Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.

Mobile learning and video experience in delivering digital content

In recent years, the worldwide distribution of smartphone devices has been growing rapidly. Mobile technologies are evolving fast, a situation which provides new possibilities for mobile learning applications. Along with new delivery methods, this development enables new concepts for learning. This study focuses on the effectiveness and experience of a mobile learning video promoting the key features of a specific device. Through relevant learning theories, mobile technologies and empirical findings, the thesis presents the key elements for a mobile learning video that are essential for effective learning. This study also explores how previous experience with mobile services and knowledge of a mobile handset relate to final learning results. Moreover, this study discusses the optimal delivery mechanisms for a mobile video. The target group for the study consists of twenty employees of a Sanoma Company. The main findings show that the individual experience of learning and the actual learning results may differ and that the design for certain video elements, such as sound and the presentation of technical features, can have an impact on the experience and effectiveness of a mobile learning video. Moreover, a video delivery method based on cloud technologies and HTML5 is suggested to be used in parallel with standalone applications.

Internet-based breast cancer patient’s pathway as an empowering patient educational tool

This three-phase study was conducted to examine the effect of the Breast Cancer Patient’s Pathway program (BCPP) on breast cancer patients’ empowering process from the viewpoint of the difference between knowledge expectations and perceptions of received knowledge, knowledge level, quality of life, anxiety and treatment-related side effects during the breast cancer treatment process. The BCPP is an Internet-based patient education tool describing a flow chart of the patient pathway during the breast treatment process, from breast cancer diagnostic tests to the follow-up after treatments. The ultimate goal of this study was to evaluate the effect of the BCPP to the breast cancer patient’s empowerment by using the patient pathway as a patient education tool. In phase I, a systematic literature review was carried out to chart the solutions and outcomes of Internet-based educational programs for breast cancer patients. In phase II, a Delphi study was conducted to evaluate the usability of web pages and adequacy of their content. In phase III, the BCPP program was piloted with 10 patients and patients were randomised to an intervention group (n=50) and control group (n=48). According to the results of this study, the Internet is an effective patient education tool for increasing knowledge, and BCPP can be used as a patient education method supporting other education methods. However, breast cancer patients’ perceptions of received knowledge were not fulfilled; their knowledge expectations exceed the perceived amount of received knowledge. Although control group patients’ knowledge expectations were met better with the knowledge they received in hospital compared to the patients in the intervention group, no statistical differences were found between the groups in terms of quality of life, anxiety and treatment-related side effects. However, anxiety decreased faster in the intervention group when looking at internal differences between the groups at different measurement times. In the intervention group the relationship between the difference between knowledge expectations and perceptions of received knowledge correlated significantly with quality of life and anxiety. Their knowledge level was also significant higher than in the control group. These results support the theory that the empowering process requires patient’s awareness of knowledge expectations and perceptions of received knowledge. There is a need to develop patient education to meet patients’ perceptions of received knowledge, including oral and written education and BCPP, to fulfil patient’s knowledge expectations and facilitate the empowering process. Further research is needed on the process of cognitive empowerment with breast cancer patients. There is a need for new patient education methods to increase breast cancer patients’ awareness of knowing.

Developing Service-based Solutions for Water Treatment in Finnish Mining Industry

The aim of this master’s thesis is to analyze the mining industry customers' current and future needs for the water treatment services and discover new business development opportunities in the context of mine water treatment. In addition, the study focuses on specifying service offerings needed and evaluate suitable revenue generation models for them. The main research question of the study is: What kind of service needs related to water treatment can be identified in the Finnish mining industry? The literature examined in the study focused on industrial service classification and new service development process as well as the revenue generation of services. A qualitative research approach employing a case study method was chosen for the study. The present study uses customer and expert interviews as primary data source, complemented by archival data. The primary data was gathered by organizing total of 13 interviews, and the interviews were analyzed by using qualitative content analysis. The abductive-logic was chosen as the way of conducting scientific reasoning in this study. As a result, new service proposals were developed for Finnish mine industry suppliers. The main areas of development were on asset efficiency services and process support services. The service needs were strongly associated with suppliers’ know-how of water treatment process optimization, cost-effectiveness as well as on alternative technologies. The study provides an insight for managers that wish to pursue a water treatment services as a part of their business offering.

