Dissertations / Theses on the topic 'Content search'

Document summarization has been an intriguing task of Computational linguistics. A number of definitions have been proposed in References, all of which consider document summarization as a problem of text compression. One of the most complete definitions by Sparck-Jones states that " ... a summary is a reductive transformation of source text to summary text through content condensation by selection and/or generalisation on what is important in the source ... ". The importance of document summarization does not lie only in presenting information in a shortened form, but also in selecting the most appropriate content to present. Moreover, a main feature in summarization is the number of sources from which a summary may be produced; thus, single-document and multi -document have been proposed, denoting the number of sources from which the summary will be produced. In addition, another categorization that may be extracted from this definition refers to the importance of the source, and what the potential user thinks is important. This leads to the definition of generic and query-based or task focused summarization, where generic implies that the summarizer should extract information according to the main topics discussed in the document, while query-based summarization focuses on extracting information according to simple or more complex questions on the document. Moreover, importance of content can be extracted through knowledge-rich (supervised and semi- supervised summarization) and knowledge lean approaches (unsupervised or shallow summarization). The last categorization refers to the type generation of the summary, the two main categories being: extractive summarization, where sentences are maintained in the summarization process unaltered; and abstraction, where the sentences are either semantically altered or compressed. The research depicted in this thesis, presents novel document summarization approaches based on the theories of Machine Learning (ML) and Natural Language Processing (NLP) for generic single-document extractive summarization. The motivation to target on Greek langaguage came from the lack of a Greek summarization system. Most notably, only one system for Greek Summarization system exists in the literature (GreekSum). The research undertaken resulted in: the development of a stemming algorithm used for noun and adjective identification, based on grammatical analysis on Greek language; the development of a novel statistical classification scheme, initially aimed to document summarization, that is proven to outperform other statistical summarizers as Narve Bayes Classifier (NBC) and Language Models (LM); the development of a supervised statistical summarization algorithm based on document classification techniques (Text Classification Assisted Summarization for Greek Language-TCASGL); and the development of a knowledge-lean summarization algorithm (Generic Unsupervised Text Summarization - GUTS), using shallow semantic document analysis and statistics. The results demonstrate that the classification algorithm significantly outperforms widely available statistical algorithms, while the ML approach yielded comparable results to other supervised systems. In addition to that, GUTS was shown to perform equally well with knowledge rich approaches.

Search engines have forever changed the way people access and discover knowledge, allowing information about almost any subject to be quickly and easily retrieved within seconds. As increasingly more material becomes available electronically the influence of search engines on our lives will continue to grow. This presents the problem of how to find what information is contained in each search engine, what bias a search engine may have, and how to select the best search engine for a particular information need. This research introduces a new method, search engine content analysis, in order to solve the above problem. Search engine content analysis is a new development of traditional information retrieval field called collection selection, which deals with general information repositories. Current research in collection selection relies on full access to the collection or estimations of the size of the collections. Also collection descriptions are often represented as term occurrence statistics. An automatic ontology learning method is developed for the search engine content analysis, which trains an ontology with world knowledge of hundreds of different subjects in a multilevel taxonomy. This ontology is then mined to find important classification rules, and these rules are used to perform an extensive analysis of the content of the largest general purpose Internet search engines in use today. Instead of representing collections as a set of terms, which commonly occurs in collection selection, they are represented as a set of subjects, leading to a more robust representation of information and a decrease of synonymy. The ontology based method was compared with ReDDE (Relevant Document Distribution Estimation method for resource selection) using the standard R-value metric, with encouraging results. ReDDE is the current state of the art collection selection method which relies on collection size estimation. The method was also used to analyse the content of the most popular search engines in use today, including Google and Yahoo. In addition several specialist search engines such as Pubmed and the U.S. Department of Agriculture were analysed. In conclusion, this research shows that the ontology based method mitigates the need for collection size estimation.

Search engines have forever changed the way people access and discover knowledge, allowing information about almost any subject to be quickly and easily retrieved within seconds. As increasingly more material becomes available electronically the influence of search engines on our lives will continue to grow. This presents the problem of how to find what information is contained in each search engine, what bias a search engine may have, and how to select the best search engine for a particular information need. This research introduces a new method, search engine content analysis, in order to solve the above problem. Search engine content analysis is a new development of traditional information retrieval field called collection selection, which deals with general information repositories. Current research in collection selection relies on full access to the collection or estimations of the size of the collections. Also collection descriptions are often represented as term occurrence statistics. An automatic ontology learning method is developed for the search engine content analysis, which trains an ontology with world knowledge of hundreds of different subjects in a multilevel taxonomy. This ontology is then mined to find important classification rules, and these rules are used to perform an extensive analysis of the content of the largest general purpose Internet search engines in use today. Instead of representing collections as a set of terms, which commonly occurs in collection selection, they are represented as a set of subjects, leading to a more robust representation of information and a decrease of synonymy. The ontology based method was compared with ReDDE (Relevant Document Distribution Estimation method for resource selection) using the standard R-value metric, with encouraging results. ReDDE is the current state of the art collection selection method which relies on collection size estimation. The method was also used to analyse the content of the most popular search engines in use today, including Google and Yahoo. In addition several specialist search engines such as Pubmed and the U.S. Department of Agriculture were analysed. In conclusion, this research shows that the ontology based method mitigates the need for collection size estimation.

