer firm can use its suppliers to enhance its ability to access distant knowledge held by other organizations that the buyer is only connected to indirectly through its suppliers.

The external dimension of the EU’s fight against organized crime:the search for coherence between rhetoric and practice

SSince the external dimension of the European Union’s Justice and Home Affairs (JHA) began to be considered, a substantial amount of literature has been dedicated to discussing how the EU is cooperating with non-member states in order to counter problems such as terrorism, organized crime and illegal migration. According to the EU, the degree of security interconnectedness has become so relevant that threats can only be adequately controlled if there is effective concerted regional action. This reasoning has led the EU to develop a number of instruments, which have resulted in the exporting of certain elements of its JHA policies, either through negotiation or socialization. Although the literature has explored how this transfer has been applied to the field of terrorism and immigration, very little has been written on the externalisation of knowledge, practice and norms in the area of organized crime. This article proposes to bridge this gap by looking at EU practice in the development of the external dimension of organized crime policies, through the theoretical lens of the EU governance framework.

The search for migraine genes : an overview of current knowledge

Migraine is a complex familial condition that imparts a significant burden on society. There is evidence for a role of genetic factors in migraine, and elucidating the genetic basis of this disabling condition remains the focus of much research. In this review we discuss results of genetic studies to date, from the discovery of the role of neural ion channel gene mutations in familial hemiplegic migraine (FHM) to linkage analyses and candidate gene studies in the more common forms of migraine. The success of FHM regarding discovery of genetic defects associated with the disorder remains elusive in common migraine, and causative genes have not yet been identified. Thus we suggest additional approaches for analysing the genetic basis of this disorder. The continuing search for migraine genes may aid in a greater understanding of the mechanisms that underlie the disorder and potentially lead to significant diagnostic and therapeutic applications.

Knowledge discovery from tree databases using balanced optimal search

This research is a step forward in discovering knowledge from databases of complex structure like tree or graph. Several data mining algorithms are developed based on a novel representation called Balanced Optimal Search for extracting implicit, unknown and potentially useful information like patterns, similarities and various relationships from tree data, which are also proved to be advantageous in analysing big data. This thesis focuses on analysing unordered tree data, which is robust to data inconsistency, irregularity and swift information changes, hence, in the era of big data it becomes a popular and widely used data model.

Standardization efforts: The relationship between knowledge dimensions, search processes and innovation outcomes

We explore how a standardization effort (i.e., when a firm pursues standards to further innovation) involves different search processes for knowledge and innovation outcomes. Using an inductive case study of Vanke, a leading Chinese property developer, we show how varying degrees of knowledge complexity and codification combine to produce a typology of four types of search process: active, integrative, decentralized and passive, resulting in four types of innovation outcome: modular, radical, incremental and architectural. We argue that when the standardization effort in a firm involves highly codified knowledge, incremental and architectural innovation outcomes are fostered, while modular and radical innovations are hindered. We discuss how standardization efforts can result in a second-order innovation capability, and conclude by calling for comparative research in other settings to understand how standardization efforts can be suited to different types of search process in different industry contexts.

A New Definition of A Priori Knowledge: In Search of a Modal Basis

In this paper I will offer a novel understanding of a priori knowledge. My claim is that the sharp distinction that is usually made between a priori and a posteriori knowledge is groundless. It will be argued that a plausible understanding of a priori and a posteriori knowledge has to acknowledge that they are in a constant bootstrapping relationship. It is also crucial that we distinguish between a priori propositions that hold in the actual world and merely possible, non-actual a priori propositions, as we will see when considering cases like Euclidean geometry. Furthermore, contrary to what Kripke seems to suggest, a priori knowledge is intimately connected with metaphysical modality, indeed, grounded in it. The task of a priori reasoning, according to this account, is to delimit the space of metaphysically possible worlds in order for us to be able to determine what is actual.