The nature of communication structures of university students in an international web-based computer-supported collaborative learning (CSCL) course

Communication, the flow of ideas and information between individuals in a social context, is the heart of educational experience. Constructivism and constructivist theories form the foundation for the collaborative learning processes of creating and sharing meaning in online educational contexts. The Learning and Collaboration in Technology-enhanced Contexts (LeCoTec) course comprised of 66 participants drawn from four European universities (Oulu, Turku, Ghent and Ramon Llull). These participants were split into 15 groups with the express aim of learning about computer-supported collaborative learning (CSCL). The Community of Inquiry model (social, cognitive and teaching presences) provided the content and tools for learning and researching the collaborative interactions in this environment. The sampled comments from the collaborative phase were collected and analyzed at chain-level and group-level, with the aim of identifying the various message types that sustained high learning outcomes. Furthermore, the Social Network Analysis helped to view the density of whole group interactions, as well as the popular and active members within the highly collaborating groups. It was observed that long chains occur in groups having high quality outcomes. These chains were also characterized by Social, Interactivity, Administrative and Content comment-types. In addition, high outcomes were realized from the high interactive cases and high-density groups. In low interactive groups, commenting patterned around the one or two central group members. In conclusion, future online environments should support high-order learning and develop greater metacognition and self-regulation. Moreover, such an environment, with a wide variety of problem solving tools, would enhance interactivity.

Sport-based corporate social responsibility in European professional football

In the past few decades, sport has become a major business with remarkable international reach. As part of the commercial sector of sport, professional sport is said to be intrinsically different from other businesses due to its unique characteristics, such as the peculiar economics and the intense loyalty of fans. Simultaneously with the growing business aspect, sport continues to have great social and cultural impacts on our society. Sport has also become an increasingly popular means of attending social problems due to its alleged suitability for such purposes and its popular appeal. A great number of actors in the professional sport industry have long been involved in socially responsible activities, many of which have been sport-related. While Corporate Social Responsibility (CSR) has been extensively studied in general, its role in the professional sport industry has received less attention in the academic research until recently. It has been argued that due to the unique characteristics of professional sport, CSR should also be studied in this particular context. The objective of this study was to contribute to filling the research gap and increase the understanding of CSR in the context of professional sport by examining sport-related CSR realized by professional football clubs in Europe. The theoretical part of this study leaned on previous literature about using sport as a means of attending social issues and the role of CSR in professional sport industry. The empirical part of the study was carried out through web site analyses and interviews. The clubs to be examined were chosen by using purposive sampling technique and taking into consideration the accessibility and suitability of information the clubs could offer. The method used for analyzing the data was qualitative content analysis. The empirical findings were largely in line with the theoretical framework of the study. The sportrelated CSR of the clubs was concentrated on teaching the participants diverse skills and values, improving their health, encouraging social inclusion, supporting disabled people, and promoting overall participation in sport. The clubs also emphasized the importance of local communities as targets of their CSR. CSR had been an integral part of the clubs’ activities from the beginning, but there were remarkable differences between large and small clubs in terms of structured organization and realization of their CSR. Measuring and evaluation of CSR appeared to be a challenge for most clubs regardless of their size and resources. The motives for the clubs to engage in CSR seemed to be related to the clubs’ values or to their stakeholders’ interests. In general, the clubs’ CSR went beyond what the society is likely to expect from them in legal or ethical sense.

Local Features in Image and Video Processing - Object Class Matching and Video Shot Detection

The usage of digital content, such as video clips and images, has increased dramatically during the last decade. Local image features have been applied increasingly in various image and video retrieval applications. This thesis evaluates local features and applies them to image and video processing tasks. The results of the study show that 1) the performance of different local feature detector and descriptor methods vary significantly in object class matching, 2) local features can be applied in image alignment with superior results against the state-of-the-art, 3) the local feature based shot boundary detection method produces promising results, and 4) the local feature based hierarchical video summarization method shows promising new new research direction. In conclusion, this thesis presents the local features as a powerful tool in many applications and the imminent future work should concentrate on improving the quality of the local features.