The thesis investigates how content produced by end-users on the World Wide Web — referred to as user-generated content — can enhance the news vertical aspect of a universal Web search engine, such that news-related queries can be satisfied more accurately, comprehensively and in a more timely manner. We propose a news search framework to describe the news vertical aspect of a universal web search engine. This framework is comprised of four components, each providing a different piece of functionality. The Top Events Identification component identifies the most important events that are happening at any given moment using discussion in user-generated content streams. The News Query Classification component classifies incoming queries as news-related or not in real-time. The Ranking News-Related Content component finds and ranks relevant content for news-related user queries from multiple streams of news and user-generated content. Finally, the News-Related Content Integration component merges the previously ranked content for the user query into theWeb search ranking. In this thesis, we argue that user-generated content can be leveraged in one or more of these components to better satisfy news-related user queries. Potential enhancements include the faster identification of news queries relating to breaking news events, more accurate classification of news-related queries, increased coverage of the events searched for by the user or increased freshness in the results returned. Approaches to tackle each of the four components of the news search framework are proposed, which aim to leverage user-generated content. Together, these approaches form the news vertical component of a universal Web search engine. Each approach proposed for a component is thoroughly evaluated using one or more datasets developed for that component. Conclusions are derived concerning whether the use of user-generated content enhances the component in question using an appropriate measure, namely: effectiveness when ranking events by their current importance/newsworthiness for the Top Events Identification component; classification accuracy over different types of query for the News Query Classification component; relevance of the documents returned for the Ranking News-Related Content component; and end-user preference for rankings integrating user-generated content in comparison to the unalteredWeb search ranking for the News-Related Content Integration component. Analysis of the proposed approaches themselves, the effective settings for the deployment of those approaches and insights into their behaviour are also discussed. In particular, the evaluation of the Top Events Identification component examines how effectively events — represented by newswire articles — can be ranked by their importance using two different streams of user-generated content, namely blog posts and Twitter tweets. Evaluation of the proposed approaches for this component indicates that blog posts are an effective source of evidence to use when ranking events and that these approaches achieve state-of-the-art effectiveness. Using the same approaches instead driven by a stream of tweets, provide a story ranking performance that is significantly more effective than random, but is not consistent across all of the datasets and approaches tested. Insights are provided into the reasons for this with regard to the transient nature of discussion in Twitter. Through the evaluation of the News Query Classification component, we show that the use of timely features extracted from different news and user-generated content sources can increase the accuracy of news query classification over relying upon newswire provider streams alone. Evidence also suggests that the usefulness of the user-generated content sources varies as news events mature, with some sources becoming more influential over time as new content is published, leading to an upward trend in classification accuracy. The Ranking News-Related Content component evaluation investigates how to effectively rank content from the blogosphere and Twitter for news-related user queries. Of the approaches tested, we show that learning to rank approaches using features specific to blog posts/tweets lead to state-of-the-art ranking effectiveness under real-time constraints. Finally this thesis demonstrates that the majority of end-users prefer rankings integrated with usergenerated content for news-related queries to rankings containing only Web search results or integrated with only newswire articles. Of the user-generated content sources tested, the most popular source is shown to be Twitter, particularly for queries relating to breaking events. The central contributions of this thesis are the introduction of a news search framework, the approaches to tackle each of the four components of the framework that integrate user-generated content and their subsequent evaluation in a simulated real-time setting. This thesis draws insights from a broad range of experiments spanning the entire search process for news-related queries. The experiments reported in this thesis demonstrate the potential and scope for enhancements that can be brought about by the leverage of user-generated content for real-time news search and related applications.

In this thesis, I propose a method for establishing a personalized recommendation system for re-ranking web search results and recommending web contents. The method is based on personal reading interest which can be reflected by the user’s dwell time on each document or webpage. I acquire document-level dwell times via a customized web browser, or a mobile device. To obtain better precision, I also explore the possibility of tracking gaze position and facial expression, from which I can determine the attractiveness of different parts of a document. Inspired by idea of Google Knowledge Graph, I also establish a graph-based ontology to maintain a user profile to describe the user’s personal reading interest. Each node in the graph is a concept, which represents the user’s potential interest on this concept. I also use the dwell time to measure concept-level interest, which can be inferred from document-level user dwell times. The graph is generated based on the Wikipedia. According to the estimated concept-level user interest, my algorithm can estimate a user’s potential dwell time over a new document, based on which personalized webpage re-ranking can be carried out. I compare the rankings produced by my algorithm with rankings generated by popular commercial search engines and a recently proposed personalized ranking algorithm. The results clearly show the superiority of my method. I also use my personalized recommendation framework in other applications. A good example is personalized document summarization. The same knowledge graph is employed to estimate the weight of every word in a document; combining with a traditional document summarization algorithm which focused on text mining, I could generate a personalized summary which emphasize the user’s interest in the document. To deal with images and videos, I present a new image search and ranking algorithm for retrieving unannotated images by collaboratively mining online search results, which consists of online images and text search results. The online image search results are leveraged as reference examples to perform content-based image search over unannotated images. The online text search results are used to estimate individual reference images’ relevance to the search query as not all the online image search results are closely related to the query. Overall, the key contribution of my method lies in its ability to deal with unreliable online image search results through jointly mining visual and textual aspects of online search results. Through such collaborative mining, my algorithm infers the relevance of an online search result image to a text query. Once I estimate a query relevance score for each online image search result, I can selectively use query specific online search result images as reference examples for retrieving and ranking unannotated images. To explore the performance of my algorithm, I tested it both on a standard public image datasets and several modestly sized personal photo collections. I also compared the performance of my method with that of two peer methods. The results are very positive, which indicate that my algorithm is superior to existing content-based image search algorithms for retrieving and ranking unannotated images. Overall, the main advantage of my algorithm comes from its collaborative mining over online search results both in the visual and the textual domains.<br>published_or_final_version<br>Computer Science<br>Doctoral<br>Doctor of Philosophy

Abstract Growth in storage capacity has led to large digital video repositories and complicated the discovery of specific information without the laborious manual annotation of data. The research focuses on creating a retrieval system that is ultimately independent of manual work. To retrieve relevant content, the semantic gap between the searcher's information need and the content data has to be overcome using content-based technology. Semantic gap constitutes of two distinct elements: the ambiguity of the true information need and the equivocalness of digital video data. The research problem of this thesis is: what computational content-based models for retrieval increase the effectiveness of the semantic retrieval of digital video? The hypothesis is that semantic search performance can be improved using pattern recognition, data abstraction and clustering techniques jointly with human interaction through manually created queries and visual browsing. The results of this thesis are composed of: an evaluation of two perceptually oriented colour spaces with details on the applicability of the HSV and CIE Lab spaces for low-level feature extraction; the development and evaluation of low-level visual features in example-based retrieval for image and video databases; the development and evaluation of a generic model for simple and efficient concept detection from video sequences with good detection performance on large video corpuses; the development of combination techniques for multi-modal visual, concept and lexical retrieval; the development of a cluster-temporal browsing model as a data navigation tool and its evaluation in several large and heterogeneous collections containing an assortment of video from educational and historical recordings to contemporary broadcast news, commercials and a multilingual television broadcast. The methods introduced here have been found to facilitate semantic queries for novice users without laborious manual annotation. Cluster-temporal browsing was found to outperform the conventional approach, which constitutes of sequential queries and relevance feedback, in semantic video retrieval by a statistically significant proportion.

