Molecular interactions enable us to understand the complexity of the human living system and how it can be exploited or malfunction to cause disease. The biomedical literature presents detailed knowledge of molecular functions and therefore represents a valuable reservoir of data for studying disease. However, extracting this data efficiently is difficult as it is spread over millions of publications in text that is not machine-readable. In this thesis we investigate how text mining can be used to automatically extract data for molecular interactions and their context relevant to disease. We focus on two globally relevant classes of diseases of which manifest from contrasting mechanisms: pain-related diseases and diseases caused by pathogenic organisms. Using HIV-1 as a case study, we first show that text mining can be used to partially recreate a large, manually curated database of HIV-1-human molecular interactions derived from the literature. We highlight both weaknesses in the quality of the data produced by the text-mining approach and strengths in it being able to extract this data rapidly, identifying instances missed in the manual curation and its potential as a support tool. We then expand on this approach by showing how an entirely new database of protein interactions relevant to pain can be created efficiently and accurately using text mining to generate the data and manual curation to validate the data quality. The following chapter then presents an analysis of 1,002 unique pain-related protein-protein interactions derived from this database, showing that it is of greater relevance to pain research than databases of pain interactions created from other common starting points. We highlight its value by, for example, identifying new drug repurposing opportunities and exploring differences in specific pain diseases using the contextual detail afforded by the text mining. Finally, we expand further on our approach to extracting molecular interactions from the literature, by showing how interactions between human proteins and pathogens can be curated across pathogenic organisms. We demonstrate how these techniques can be used to expand our knowledge of human pathogen related interaction data already stored in public databases, by identifying 42 new HIV-1-human molecular interactions, 108 new interactions between pathogen species and human proteins and 33 new human proteins that were found to interact with pathogens. Together, the results show that contexualised text mining, when supported by manual curation, can be used to extract molecular interactions for contrasting disease types in an efficient and accurate manner.

This dissertation addresses the problem of analysing digital data in digital forensics. It will be shown that text mining methods can be adapted and applied to digital forensics to aid analysts to more quickly, efficiently and accurately analyse data to reveal truly useful information. Investigators who wish to utilise digital evidence must examine and organise the data to piece together events and facts of a crime. The difficulty with finding relevant information quickly using the current tools and methods is that these tools rely very heavily on background knowledge for query terms and do not fully utilise the content of the data. A novel framework in which to perform evidence discovery is proposed in order to reduce the quantity of data to be analysed, aid the analysts' exploration of the data and enhance the intelligibility of the presentation of the data. The framework combines information extraction techniques with visual exploration techniques to provide a novel approach to performing evidence discovery, in the form of an evidence discovery system. By utilising unrestricted, unsupervised information extraction techniques, the investigator does not require input queries or keywords for searching, thus enabling the investigator to analyse portions of the data that may not have been identified by keyword searches. The evidence discovery system produces text graphs of the most important concepts and associations extracted from the full text to establish ties between the concepts and provide an overview and general representation of the text. Through an interactive visual interface the investigator can explore the data to identify suspects, events and the relations between suspects. Two models are proposed for performing the relation extraction process of the evidence discovery framework. The first model takes a statistical approach to discovering relations based on co-occurrences of complex concepts. The second model utilises a linguistic approach using named entity extraction and information extraction patterns. A preliminary study was performed to assess the usefulness of a text mining approach to digital forensics as against the traditional information retrieval approach. It was concluded that the novel approach to text analysis for evidence discovery presented in this dissertation is a viable and promising approach. The preliminary experiment showed that the results obtained from the evidence discovery system, using either of the relation extraction models, are sensible and useful. The approach advocated in this dissertation can therefore be successfully applied to the analysis of textual data for digital forensics Copyright / Dissertation (MSc)--University of Pretoria, 2010. / Computer Science / unrestricted

In the last decade, large numbers of social media services have emerged and been widely used in people's daily life as important information sharing and acquisition tools. With a substantial amount of user-contributed text data on social media, it becomes a necessity to develop methods and tools for text analysis for this emerging data, in order to better utilize it to deliver meaningful information to users. Previous work on text analytics in last several decades is mainly focused on traditional types of text like emails, news and academic literatures, and several critical issues to text data on social media have not been well explored: 1) how to detect sentiment from text on social media; 2) how to make use of social media's real-time nature; 3) how to address information overload for flexible information needs. In this dissertation, we focus on these three problems. First, to detect sentiment of text on social media, we propose a non-negative matrix tri-factorization (tri-NMF) based dual active supervision method to minimize human labeling efforts for the new type of data. Second, to make use of social media's real-time nature, we propose approaches to detect events from text streams on social media. Third, to address information overload for flexible information needs, we propose two summarization framework, dominating set based summarization framework and learning-to-rank based summarization framework. The dominating set based summarization framework can be applied for different types of summarization problems, while the learning-to-rank based summarization framework helps utilize the existing training data to guild the new summarization tasks. In addition, we integrate these techneques in an application study of event summarization for sports games as an example of how to better utilize social media data.