Assessment of Young Learners’ English Proficiency in Bilingual Content Instruction CLIL

This applied linguistic study in the field of second language acquisition investigated the assessment practices of class teachers as well as the challenges and visions of language assessment in bilingual content instruction (CLIL) at primary level in Finnish basic education. Furthermore, pupils’ and their parents’ perceptions of language assessment and LangPerform computer simulations as an alternative, modern assessment method in CLIL contexts were examined. The study was conducted for descriptive and developmental purposes in three phases: 1) a CLIL assessment survey; 2) simulation 1; and 3) simulation 2. All phases had a varying number of participants. The population of this mixed methods study were CLIL class teachers, their pupils and the pupils’ parents. The sampling was multi-staged and based on probability and random sampling. The data were triangulated. Altogether 42 CLIL class teachers nationwide, 109 pupils from the 3rd, 4th and 5th grade as well as 99 parents from two research schools in South-Western Finland participated in the CLIL assessment survey followed by an audio-recorded theme interview of volunteers (10 teachers, 20 pupils and 7 parents). The simulation experimentations 1 and 2 produced 146 pupil and 39 parental questionnaires as well as video interviews of volunteered pupils. The data were analysed both quantitatively using percentages and numerical frequencies and qualitatively employing thematic content analysis. Based on the data, language assessment in primary CLIL is not an established practice. It largely appears to be infrequent, incidental, implicit and based on impressions rather than evidence or the curriculum. The most used assessment methods were teacher observation, bilingual tests and dialogic interaction, and the least used were portfolios, simulations and peer assessment. Although language assessment was generally perceived as important by teachers, a fifth of them did not gather assessment information systematically, and 38% scarcely gave linguistic feedback to pupils. Both pupils and parents wished to receive more information on CLIL language issues; 91% of pupils claimed to receive feedback rarely or occasionally, and 63% of them wished to get more information on their linguistic coping in CLIL subjects. Of the parents, 76% wished to receive more information on the English proficiency of their children and their linguistic development. This may be a response to indirect feedback practices identified in this study. There are several challenges related to assessment; the most notable is the lack of a CLIL curriculum, language objectives and common ground principles of assessment. Three diverse approaches to language in CLIL that appear to affect teachers’ views on language assessment were identified: instrumental (language as a tool), dual (language as a tool and object of learning) and eclectic (miscellaneous views, e.g. affective factors prioritised). LangPerform computer simulations seem to be perceived as an appropriate alternative assessment method in CLIL. It is strongly recommended that the fundamentals for assessment (curricula and language objectives) and a mutual assessment scheme should be determined and stakeholders’ knowledge base of CLIL strengthened. The principles of adequate assessment in primary CLIL are identified as well as several appropriate assessment methods suggested.

Model-Based Tracking Initialization in Ship Building Environment

An augmented reality (AR) device must know observer’s location and orientation, i.e. observer’s pose, to be able to correctly register the virtual content to observer’s view. One possible way to determine and continuously follow-up the pose is model-based visual tracking. It supposes that a 3D model of the surroundings is known and that there is a video camera that is fixed to the device. The pose is tracked by comparing the video camera image to the model. Each new pose estimate is usually based on the previous estimate. However, the first estimate must be found out without a prior estimate, i.e. the tracking must be initialized, which in practice means that some model features must be identified from the image and matched to model features. This is known in literature as model-to-image registration problem or simultaneous pose and correspondence problem. This report reviews visual tracking initialization methods that are suitable for visual tracking in ship building environment when the ship CAD model is available. The environment is complex, which makes the initialization non-trivial. The report has been done as part of MARIN project.

Repository content - lost in Discovery?