Aquesta tesi tracta de cercadors d'audio basats en contingut. Específicament, tracta de desenvolupar tecnologies que permetin fer més estret l'interval semàntic o --semantic gap' que, a avui dia, limita l'ús massiu de motors de cerca basats en contingut. Els motors de cerca d'àudio fan servir metadades, en la gran majoria generada per editors, per a gestionar col.leccions d'àudio. Tot i ser una tasca àrdua i procliu a errors, l'anotació manual és la pràctica més habitual. Els mètodes basats en contingut àudio, és a dir, aquells algorismes que extreuen automàticament etiquetes descriptives de fitxers d'àudio, no són generalment suficientment madurs per a permetre una interacció semàntica. En la gran majoria, els mètodes basats en contingut treballen amb descriptors de baix nivell, mentre que els descriptors d'alt nivell estan més enllà de les possibilitats actuals. En la tesi explorem mètodes, que considerem pas previs per a atacar l'interval semàntic.<br>This dissertation is about audio content-based search. Specifically, it is on developing technologies for bridging the semantic gap that currently prevents wide-deployment of audio content-based search engines.<br/>Audio search engines rely on metadata, mostly human generated, to manage collections of audio assets.<br/>Even though time-consuming and error-prone, human labeling is a common practice.<br/>Audio content-based methods, algorithms that automatically extract description from audio files, are generally not mature enough to provide a user friendly representation for interacting with audio content. Mostly, content-based methods are based on low-level descriptions, while high-level or semantic descriptions are beyond current capabilities. In this thesis we explore technologies that can help close the semantic gap.

Inefficiencies associated with online information search are becoming increasingly prevalent in digital environments due to a surge in Consumer Generated Content (CGC). Despite growing scholarly interest in investigating users' information search behavior in CGC environments, there is a paucity of studies that explores the phenomenon from a theory-guided angle. Drawing on Information Foraging Theory (IFT), we re-conceptualize online information search as a form of adaptive user behavior in response to system design constraints. Through this theoretical lens, we advance separate taxonomies for online information search tactics and strategies, both of which constitute essential building blocks of the search process. Furthermore, we construct a research framework that bridges the gap between online information search tactics and strategies by articulating how technology-enabled search tactics contribute to the fulfillment of strategic search goals. We validate our research framework via an online experiment by recruiting participants from Amazon Mechanical Turk (AMT). Participants were tasked to perform searches on custom-developed online review websites, which were modeled after a popular online review website and populated with real restaurant review data. Empirical findings reveal that the provision of different search features indeed engenders distinct search tactics, thereby allowing users varying levels of search determination control and search manipulation control. In turn, both types of search controls affects users' result anticipation and search costs, which when combined, determine the efficiency of goal-oriented search strategy and the utility of exploratory search strategy. This study provides valuable insights that can guide future research and practice.

Through a little investigation by file types it is possible to easily find that one of the most popular search engines has in its indexes about 10 billion of images. Even considering that this data is probably an underestimate of the real number, however, immediately it gives us an idea of how the images are a key component in human communication. This so exorbitant number puts us in the face of the enormous difficulties encountered when one has to deal with them. Until now, the images have always been accompanied by textual data: description, tags, labels, ... which are used to retrieve them fromthe archives. However it is clear that their increase, occurred in recent years, does not allow this type cataloguing. Furthermore, for its own nature, a manual cataloguing is subjective, partial and without doubt subject to error. To overcome this situation in recent years it has gotten a footing a kind of search based on the intrinsic characteristics of images such as colors and shapes. This information is then converted into numerical vectors, and through their comparison it is possible to find images that have similar characteristics. It is clear that a search, on this level of representation of the images, is far from the user perception that of the images. To allow the interaction between users and retrieval systems and improve the performance, it has been decided to involve the user in the search allowing to him to give a feedback of relevance of the images retrieved so far. In this the kind of image that are interesting for user can be learnt by the system and an improvement in the next iteration can be obtained. These techniques, although studied for many years, still present open issues. High dimensional feature spaces, lack of relevant training images, and feature spaceswith lowdiscriminative capability are just some of the problems encountered. In this thesis these problems will be faced by proposing some innovative solutions both to improve performance obtained by methods proposed in the literature, and to provide to retrieval systems greater generalization capability. Techniques of data fusion, both at the feature space level and at the level of different retrieval techniques, will be presented, showing that the former allow greater discriminative capability while the latter provide more robustness to the system. To overcome the lack of images of training it will be proposed a method to generate synthetic patterns allowing in this way a more balanced learning. Finally, new methods to measure similarity between images and to explore more efficiently the feature space will be proposed. The presented results show that the proposed approaches are indeed helpful in resolving some of the main problems in content based image retrieval.

This thesis aims to explore and describe the content diversity of Siri’s search results in the polarized context of US politics. To do so, a crowdsourced audit was conducted. A diverse sample of 134 US-based Siri users between the ages of 18-64 performed five identical queries about the politically controversial issues of gun laws, immigration, the death penalty, taxes and abortion. The data were viewed through a theoretical framework using the concepts of algorithmic bias and media-centric fragmentation. The results suggest that Siri’s search algorithm produces a long tail distribution of search results: Forty-two percent of the participants received the six most frequent answers, while 22% of the users received unique answers. These statistics indicate that Siri’s search algorithm causes moderate concentration and low fragmentation. The age and, surprisingly, the political orientation of users, do not seem to be driving either concentration or fragmentation. However, the users' gender and location appears to cause low concentration. The finding that Siri’s search algorithm produces a long tail of replies challenges previous research on the content diversity of search results, which found no evidence of fragmentation. However, due to the limited scope of this study, these findings cannot be generalized to a larger population. Further research is needed to support or refute them.

Content Management Systems (CMS) generally store data in a way that the content is distributed among several relational database tables or stored in files as a whole without any distinctive characteristics. These storage mechanisms cannot provide the management of semantic information about the data. They lack semantic retrieval, search and browsing of the stored content. To enhance non-semantic CMSes with advanced semantic features, the semantics within the CMS itself and additional semantic information related with the actual managed content should also be taken into account. However, extracting implicit knowledge from the legacy CMSes, lifting to a semantic content management system environment and providing semantic operations on the content is a challenging task which includes adoption of several latest advancements in information extraction (IE), information retrieval (IR) and Semantic Web areas. In this study, we propose an integrative approach including automatic lifting of content from legacy systems, automatic annotation of data with the information retrieved from the Linked Open Data (LOD) cloud and several semantic operations on the content in terms of storage and search. We use a simple RDF path language to create custom, semantic indexes and filter annotations obtained from LOD cloud in a way that is eligible for specific use cases. Filtered annotations are materialized along with the actual content of document in dedicated indexes. This semantix indexing infrastructure allows semantically meaningful search facilities on top of it. We realize our approach in the scope of Apache Stanbol project, which is a subproject developed in the scope of IKS project, by focusing on document storage and retrival parts of it. We evaluate our approach in healthcare domain with different domain ontologies (SNOMED/CT, ART, RXNORM) in addition to DBpedia as parts of LOD cloud which are used annotate documents and content obtained from different health portals.

