In this thesis, we plan to use a group-based semantic-expansion approach to design a new personalised system framework. Semantic web and group preference offer solution to the above problem. In this thesis, ontologies and semantic techniques are applied in different components of the framework. Information has been gathered from different resources and each of the resource might be using various types of identifiers for the same concept, therefore semantic web technologies are used to find out if the concept is the same or not. On the other hand, we create group preference in our personalization system. If the system fails to obtain personal preference from new user, group preference supports the system providing recommendation to the new user according to group classification.

The explosive growth of the World-Wide-Web and the emergence of ecommerce are the major two factors that have led to the development of recommender systems (Resnick and Varian, 1997). The main task of recommender systems is to learn from users and recommend items (e.g. information, products or books) that match the users’ personal preferences. Recommender systems have been an active research area for more than a decade. Many different techniques and systems with distinct strengths have been developed to generate better quality recommendations. One of the main factors that affect recommenders’ recommendation quality is the amount of information resources that are available to the recommenders. The main feature of the recommender systems is their ability to make personalised recommendations for different individuals. However, for many ecommerce sites, it is difficult for them to obtain sufficient knowledge about their users. Hence, the recommendations they provided to their users are often poor and not personalised. This information insufficiency problem is commonly referred to as the cold-start problem. Most existing research on recommender systems focus on developing techniques to better utilise the available information resources to achieve better recommendation quality. However, while the amount of available data and information remains insufficient, these techniques can only provide limited improvements to the overall recommendation quality. In this thesis, a novel and intuitive approach towards improving recommendation quality and alleviating the cold-start problem is attempted. This approach is enriching the information resources. It can be easily observed that when there is sufficient information and knowledge base to support recommendation making, even the simplest recommender systems can outperform the sophisticated ones with limited information resources. Two possible strategies are suggested in this thesis to achieve the proposed information enrichment for recommenders: • The first strategy suggests that information resources can be enriched by considering other information or data facets. Specifically, a taxonomy-based recommender, Hybrid Taxonomy Recommender (HTR), is presented in this thesis. HTR exploits the relationship between users’ taxonomic preferences and item preferences from the combination of the widely available product taxonomic information and the existing user rating data, and it then utilises this taxonomic preference to item preference relation to generate high quality recommendations. • The second strategy suggests that information resources can be enriched simply by obtaining information resources from other parties. In this thesis, a distributed recommender framework, Ecommerce-oriented Distributed Recommender System (EDRS), is proposed. The proposed EDRS allows multiple recommenders from different parties (i.e. organisations or ecommerce sites) to share recommendations and information resources with each other in order to improve their recommendation quality. Based on the results obtained from the experiments conducted in this thesis, the proposed systems and techniques have achieved great improvement in both making quality recommendations and alleviating the cold-start problem.

Advertisers setting up search engine advertising campaigns for the first time need to place bids on keywords, but typically lack experience and data to determine ranks that maximize a keyword's profit (generally referred to as a cold-start problem). This article aims at solving the problem of bidding on keywords in newly set-up search engine advertising campaigns. We suggest that advertisers collect data from the Google Keyword Planner to obtain precise estimates of the percentage increases in prices per click and clickthrough rates, which are needed to calculate optimal bids (exact approach). Together with the profit contribution per conversion and the conversion rate, the advertiser might then set bids that maximize profit. In case advertisers cannot afford to collect the required data, we suggest two proxy approaches and evaluate their performance using the exact approach as a benchmark. The empirical study shows that both proxy approaches perform reasonably well-the easier approach to implement (proxy 2) sometimes performs even better than the more sophisticated one (proxy 1). As a consequence, advertisers might just use this very simple proxy when bidding on keywords in newly set-up SEA campaigns. This research extends the stream of literature on how to determine optimal bids, which so far focuses on campaigns that are already running and where the required data to calculate bids is already available. This research offers a novel approach of determining bids when advertisers lack the aforementioned information.