Presentation at Open Repositories 2014, Helsinki, Finland, June 9-13, 2014

Packaging Content for Aggregated and Federated Repositories through APTrust

Presentation at Open Repositories 2014, Helsinki, Finland, June 9-13, 2014

Service oriented archive based on Fedora Commons

Presentation at Open Repositories 2014, Helsinki, Finland, June 9-13, 2014

Biomedical Event Extraction with Machine Learning

Biomedical natural language processing (BioNLP) is a subfield of natural language processing, an area of computational linguistics concerned with developing programs that work with natural language: written texts and speech. Biomedical relation extraction concerns the detection of semantic relations such as protein-protein interactions (PPI) from scientific texts. The aim is to enhance information retrieval by detecting relations between concepts, not just individual concepts as with a keyword search. In recent years, events have been proposed as a more detailed alternative for simple pairwise PPI relations. Events provide a systematic, structural representation for annotating the content of natural language texts. Events are characterized by annotated trigger words, directed and typed arguments and the ability to nest other events. For example, the sentence “Protein A causes protein B to bind protein C” can be annotated with the nested event structure CAUSE(A, BIND(B, C)). Converted to such formal representations, the information of natural language texts can be used by computational applications. Biomedical event annotations were introduced by the BioInfer and GENIA corpora, and event extraction was popularized by the BioNLP'09 Shared Task on Event Extraction. In this thesis we present a method for automated event extraction, implemented as the Turku Event Extraction System (TEES). A unified graph format is defined for representing event annotations and the problem of extracting complex event structures is decomposed into a number of independent classification tasks. These classification tasks are solved using SVM and RLS classifiers, utilizing rich feature representations built from full dependency parsing. Building on earlier work on pairwise relation extraction and using a generalized graph representation, the resulting TEES system is capable of detecting binary relations as well as complex event structures. We show that this event extraction system has good performance, reaching the first place in the BioNLP'09 Shared Task on Event Extraction. Subsequently, TEES has achieved several first ranks in the BioNLP'11 and BioNLP'13 Shared Tasks, as well as shown competitive performance in the binary relation Drug-Drug Interaction Extraction 2011 and 2013 shared tasks. The Turku Event Extraction System is published as a freely available open-source project, documenting the research in detail as well as making the method available for practical applications. In particular, in this thesis we describe the application of the event extraction method to PubMed-scale text mining, showing how the developed approach not only shows good performance, but is generalizable and applicable to large-scale real-world text mining projects. Finally, we discuss related literature, summarize the contributions of the work and present some thoughts on future directions for biomedical event extraction. This thesis includes and builds on six original research publications. The first of these introduces the analysis of dependency parses that leads to development of TEES. The entries in the three BioNLP Shared Tasks, as well as in the DDIExtraction 2011 task are covered in four publications, and the sixth one demonstrates the application of the system to PubMed-scale text mining.