Background: Emergency contraception (EC) effectively reduces the risk of pregnancy following unprotected or under-protected sexual intercourse. EC has recently become available without a prescription in Canada and the USA in order to improve its availability and use, especially by young women. At the same time, the Internet is increasingly relied on as a source for health information, and Internet use is now nearly universal in younger age groups. Nevertheless, how people use the web to look for information about EC, and the qualities of information they find, are not well understood. The objectives of this thesis were to: (1) investigate the change in Internet search patterns for EC-related search terms in Canada and the USA over time and through shifts in EC-related policy and (2) assess the qualities (e.g., credibility, readability) and source (e.g., financial affiliations) of web-based information available about EC. Methods: The impact of policy changes in Canada and the USA on Internet search volume was estimated using interrupted time series analysis for two search terms:“morning after pill” and “Plan B”. Quality ratings and readability scores were generated for the ten most frequently found websites in Canada and the USA. Results: Policy changes making EC available without a prescription in Canada and the USA appear to have had an impact on information-seeking patterns online for EC-related search terms. The university-based website ec.princeton.edu and Wikipedia (en.wikipedia.org) were found frequently in searches for EC-related search terms in both countries. The website sponsored by the makers of Plan B®, www.planb.ca, was found most frequently in Canadian searches. All websites achieved fair to medium ratings in a systematic quality assessment, and 14 (83%) of websites had a reading grade level higher than the reading grade levels recommended for written health information. Discussion: Regulatory changes making EC available without a prescription appear to have affected the frequency with which people look online for information about EC. Public health agencies may want to improve the quality, readability and prominence of their web pages in online searches for EC-related search terms to ensure easy and convenient access to comprehensible, unbiased, and high quality web-based materials.

The aim of this thesis is to evaluate the use of Search-based Procedural Content generation (SBPCG) to help a designer create levels for different game styles. I show how SBPCG can be used for level generation in different game genres by surveying both paper and released commercial solutions. I then provide empirical data by using a Genetic Algorithm (GA) to evolve levels in two different game types, first one being a space puzzle game, and the second a platform game. Constraints from a level designer provide a base to create fitness functions for both games with success. Even though difficulties with level representation make it hard for a designer to work with this technique directly, the generated levels show that the technique has promising potential to aid level designers with their work.

The purpose of this research was to evaluate user interaction with learning object metadata surrogates both in terms of content and presentation. The main objectives of this study were: (1) to review the literature on learning object metadata and user-centred evaluation of metadata surrogates in the context of cognitive information retrieval (including user-centred relevance and usability research); (2) to develop a framework for the evaluation of user interaction with learning object metadata surrogates in search result interfaces; (3) to investigate the usability of metadata surrogates in search result interfaces of learning object repositories (LORs) in terms of various presentation aspects (such as amount of information, structure and highlighting of query terms) as a means for facilitating the user relevance judgment process; (4) to investigate in-depth the type of content that should be included in learning object metadata surrogates in order to facilitate the process of relevance judgment; (5) to provide a set of recommendations—guidelines for the design of learning object metadata surrogates in search result interfaces both in terms of content and presentation.

Since multimedia data, such as images and videos, are way more expressive and informative than ordinary text-based data, people find it more attractive to communicate and express with them. Additionally, with the rising popularity of social networking tools such as Facebook and Twitter, multimedia information retrieval can no longer be considered a solitary task. Rather, people constantly collaborate with one another while searching and retrieving information. But the very cause of the popularity of multimedia data, the huge and different types of information a single data object can carry, makes their management a challenging task. Multimedia data is commonly represented as multidimensional feature vectors and carry high-level semantic information. These two characteristics make them very different from traditional alpha-numeric data. Thus, to try to manage them with frameworks and rationales designed for primitive alpha-numeric data, will be inefficient. An index structure is the backbone of any database management system. It has been seen that index structures present in existing relational database management frameworks cannot handle multimedia data effectively. Thus, in this dissertation, a generalized multidimensional index structure is proposed which accommodates the atypical multidimensional representation and the semantic information carried by different multimedia data seamlessly from within one single framework. Additionally, the dissertation investigates the evolving relationships among multimedia data in a collaborative environment and how such information can help to customize the design of the proposed index structure, when it is used to manage multimedia data in a shared environment. Extensive experiments were conducted to present the usability and better performance of the proposed framework over current state-of-art approaches.

The vast amount of data available over the World Wide Web has created the necessity for new initiatives that translate this data into useful information for users. Due to human's acute visual perception, applications that utilize information visualization CIV) methodologies may ease user frustration when facing an abundance of search results from an Internet query. The recent introduction of ditto.com, an Internet search engine that provides users with a graphical depiction of search results documents, is a recent initiative that employs IV methodologies. This research is based upon the usability of traditional information retrieval systems and Internet search applications, and the impact IV methodologies have had on these systems. A usability evaluation was recently implemented to determine if IV methodologies can facilitate users' search needs when searching for information over the Internet. Fifteen randomly selected participants that match the diversity of Web users were asked to compare two Internet search results interfaces: Yahoo! a search engine that provides users with text-based search results and the graphical displays found within ditto.com. Descriptive data was collected through usability questionnaires and observing users search for information. Measurable data was collected by testing the performance of each search engine as the users search for ready-reference questions. Time to complete search tasks, the accuracy of the tasks, and number of error rates was collected from this session. Users were asked to provide their preference for one of the search engines. The data was analyzed for mean averages, occurrence of specific incidents that help or hindered users, and distribution of results with user experience. The results of this study are presented in a narrative report of users' preferences and concerns.

Automated content-based answer search is proposed to use at the implementation of the firstline technical support for the product. The main task is to inform users about product and to solve the most common problems in product work. Question-answering system and intelligent chat bot are proposed as technologies that can be used in implementing this task.

This report has conducted a research of comparing a few open source search engines. The research contains two evaluation processes, the first evaluation will evaluate each open source search engine found on today’s market. Points will be given between one to five points depending on how well the open source search engine meets the requirements. The open source search engine with the highest score will then be chosen for implementation. The first evaluation resulted in Elasticsearch being the selected open source search engine and will continue to the implementation phase. The second evaluation will be measuring the system performance and the relevance of the SERP (Search Engine Results Pages). This phase will evaluate the system performance by taking time measurements on how long it takes for the search engines to deliver the SERP. The relevance of the search results will be judge by a group of CSN employers. The group will be giving point be-tween one to five points depending on the relevance of the SERP. It will eval-uate Elasticsearch with the search engine CSN are using today on their web-site (www.csn.se). This phase resulted in Elasticsearch being the better in performance measurements but not in the relevance of the SERP. This was discussed and came to the conclusion that most points were lost because of the first search result Elasticsearch delivered. If this search result was re-moved Elasticsearch could deliver as good results as the old search engine. The survey came to the conclusion that Elasticsearch is recommended for CSN if certain problem areas could be corrected before implementation into their systems.