The item cold-start problem, which describes the difficulty of recommendation systems in recommending new items to users, remains a great challenge for recommendation systems that rely on past user-item interaction data. A popular technique in the current research surrounding the cold-start problem is the use of hybrid models that combine two or more recommendation strategies that may contribute with their individual advantages. This thesis investigates the use of a hybrid model which combines Sentence BERT embeddings with a recommendation model based on Factorization Machines (FM). The research question is stated as: How does a hybrid recommendation system based on Factorization Machines with frozen Sentence BERT embeddings perform in terms of solving the cold-start problem?. Three experiments were conducted to answer the research question. These involved finding an optimal pre-trained Sentence BERT model, investigating the difference in performance between an FM-model and a hybrid FM-model, as well as the difference in ranking of an item depending on whether or not the hybrid FM-model has been trained on the item. The results show that the best pre-trained Sentence BERT model for producing meaningful embeddings is the paraphrase-MiniLM-L3-v2 model, that a hybrid FM-model and a standard FM-model perform almost equally in terms of precision and recall at 50, and that there is a weak correlation between the item-frequency and how the hybrid FM-model ranks an item when trained and not trained on the item. The answer to the research question is that a recommendation model based on Factorization Machines with frozen Sentence BERT embeddings displays low precision at 50 and recall at 50 values with the given parameters in comparison to the values given in an optimal recommendation scenario. The hybrid FM-model shows cold-start potential due to displaying similar results to the standard FM-model, but these values are so low that further investigation with other parameters is needed for a clearer conclusion. / Kallstartsproblem för artiklar som beskriver svårigheten hos rekommendationssystem gällande uppgiften att rekommendera nya artiklar till användare, är fortsatt en stor utmaning för rekommendationssystem som förlitar sig på data som beskriver interaktioner mellan användare och artiklar. En populär teknik inom den aktuella forskningen gällande kallstartsproblemet är användandet av hybridmodeller som kombinerar två eller flera rekommendationsstrategier och som potentiellt kan bidra med sina individuella fördelar. Detta examensarbete undersöker användandet av en hybridmodell som kombinerar menings-BERT inbäddningar med en rekommendationsmodell baserad på faktoriseringsmaskiner (FM). Frågeställningen lyder: Hur väl kan kallstartsproblemet för rekommendationer lösas med en hybridmodell baserad på faktoriseringsmaskiner med frusna menings-BERT-inbäddningar?. Tre experiment utfördes för att svara på frågeställningen. Dessa experiment innebar att hitta en optimal förtränad menings-BERT-modell, undersöka skillnaden i utförandet mellan en FM-modell och en hybrid FM-modell, samt skillnaden i ranking av en artikel baserat på huruvida hybridmodellen tränats eller inte tränats på artikeln. Resultaten visar att den bästa förtränade menings-BERT-modellen gällande skapandet av meningsfulla inbäddningar är paraphrase-MiniLM-L3-v2-modellen, att en hybrid FM-modell och en FM-modell genererar nästan identiska resultat baserat på precision och återkallelse för de första 50 resultaten och att det finns en svag korrelation mellan artikel-frekvens och hur hybridmodellen rankar en artikel när hybridmodellen tränats eller inte tränats på artikeln. Svaret på frågeställningen är att en hybrid FM-modell med frusna menings-BERT-inbäddningar visar låga resultat för precision och återkallelse för de första 50 resultaten givet de använda parametrarna jämfört med de värden som skulle genererats i ett optimalt rekommendationsscenario. Den hybrida FM-modellen visar kallstartspotential då den visar liknande resultat som FM-modellen, men dessa värden är så låga att frågan behöver undersökas ytterligare för tydligare resultat.

User contributions are vital to online communities; therefore it is important to know how to motivate user participation to ensure flow and quality of content, and to generate more traffic and revenue to community owners. In contrast to previous research which has explored the motivations of participants in already existing online communities, I investigate whether a new niche online community with a particular focus(women in Science and Engineering sharing their personal experiences through stories) can be started through a design that follows best practices for community design and principles derived from theories of motivation. The design of the WISETales community is based upon insights from literature in three main areas: social psychology, computer science, and gender studies. A social visualization which serves informational, navigational and motivational tool was also developed. One pilot study and two exploratory studies were carried out to evaluate the need for such a community, its design and interface usability. The design of the community and visualization, along with the results from the studies, their analysis and discussion are presented in the thesis.

User contributions are vital to online communities; therefore it is important to know how to motivate user participation to ensure flow and quality of content, and to generate more traffic and revenue to community owners. In contrast to previous research which has explored the motivations of participants in already existing online communities, I investigate whether a new niche online community with a particular focus(women in Science and Engineering sharing their personal experiences through stories) can be started through a design that follows best practices for community design and principles derived from theories of motivation. The design of the WISETales community is based upon insights from literature in three main areas: social psychology, computer science, and gender studies. A social visualization which serves informational, navigational and motivational tool was also developed. One pilot study and two exploratory studies were carried out to evaluate the need for such a community, its design and interface usability. The design of the community and visualization, along with the results from the studies, their analysis and discussion are presented in the thesis.