Cancer is a serious threat to human health and is now one of major causes of death worldwide. However, the complexity of the cancer makes the development of new and specific diagnostic tools particularly challenging. A number of different strategies have been developed for biomarker discovery in cancer using microarray data. The problem that typically needs to be addressed is the scale of the data sets; we simply do not have (or are likely to obtain) sufficient data for classical machine learning approaches for biomarker discovery to be properly validated. Obtaining a biomarker that is specific to a particular cancer is also very challenging. The initial promise that was held out for gene microarray work for the development of cancer biomarkers has not yet yielded the hoped for breakthroughs. This work discusses the construction of a strategy for a systematic approach to biomarker discovery validation using lung cancer gene expression microarray data based around non-small cell cancer and in patients which either stayed disease free after surgery (a five year window) or in which the disease progressed and re-occurred. As a means of assisting the validation purposes we have therefore looked at new methodologies for using existing biological knowledge to support machine learning biomarker discovery techniques. We employ text mining strategy using previously published literature for correlating biological concepts to a given phenotype. Pathway driven approaches through the use of Web Services and workflows, enabled the large-scale dataset to be analysed systematically. The results showed that it was possible, at least using this specific data set, to clearly differentiate between progressive disease and disease free patients using a set of biomarkers implicated in neuroendocrine signaling. A validation of the biomarkers identified was attempted in three separately published data sets. This analysis showed that although there was support for some of our findings in one of these data sets, this appeared to be a function of the close similarity in experimental design followed rather than through specific of the analysis method developed.

Cieľom tejto práce je preskúmať dostupné systémy pre extrakciu informácií a možnosti ich použitia v projekte MedIEQ. Teoretickú časť obsahuje úvod do oblasti extrakcie informácií. Popisujem účel, potreby a použitie a vzťah k iným úlohám spracovania prirodzeného jazyka. Prechádzam históriou, nedávnym vývojom, meraním výkonnosti a jeho kritikou. Taktiež popisujem všeobecnú architektúru IE systému a základné úlohy, ktoré má riešiť, s dôrazom na extrakciu entít. V praktickej časti sa nacházda prehľad algoritmov používaných v systémoch pre extrakciu informácií. Opisujem oba typy algoritmov ? pravidlové aj štatistické. V ďalšej kapitole je zoznam a krátky popis existujúcich voľných systémov. Nakoniec robím vlastný experiment s dvomi systémami ? LingPipe a GATE na vybraných korpusoch. Meriam rôzne výkonnostné štatistiky. Taktiež som vytvoril malý slovník a regulárny výraz pre email aby som demonštroval taktiež pravidlá pre extrahovanie určitých špecifických informácií.

With the advent of online publishing of scientific research came an avalanche of electronic resources and repositories containing knowledge encoded in some form or another. In the domain of biomedical sciences, research is now being published at a faster-than-ever pace, with several thousand articles per day. It is impossible for any human being to process that amount of information in due time, let alone apply it to their own needs. Thus appeared the necessity of being able to automatically retrieve relevant documents and extract useful information from text. Although it is now possible to distil essential factual knowledge from text, it is difficult to interpret the connections between the extracted facts. These connections, also known as discourse relations, make the text coherent and cohesive, and their automatic discovery can lead to a better understanding of the conveyed knowledge. One fundamental discourse relation is causality, as it is the one which explains reasons and allows for inferences to be made. This thesis is the first comprehensive study which focusses on recognising discourse causality in biomedical scientific literature. We first construct a manually annotated corpus of discourse causality and analyse its characteristics. Then, a methodology for automatically recognising causal relations using text mining and natural language processing techniques is presented. Furthermore, we investigate the automatic identification of additional information about the polarity, certainty, knowledge type and source of causal relations. The entire methodology is evaluated by empirical experiments, whose results show that it is possible to successfully extract causal relations from biomedical literature. Finally, we provide an example of a direct application of our research and offer ideas for further research directions and possible improvements to our methodology.