During the last decade, the Internet has reinvented the music industry. Physical media have evolved towards online products and services. As a consequence of this transition, online music corpora have reached a massive scale and are constantly being enriched with new documents. At the same time, a great quantity of cultural heritage content remains undisclosed because of the lack of metadata to describe and contextualize it. This has created a need for music retrieval and discovery technologies that allow users to interact with all these music repositories efficiently and effectively. Music Information Retrieval (MIR) is the research field that studies methods and tools for improving such interaction as well as access to music documents. Most of the research works in MIR focuses on content-based approaches, which exploit the analysis of the audio signal of a song to extract significant descriptors of the music content. These content descriptors may be processed and used in different application scenarios, such as retrieval, recommendation, dissemination, musicology analysis, and so on. The thesis explores novel automatic (content-based) methodologies for music retrieval which are based on semantic textual descriptors, acoustic similarity, and a combination of the two; we show empirically how the proposed approaches lead to efficient and competitive solutions with respect to other alternative state-of-the-art strategies. Part of the thesis focuses on music discovery systems, that is search engines where users do not look for a specific song or artist, but may have some general criteria they wish to satisfy. These criteria are commonly expressed in the form of tags, that is short phrases that capture relevant characteristics of the songs, such as genre, instrumentation, emotions, and so on. Because of the scale of current collections, manually assigning tags to songs is becoming an infeasible task; for this reason the automatic tagging of music content is now considered a core challenge in the design of fully functional music retrieval systems. State-of-the-art content-based systems for music annotation (which are usually called auto-taggers) model the acoustic patterns of the songs associated with each tag in a vocabulary through machine learning approaches. Based on these tag models, auto-taggers generate a vector of tag weights when annotating a new song. This vector may be interpreted as a semantic multinomial (SMN), that is a distribution characterizing the relevance of each tag to a song, which can be used for music annotation and retrieval. A first original contribution reported in the thesis aims at improving state-of-the-art auto-taggers by considering tag co-occurrences. While a listener may derive semantic associations for audio clips from direct auditory cues (e.g. hearing “bass guitar”) as well as from context (e.g. inferring “bass guitar” in the context of a “rock” song), auto-taggers ignore this context. Indeed, although contextual relationships correlate tags, many state-of-the-art auto-taggers model tags independently. We present a novel approach for improving automatic music annotation by modeling contextual relationships between tags. A Dirichlet mixture model (DMM) is proposed as a second, additional stage in the modeling process to supplement any auto-tagging system that generates a semantic multinomial over a vocabulary of tags. For each tag in the vocabulary, a DMM captures the broader context defined by the tag by modeling tag co-occurrence patterns in the SMNs of songs associated with the tag. When annotating songs, the DMMs refine SMN annotations by leveraging contextual evidence. Experimental results demonstrate the benefits of combining a variety of auto-taggers with this generative context model; it generally outperforms other approaches to context modeling as well. The use of tags alone allows for efficient and effective music retrieval mechanisms; however, automatic tagging strategies may lead to noisy representations that may negatively affect the effectiveness of retrieval algorithms. Yet, search and discovery operations across music collections can be also carried out matching users interests or exploiting acoustic similarity. One major issue in music information retrieval is how to combine such noisy and heterogeneous information sources in order to improve retrieval effectiveness. At this aim, the thesis explores a statistical retrieval framework based on combining tags and acoustic similarity through a hidden Markov model. The retrieval mechanism relies on an application of the Viterbi algorithm which highlights the sequence of songs that best represents a user query. The model is presented for improving state-of-the-art music search and discovery engines by delivering more relevant ranking lists. In fact, through an empirical evaluation we show how the proposed model leads to better performances than retrieval approaches which rank songs according to individual information sources alone or which use a combination of them. Additionally, the high generality of the framework makes it suitable for other media as well, such as images and videos. Besides music discovery, the thesis challenges also the problem of music identification, the goal which is to match different recordings of the same songs (i.e. finding covers of a given query). At this aim we present two novel music descriptors based on the harmonic content of the audio signals. Their main purpose is to provide a compact representation which is likely to be shared by different performances of the same music score. At the same time, they also aim at reducing the storage requirements of the music representation as well as enabling efficient retrieval over large music corpora. The effectiveness of these two descriptors, combined in a single scalable system, has been tested for classical music identification, which is probably the applicative scenario that mostly needs automatic strategies for labeling unknown recordings. Scalability is guaranteed by an index-based pre-retrieval step which handles music features as textual words; in addition, precision in the identification is brought by alignment carried out through an application of hidden Markov models. Results with a collection of more than ten thousand recordings have been satisfying in terms of efficiency and effectiveness.<br>Nell’ultimo decennio l’avvento di Internet ha reinventato l’industria musicale, in particolare i supporti fisici si sono evoluti verso prodotti e servizi reperibili online. Questa transizione ha portato le collezioni musicali disponibili su Internet ad avere dimensioni enormi e in continua crescita, a causa del quotidiano inserimento di nuovo contenuto musicale. Allo stesso tempo, una buona parte dei documenti musicali tipici del patrimonio culturale rimane inaccessibile, a causa della mancanza di dati che li descrivano e li contestualizzino. Tutto ciò evidenzia la necessità di nuove tecnologie che permettano agli utenti di interagire con tutte queste collezioni musicali in modo effettivo ed efficiente. Il reperimento d’informazioni musicali (i.e. MIR) è il settore di ricerca che studia le tecniche e gli strumenti per migliorare sia questa interazione, sia l’accesso ai documenti musicali. La maggior parte della ricerca effettuata nel MIR riguarda tecniche automatiche basate sul contenuto (i.e. content-based), le quali analizzano il segnale audio di una canzone ed estraggono dei descrittori, che ne caratterizzano, appunto, il contenuto. Questi descrittori possono essere elaborati ed utilizzati in varie applicazioni: motori di ricerca, divulgazione, analisi musicologa e così via. La tesi presenta dei modelli originali content-based per motori di ricerca musicali di vario genere, che si basano, sia su descrittori semantici testuali e su similarità acustica, sia su una loro combinazione. Attraverso esperimenti pratici, dimostreremo come i modelli proposti ottengano prestazioni efficienti e competitive se confrontate con alcuni dei sistemi alternativi presenti nello stato dell’arte. Una buona parte della tesi si concentra sui sistemi di music discovery, ovvero motori di ricerca nei quali gli utenti non cercano una canzone o un’artista specifico, ma hanno perlopiù un criterio generale che vogliono soddisfare. Questi criteri di ricerca sono in genere espressi sottoforma di tag, ovvero annotazioni che caratterizzano gli aspetti rilevanti delle canzoni (e.g. genere, strumenti, emozioni). A causa delle dimensioni raggiunte ormai dalle varie collezioni, l’assegnazione manuale dei tag alle canzoni è però diventata un’operazione impraticabile. Per questa ragione, i modelli che assegnano i tag in modo automatico sono diventati dei punti chiave nella progettazione dei motori di ricerca musicale. I sistemi content-based per l’assegnazione automatica di tag (i.e. auto-tagger) generalmente si basano su approcci di machine learning, che modellano le caratteristiche audio delle canzoni associate ad un certo tag. Questi modelli sono poi utilizzati per annotare le nuove canzoni generando un vettore di pesi, uno per ogni tag nel vocabolario, che misurano la rilevanza che ogni tag ha per quella canzone (i.e. SMN). Un primo contributo originale della tesi ha l’obiettivo di migliorare lo stato dell’arte degli auto-tagger, modellando le co-occorrenze tra i tag. Infatti mentre una persona può associare tag a una canzone sia direttamente (e.g. ascolta lo strumento“basso”), sia dal contesto (e.g. intuisce“basso” sapendo che la canzone `e di genere “rock”), gli auto-tagger diversamente ignorano questo contesto. Infatti, nonostante le relazioni contestuali correlino i tag, la maggior parte degli auto-tagger modella ogni tag in modo indipendente. Il nostro sistema pertanto cerca di migliorare l’assegnazione automatica di tag, modellando le relazioni contestuali che occorrono tra i vari tag di un vocabolario. Per far questo utilizziamo un modello di misture di Dirichlet (DMM) al fine di migliorare qualsiasi auto-tagger che genera delle SMN. Per ogni tag nel vocabolario, una DMM è usata per catturare le co-occorrenze con gli altri tag nelle SMN delle canzoni associate con quel tag. Quando una nuova canzone è annotata, il DMM rifinisce le SMN prodotte da un auto-tagger sfruttando le sue caratteristiche contestuali. I risultati sperimentali dimostrano i benefici di combinare vari auto-tagger con le DMM; in aggiunta, i risultati migliorano rispetto anche a quelli ottenuti con modelli contestuali alternativi dello stato dell’arte. L’uso dei tag permette di costruire efficienti ed effettivi motori di ricerca musicali; tuttavia le strategie automatiche per l’assegnazione di tag a volte ottengono rappresentazioni non precise che possono influenzare negativamente le funzioni di reperimento. Al tempo stesso, le ricerca di documenti musicali può essere anche fatta confrontando gli interessi degli utenti o sfruttando le similarit`a acustiche tra le canzoni. Uno dei principali problemi aperti nel MIR è come combinare tutte queste diverse informazioni per migliorare le funzioni di ricerca. Ponendosi questo obiettivo, la tesi propone un modello di reperimento statistico basato sulla combinazione tra i tag e la similarità acustica mediante un modello di Markov nascosto. Il meccanismo di ricerca si basa su un’applicazione dell’algoritmo di Viterbi, il quale estrae dal modello la sequenza di canzoni che meglio rappresenta la query. L’obiettivo è di migliorare lo stato dell’arte dei sistemi di ricerca musicale e, in particolare, di music discovery fornendo all’utente liste di canzoni maggiormente rilevanti. Gli esperimenti infatti mostrano come il modello proposto risulta migliore sia di algoritmi che ordinano le canzoni utilizzando un’informazione sola, sia di quelli che le combinano in modo diverso. In aggiunta, l’alta generalità a del modello lo rende adatto anche ad altri settori multimediali, come le immagini e i video. In parallelo con i sistemi di music discovery, la tesi affronta anche il problema di identificazione musicale (i.e. music identification), il cui obiettivo è quello di associare tra loro diverse registrazioni audio che condividono lo stesso spartito musicale (i.e. trovare le versioni cover di una certa query). In funzione di questo, la tesi presenta due descrittori che si basano sulla progressione armonica della musica. Il loro scopo principale è quello di fornire una rappresentazione compatta del segnale audio che possa essere condivisa dalle canzoni aventi lo stesso spartito musicale. Al tempo stesso, mirano anche a ridurre lo spazio di memoria occupato e a permettere operazioni di ricerca efficienti anche in presenza di grandi collezioni. La validità dei due descrittori è stata verificata per l’identificazione di musica classica, ovvero lo scenario che maggiormente necessita di strategie automatiche per la gestione di registrazioni audio non catalogate. La scalabilità del sistema è garantita da una pre-ricerca basata su un indice che gestisce i descrittori musicali come fossero parole di un testo; in aggiunta, la precisione dell’identificazione è aumentata mediante un’operazione di allineamento eseguita utilizzando i modelli di Markov nascosti. I risultati sperimentali ottenuti con una collezione di più di diecimila registrazioni audio sono stati soddisfacenti sia da un punto di vista di efficienza sia di efficacia.