Enticing general engineering undergrads into signal processing with Photo-Editing and image search

The University of Cambridge is unusual in that its Department of Engineering is a single department which covers virtually all branches of engineering under one roof. In their first two years of study, our undergrads study the full breadth of engineering topics and then have to choose a specialization area for the final two years of study. Here we describe part of a course, given towards the end of their second year, which is designed to entice these students to specialize in signal processing and information engineering topics for years 3 and 4. The course is based around a photo editor and an image search application, and it requires no prior knowledge of the z-transform or of 2-dimensional signal processing. It does assume some knowledge of 1-D convolution and basic Fourier methods and some prior exposure to Matlab. The subject of this paper, the photo editor, is written in standard Matlab m-files which are fully visible to the students and help them to see how specific algorithms are implemented in detail. © 2011 IEEE.

Collating the knowledge base for core outcome set development: developing and appraising the search strategy for a systematic review

Background: The COMET (Core Outcome Measures in Effectiveness Trials) Initiative is developing a publicly accessible online resource to collate the knowledge base for core outcome set development (COS) and the applied work from different health conditions. Ensuring that the database is as comprehensive as possible and keeping it up to date are key to its value for users. This requires the development and application of an optimal, multi-faceted search strategy to identify relevant material. This paper describes the challenges of designing and implementing such a search, outlining the development of the search strategy for studies of COS development, and, in turn, the process for establishing a database of COS.

Methods: We investigated the performance characteristics of this strategy including sensitivity, precision and numbers needed to read. We compared the contribution of databases towards identifying included studies to identify the best combination of methods to retrieve all included studies.

Results: Recall of the search strategies ranged from 4% to 87%, and precision from 0.77% to 1.13%. MEDLINE performed best in terms of recall, retrieving 216 (87%) of the 250 included records, followed by Scopus (44%). The Cochrane Methodology Register found just 4% of the included records. MEDLINE was also the database with the highest precision. The number needed to read varied between 89 (MEDLINE) and 130 (SCOPUS).

Conclusions: We found that two databases and hand searching were required to locate all of the studies in this review. MEDLINE alone retrieved 87% of the included studies, but actually 97% of the included studies were indexed on MEDLINE. The Cochrane Methodology Register did not contribute any records that were not found in the other databases, and will not be included in our future searches to identify studies developing COS. SCOPUS had the lowest precision rate (0.77) and highest number needed to read (130). In future COMET searches for COS a balance needs to be struck between the work involved in screening large numbers of records, the frequency of the searching and the likelihood that eligible studies will be identified by means other than the database searches.

Communicating neurons: a connectionist spiking neuron implementation of stochastic diffusion search

An information processing paradigm in the brain is proposed, instantiated in an artificial neural network using biologically motivated temporal encoding. The network will locate within the external world stimulus, the target memory, defined by a specific pattern of micro-features. The proposed network is robust and efficient. Akin in operation to the swarm intelligence paradigm, stochastic diffusion search, it will find the best-fit to the memory with linear time complexity. information multiplexing enables neurons to process knowledge as 'tokens' rather than 'types'. The network illustrates possible emergence of cognitive processing from low level interactions such as memory retrieval based on partial matching. (C) 2007 Elsevier B.V. All rights reserved.

Entity-based semantic search on conversational transcripts semantic

This paper describes the implementation of a semantic web search engine on conversation styled transcripts. Our choice of data is Hansard, a publicly available conversation style transcript of parliamentary debates. The current search engine implementation on Hansard is limited to running search queries based on keywords or phrases hence lacks the ability to make semantic inferences from user queries. By making use of knowledge such as the relationship between members of parliament, constituencies, terms of office, as well as topics of debates the search results can be improved in terms of both relevance and coverage. Our contribution is not algorithmic instead we describe how we exploit a collection of external data sources, ontologies, semantic web vocabularies and named entity extraction in the analysis of underlying semantics of user queries as well as the semantic enrichment of the search index thereby improving the quality of results.