Uraohjaus ja urasuunnittelu ammattikorkeakoulussa

Tutkimuksen tarkoituksena oli selvittää, millaista uraohjausta ammattikorkeakoulun tuutoriopettajat antavat ja millaista uraohjausta opiskelijat haluavat. Lisäksi tavoitteena oli selvittää, löytyykö opiskelijoiden koulutusalavalinnan perusteista yhteyttä uran suunnittelutaitoihin ja ohjauksen tarpeeseen, ja tunnistavatko tuutoriopettajat opiskelijoiden erilaiset uraohjauksen tarpeet. Tutkimuksen teoreettisissa rakenteissa hyödynnettiin kolmea postmodernia urateoriaa, jotka olivat Hodkinsonin ja Sparkesin (1997) uranvalinnan päätöksentekoteoria, Mitchellin, Lewinin ja Krumbolzin (1999) suunnitellun sattuman teoria ja Savickasin (2005) uran rakentamisteoria. Tutkimusympäristönä oli Satakunnan ammattikorkeakoulu. Tutkimus oli kaksivaiheinen. Ensimmäisessä vaiheessa kerättiin harkinnanvaraisesti valituilta tuutoriopettajilta (n=14) ja opintojensa eri vaiheissa olevilta opiskelijoilta (n=65) kirjoitettu aineisto. Kvalitatiivinen aineisto analysoitiin sisällönanalyysillä. Aineiston perusteella löydettiin kolmenlaisia urasuunnittelijoita: epävarmat, uteliaat ja tietoiset. Aineiston perusteella laadittiin kyselylomake tutkimuksen toisen vaiheen tiedonkeruuta varten. Tutkimuksen toisessa vaiheessa kerättiin opintojen eri vaiheissa olevilta opiskelijoilta kyselylomakekyselynä kvantitatiivinen aineisto (n=903), joka analysoitiin tilastollisin menetelmin. Koulutusalavalinnan perusteista elämäntilanne, alan mahdollisuudet, oma toive, kutsumus, aktiivinen tiedonhaku ja halu opiskella ammattikorkeakoulussa olivat yhteydessä opiskelijan hyvään urasuunnittelukykyyn. Näillä perusteilla koulutusalansa valinneita tietoisiksi luokiteltuja urasuunnittelijoita oli 72 % vastanneista. Alavalinnan perusteista sattuman, kavereiden, sukulaisten, lukion opinto-ohjauksen ja paikkakunnan perusteella koulutusalansa valinneet luokiteltiin epävarmoiksi urasuunnittelijoiksi, ja heitä oli 28 % vastanneista. Tulokset antavat ohjaajille tukea epävarman ja muita enemmän uraohjausta tarvitsevan opiskelijan tunnistamiseen ja heidän hops-prosessinsa tehostamiseen opintojen alusta asti. Lisäksi tulosten perusteella esitetään seuraavia suosituksia: tuutoriopettajille tulisi asettaa pätevyysvaatimukseksi ohjausalan opintojen suorittaminen; opiskelijoita tulisi ohjata tunnistamaan erilaisia satunnaisesti avautuvia mahdollisuuksia ja tietoisesti hyödyntämään niitä elämässään; uraohjaukseen tulisi kytkeä mukaan työelämäyhteistyö; ohjaajien tulisi tiivistää yhteistyötä toisen asteen ohjaajien kanssa, jotta opiskelijoiden koulutusalavalinnat onnistuisivat paremmin; uraohjausta tulisi antaa tulevaisuuden kvalifikaatioiden ennakoinnin ja elinikäisten oppimisvalmiuksien näkökulmasta.

How to increase brand quality perception with audiovisual content? : Exploring the role of emotions in B2C and B2B contexts

This study discusses how audiovisual content can influence brand quality perceptions. The purpose of this study is to explore how audiovisual content creation can increase brand quality perceptions. This research problem is addressed with three sub questions, which aim at clarifying the role of emotions between content marketing and brand quality perception, explaining how different functions of audiovisual content can increase brand quality perception, and by identifying and comparing the key differences in content creation in business-to-consumer and business-to-businesscontexts. The theoretical background of the study is in brand personality, consumer emotions, consumerbrand relationships, content marketing and B2B branding literature. The empirical research part includes a single-case study. The case company was a Swiss startup that wished to build a highquality brand for both B2C and B2B segments. The empirical data was collected in September 2014. Eight interviews were conducted; seven with target segment representatives and one with an existing customer of the case company. The empirical findings were analyzed with thematic analysis and finally a 5-stage framework was created based on the findings of the research, offering a guideline for high-quality content creation. This study finds that emotions play an important role in brand quality perceptions. Psychological processes, emotion, cognition and conation, influence the engagement process of the target segment which ultimately can lead to activation and electronic word-of-mouth. Brand quality perception is the result of the overall emotion of the brand. The overall emotion derives from brand personality, brand concept, product attributes and utilitarian benefits of the brand. The entertaining and educational functions of the audiovisual content can target and evoke these emotional processes, and result in increased quality perceptions. In the B2B context, emotions are found to play a relatively smaller role in the quality perception processes. However, the significance of emotions cannot be ignored, since they can emphasize the value for the buying organization, and build on the trust and loyalty among the potential customers. The final framework presents five stages of content creation that ultimately improve brand quality perceptions. These stages help marketers to design and implement their content and evoke positive emotions in their target segment as part of a quality-based marketing strategy. Further research is warranted to quantitatively test the generalizability of the framework. Further research is also suggested to make the framework adaptable to different stages of the brand life cycle.