The ever-increasing density of computer storage devices has allowed the average user to store enormous quantities of multimedia content, and a large amount of this content is usually music. Current search techniques for musical content rely on meta-data tags which describe artist, album, year, genre, etc. Query-by-content systems, however, allow users to search based upon the actual acoustical content of the songs. Recent systems have mainly depended upon textual representations of the queries and targets in order to apply common string-matching algorithms and are often confined to a single query style (e.g., humming). These methods also lose much of the information content of the song which limits the ways in which a user may search. We present a query-by-content system which supports querying in several styles using a Self-Organizing Map as its basis. The results from testing our system show that it performs better than random orderings and is, therefore, a viable option for musical query-by-content.

This thesis presents a series of experiments where endorsement rates, latencies and measures of cognitive ability were collected, to investigate the extent to which people search for counterexamples under necessity instructions, and alternative models under possibility instructions. The research was motivated by a syllogistic reasoning study carried out by Evans, Handley, Harper, and Johnson-Laird (1999), and predictions were derived from mental model theory (Johnson-Laird, 1983; Johnson-Laird &amp; Byrne, 1991). With regard to the endorsement rate data: Experiment 1 failed to find evidence that a search for counterexamples or alternative models took place. In contrast experiment 2 (transitive inference) found some evidence to support the search for alternative models under possibility instructions, and following an improved training session, experiment 3 produced strong evidence to suggest that people searched for other models; which was mediated by cognitive ability. There was also strong evidence from experiments 4, 5 and 6 (abstract and everyday conditionals) to support the search for counterexamples and alternative models. Furthermore it was also found that people were more likely to find alternative causes when there were many that could be retrieved from their everyday knowledge, and that people carried out a search for counterexamples with many alternative causes under necessity instructions, and across few and many causal groups under possibility instructions. .The evidence from the latency data was limited and inconsistent, although people with higher cognitive ability were generally quicker in completing the tasks.

Online shopping has become an important part of people’s daily life with the rapid development of e-commerce. In some domains such as books, electronics, and CD/DVDs, online shopping has surpassed or even replaced the traditional shopping method. Compared with traditional retailing, e-commerce is information intensive. One of the key factors to succeed in e-business is how to facilitate the consumers’ approaches to discover a product. Conventionally a product search engine based on a keyword search or category browser is provided to help users find the product information they need. The general goal of a product search system is to enable users to quickly locate information of interest and to minimize users’ efforts in search and navigation. In this process human factors play a significant role. Finding product information could be a tricky task and may require an intelligent use of search engines, and a non-trivial navigation of multilayer categories. Searching for useful product information can be frustrating for many users, especially those inexperienced users. This dissertation focuses on developing a new visual product search system that effectively extracts the properties of unstructured products, and presents the possible items of attraction to users so that the users can quickly locate the ones they would be most likely interested in. We designed and developed a feature extraction algorithm that retains product color and local pattern features, and the experimental evaluation on the benchmark dataset demonstrated that it is robust against common geometric and photometric visual distortions. Besides, instead of ignoring product text information, we investigated and developed a ranking model learned via a unified probabilistic hypergraph that is capable of capturing correlations among product visual content and textual content. Moreover, we proposed and designed a fuzzy hierarchical co-clustering algorithm for the collaborative filtering product recommendation. Via this method, users can be automatically grouped into different interest communities based on their behaviors. Then, a customized recommendation can be performed according to these implicitly detected relations. In summary, the developed search system performs much better in a visual unstructured product search when compared with state-of-art approaches. With the comprehensive ranking scheme and the collaborative filtering recommendation module, the user’s overhead in locating the information of value is reduced, and the user’s experience of seeking for useful product information is optimized.

A Content Management System (CMS) is a software application for creating, publishing, editing and managing content. The future step in content management system development is building intelligence over existing content resources that are heterogeneous in nature. Intelligence collected at the knowledge base can later on be used for executing semantic queries. Expressing the relations among content resources with ontological formalisms is therefore the key to implementing such semantic features. In this work, a methodology for the semantic lifting of JSR-170 compliant content repositories to ontologies is devised. The fact that in the worst case JSR-170 enforces no particular structural restrictions on the content model poses a technical challenge both for the initial build-up and further synchronization of the knowledge base. To address this problem, some recurring structural patterns in JSR-170 compliant content repositories are exploited. The value of the ontology extraction framework is assessed through a semantic search mechanism that is built on top of the extracted ontologies. The work in this thesis is complementary to the &ldquo<br>Interactive Knowledge Stack for small to medium CMS/KMS providers (IKS)&rdquo<br>project funded by the EC (FP7-ICT-2007-3).

This MSc Thesis was performed during a special practice at The Open University, Milton Keynes, UK. In recent years a number of new approaches for visualising and browsing document collections have been developed. These approaches try to address the problems associated with the growing amounts of content available and the changing patterns in the way people interact with information. Users now demand better support for exploring document collections to discover connections, compare and contrast information. Although visual search interfaces have the potential to improve the user experience in exploring document collections compared to textual search interfaces, they have not yet become as popular among users. The reasons for this range from the design of such visual interfaces to the way these interfaces are implemented and used. This work studies these reasons and determines the factors that contribute to an improved visual browsing experience. Consequently, by taking these factors into account, a novel visual search interface that improves exploratory search and the discovery of document relations is designed, implemented and evaluated.

The thesis presents a system for content based image retrieval and mining. The research presents a design of a scalable solution for efficient retrieval of images from large image databases using image features such as color, shape and texture. A framework for automatic labeling of images and clustering of meta data in database based on the dominant shapes, textures and colors in the image is proposed. The thesis also presents a new image tagging methodology to annotate the dominant image features to the image as meta data. The users of this system can input a query image and select similar image retrieval criteria by selecting a feature type from amongst color, texture or shape. The system retrieves images from the database that match the specified pattern and displays them by relevance. The user can enter a set of keywords or a combination of keywords that form the input text query. Images in the database that match the input text query are fetched and displayed. This ensures content based similar image search even for text based search. An efficient clustering algorithm is shown to improve the image retrieval by an order of magnitude.

This thesis defines a generic framework that allows building high level services, of both data search and content distribution, for structured peer-to-peer networks (SPN). We consider a twofold genericity: (i) Extensible framework for services and applications, with a dynamic deploy over other P2P systems; and (ii) generic and portable framework over most of the SPNs.<br>Esta tesis construye un marco de trabajo genérico que permite construir servicios de alto nivel, tanto de gestión de datos como de distribución de contenidos, para redes peer-to-peer estructradas (RPE). Consideramos que la genericidad proporcionada es doble: (i) Marco de trabajo extensible para servicios y aplicaciones, con un despliegue dinámico sobre diferentes sistemas peer-to-peer; (ii) Marco de trabajo genérico y portable de la mayoría de RPEs.

Both electronic health records (EHR) and clinical decision support (CDS)are each important attributions to clinicans and clinical workflow. Elec-tronic health records provide clinicans with crucial patient information atthe point of care, while clinical decision support gives well-founded and well-documented clinical recommendations at the point of decision making.This thesis explores how patient information from EHRs could be utilized as abasis for better and more effective decision support. Additionally, two meth-ods of accessing decision support recommendations are created and studied.One is based on search, with known elements from common general andmedical search interfaces. The other performs automatic ranking of relevantrecommendations based on patient information from a popular norwegianEHR system.To find out how these two prototypes should integrate and utilize informa-tion from an EHR system, an case-based experiment was conducted. Thisis a qualitative measure of user feedback, with elements from quantitativeresearch. User satisfaction were measured by using methods such as usertesting, interviews and surveys.User feedback suggest that clinical workflow have much to gain from integrat-ing EHR with CDS, as well as computerizing and making clinical nationalrecommendations available in an EHR context. It is also clear that elementslike search and automation are important features in an integrated system,and further research of decision support may include an integration of theseelements.

This thesis considers random access memory, cache, and out- and in-line linking as selected new forms of copyright infringement involving protected digital content. It examines the legality of transient and incidental copies in random access memory and cache client or local caching and proxy server or forward caching, and discusses the legality of different types of linking: surface and deep links and in-line linking, through critical analysis of the United Kingdom law, before and after the latest amendments in 2003 to the Copyright, Designs and Patents Act 1988, in addition to the Electronic Commerce (EC Directive) Regulations 2002. A brief discussion regarding the position of some European countries before and after the implementations of several European Union Directives related to copyright infringement concerning internet is also presented. The thesis considers several arguments related to the copyright holders' exclusive rights, such as issuing copies of the work to the public, performing, showing or playing the work in public or communicating the work to the public. However, it focuses mainly on the reproduction right. Several traditional defences to infringement of copyright are presented, especially fair dealing and implied licence. Also considered are different potential technological solutions, such as date bombs, metering technology, link removal request policy or password protection, and non-technological solutions, such as licensing or obtaining an agreement to link, as means of avoiding those new forms of copyright infringement. It is concluded that it is important to amend some pieces of the United Kingdom legislation, especially section 28A of the United Kingdom Copyright, Designs and Patents Act 1988 and the United Kingdom Electronic Commerce (EC Directive) Regulations 2002, as well as some pieces of the Directive 2001/29/EC of the European parliament and of the Council of 22 May 2001 on the harmonisation of Certain Aspects of Copyright and Related Rights in the Information Society, and the Directive 2000/31/EC of the European Parliament and of the Council of 8 June 2000 on certain legal aspects of information society services, in particular electronic commerce, in the Internal Market (Directive on electronic commerce).

In content-based image retrieval (CBIR) systems, there are two general types of search: target search and category search. Unlike queries in traditional database systems, users in most cases cannot specify an ideal query to retrieve the desired results for either target search or category search in multimedia database systems, and have to rely on iterative feedback to refine their query. Efficient evaluation of such iterative queries can be a challenge, especially when the multimedia database contains a large number of entries, and the search needs many iterations, and when the underlying distance measure is computationally expensive. The overall processing costs, including CPU and disk I/O, are further emphasized if there are numerous concurrent accesses. To address these limitations involved in relevance feedback processing, we propose a generic framework, including a query model, index structures, and query optimization techniques. Specifically, this thesis has five main contributions as follows. The first contribution is an efficient target search technique. We propose four target search methods: naive random scan (NRS), local neighboring movement (LNM), neighboring divide-and-conquer (NDC), and global divide-and-conquer (GDC) methods. All these methods are built around a common strategy: they do not retrieve checked images (i.e., shrink the search space). Furthermore, NDC and GDC exploit Voronoi diagrams to aggressively prune the search space and move towards target images. We theoretically and experimentally prove that the convergence speeds of GDC and NDC are much faster than those of NRS and recent methods. The second contribution is a method to reduce the number of expensive distance computation when answering k-NN queries with non-metric distance measures. We propose an efficient distance mapping function that transfers non-metric measures into metric, and still preserves the original distance orderings. Then existing metric index structures (e.g., M-tree) can be used to reduce the computational cost by exploiting the triangular inequality property. The third contribution is an incremental query processing technique for Support Vector Machines (SVMs). SVMs have been widely used in multimedia retrieval to learn a concept in order to find the best matches. SVMs, however, suffer from the scalability problem associated with larger database sizes. To address this limitation, we propose an efficient query evaluation technique by employing incremental update. The proposed technique also takes advantage of a tuned index structure to efficiently prune irrelevant data. As a result, only a small portion of the data set needs to be accessed for query processing. This index structure also provides an inexpensive means to process the set of candidates to evaluate the final query result. This technique can work with different kernel functions and kernel parameters. The fourth contribution is a method to avoid local optimum traps. Existing CBIR systems, designed around query refinement based on relevance feedback, suffer from local optimum traps that may severely impair the overall retrieval performance. We therefore propose a simulated annealing-based approach to address this important issue. When a stuck-at-a-local-optimum occurs, we employ a neighborhood search technique (i.e., simulated annealing) to continue the search for additional matching images, thus escaping from the local optimum. We also propose an index structure to speed up such neighborhood search. Finally, the fifth contribution is a generic framework to support concurrent accesses. We develop new storage and query processing techniques to exploit sequential access and leverage inter-query concurrency to share computation. Our experimental results, based on the Corel dataset, indicate that the proposed optimization can significantly reduce average response time while achieving better precision and recall, and is scalable to support a large user community. This latter performance characteristic is largely neglected in existing systems making them less suitable for large-scale deployment. With the growing interest in Internet-scale image search applications, our framework offers an effective solution to the scalability problem.<br>Ph.D.<br>School of Electrical Engineering and Computer Science<br>Engineering and Computer Science<br>Computer Science PhD

An Evolutionary Algorithm was run in real-time for the procedural generation ofenemies in a third-person, wave based hack and slash and shoot 'em up game. Thealgorithm evaluates enemies as individuals based on their effectiveness at battlingthe player character. Every generation is presented as a new wave of enemieswhose properties have been adjusted according to the fitness of the last wave. Byconstantly making new enemies more adept at the task of the defeating the currentplayer, I attempt to automatically and naturally raise the difficulty as the gameprogresses. The goal is also to improve player satisfaction as a result. By analyzingthe response from players and observing the changes of the generated enemies, Idetermine whether or not this is an appropriate implementation of EvolutionaryAlgorithms. Results showed that the success of the algorithm varied substantiallybetween tests, giving a number of both failed and successful tests. I go throughsome of the individual data and draw conclusions on what specific conditions makesthe algorithm perform desirably.

This thesis presents the data extraction of digitized old newspaper content and the implementation of a search function to simplify for the user. This is developed as a master’s degree project at Linköping University. The application allows the user to search for interesting content in a database of articles and can be used by both genealogists, local historians and novices. The database is filled with data from OCR scanned newspapers and the user can either search the database by their own or with the help of their family tree. The family tree is implemented by reading the users GEDcom file and extracting useful information that is then used to get better search results. The result is returned to the user in the form of digital articles. The work concludes that the information from GEDcom files can be used to find new interesting facts and that the user should be allowed to affect how the data is reduced, in the form of article categorization and filtering.

In this work, we focus on automatically classifying music by mood. For this purpose, we propose computational models using information extracted from the audio signal. The foundations of such algorithms are based on techniques from signal processing, machine learning and information retrieval. First, by studying the tagging behavior of a music social network, we find a model to represent mood. Then, we propose a method for automatic music mood classification. We analyze the contributions of audio descriptors and how their values are related to the observed mood. We also propose a multimodal version using lyrics, contributing to the field of text retrieval. Moreover, after showing the relation between mood and genre, we present a new approach using automatic music genre classification. We demonstrate that genre-based mood classifiers give higher accuracies than standard audio models. Finally, we propose a rule extraction technique to explicit our models.<br>En esta tesis, nos centramos en la clasificación automática de música a partir de la detección de la emoción que comunica. Primero, estudiamos cómo los miembros de una red social utilizan etiquetas y palabras clave para describir la música y las emociones que evoca, y encontramos un modelo para representar los estados de ánimo. Luego, proponemos un método de clasificación automática de emociones. Analizamos las contribuciones de descriptores de audio y cómo sus valores están relacionados con los estados de ánimo. Proponemos también una versión multimodal de nuestro algoritmo, usando las letras de canciones. Finalmente, después de estudiar la relación entre el estado de ánimo y el género musical, presentamos un método usando la clasificación automática por género. A modo de recapitulación conceptual y algorítmica, proponemos una técnica de extracción de reglas para entender como los algoritmos de aprendizaje automático predicen la emoción evocada por la música