The beginning of this decade has seen a phenomenal raise in the amount of data generated in the world. While this increase provides us with opportunities to understand various aspects of human behavior and mechanisms behind new phenomena, the technologies, statistical techniques and theories required to gain an in depth and comprehensive understanding haven't progressed at an equal pace. As little as 5 years back, we used to deal with problems where there is insufficient prior social science or economic theory and the interest is only in prediction of the outcome or where there is an appropriate social science or economic theory and the interest is in explaining a given phenomenon. Today, we deal with problems where there is insufficient social science or economic theory but the interest is in explaining a given phenomenon. This creates a big challenge the solution to which is of equal interest to both academics and practitioners. In my research, I contribute towards addressing these challenges by building exploratory frameworks that leverage a variety of techniques including social network analysis, text and data mining, econometrics, statistical computing and visualization. My three essay dissertation focuses on understanding the antecedents to the quality of user generated content and on subscription and un-subscription behavior of users from lists on Social Media. Using a data science approach on population sized samples from Wikipedia and Twitter, I demonstrate the power of customized exploratory analyses in uncovering facts that social science or economic theory doesn't dictate and show how metrics from these analyses can be used to build prediction models with higher accuracy. I also demonstrate a method for combining exploration, prediction and explanatory modeling and propose to extend this methodology to provide causal inference. This dissertation has general implications for building better predictive and explanatory models and for mining text efficiently in Social Media.

Diversos métodos têm sido desenvolvidos para a organização da crescente quantidade de documentos textuais. Esses métodos frequentemente fazem uso de algoritmos de agrupamento para organizar documentos que referem-se a um mesmo assunto em um mesmo grupo, supondo que conteúdos de documentos de um mesmo grupo são similares. Porém, existe a possibilidade de que documentos pertencentes a grupos distintos também apresentem características semelhantes. Considerando esta situação, há a necessidade de desenvolver métodos que possibilitem a organização flexível de documentos, ou seja, métodos que possibilitem que documentos sejam organizados em diferentes grupos com diferentes graus de compatibilidade. O agrupamento fuzzy de documentos textuais apresenta-se como uma técnica adequada para este tipo de organização, uma vez que algoritmos de agrupamento fuzzy consideram que um mesmo documento pode ser compatível com mais de um grupo. Embora tem-se desenvolvido algoritmos de agrupamento fuzzy que possibilitam a organização flexível de documentos, tal organização é avaliada em termos do desempenho do agrupamento de documentos. No entanto, considerando que grupos de documentos devem possuir descritores que identifiquem adequadamente os tópicos representados pelos mesmos, de maneira geral os descritores de grupos tem sido extraídos utilizando alguma heurística sobre um conjunto pequeno de documentos, realizando assim, uma avaliação simples sobre o significado dos grupos extraídos. No entanto, uma apropriada extração e avaliação de descritores de grupos é importante porque os mesmos são termos representantes da coleção que identificam os tópicos abordados nos documentos. Portanto, em aplicações em que o agrupamento fuzzy é utilizado para a organização flexível de documentos, uma descrição apropriada dos grupos obtidos é tão importante quanto um bom agrupamento, uma vez que, neste tipo de agrupamento, um mesmo descritor pode indicar o conteúdo de mais de um grupo. Essa necessidade motivou esta tese, cujo objetivo foi investigar e desenvolver métodos para a extração de descritores de grupos fuzzy para a organização flexível de documentos. Para cumprir esse objetivo desenvolveu se: i) o método SoftO-FDCL (Soft Organization - Fuzzy Description Comes Last ), pelo qual descritores de grupos fuzzy at são extraídos após o processo de agrupamento fuzzy, visando identicar tópicos da organização flexível de documentos independentemente do algoritmo de agrupamento fuzzy utilizado; ii) o método SoftO-wFDCL ( Soft Organization - weighted Fuzzy Description Comes Last ), pelo qual descritores de grupos fuzzy at também são extraídos após o processo de agrupamento fuzzy utilizando o grau de pertinência dos documentos em cada grupo, obtidos do agrupamento fuzzy, como fator de ponderação dos termos candidatos a descritores; iii) o método HSoftO-FDCL (Hierarchical Soft Organization - Fuzzy Description Comes Last ), pelo qual descritores de grupos fuzzy hierárquicos são extraídos após o processo de agrupamento hierárquico fuzzy, identificando tópicos da organização hierárquica flexível de documentos. Adicionalmente, apresenta-se nesta tese uma aplicação do método SoftO-FDCL no contexto do programa de educação médica continuada canadense, reforçando a utilidade e aplicabilidade da organização flexível de documentos / Several methods have been developed to organize the growing number of textual documents. Such methods frequently use clustering algorithms to organize documents with similar topics into clusters. However, there are situations when documents of dffierent clusters can also have similar characteristics. In order to overcome this drawback, it is necessary to develop methods that permit a soft document organization, i.e., clustering documents into different clusters according to different compatibility degrees. Among the techniques that we can use to develop methods in this sense, we highlight fuzzy clustering algorithms (FCA). By using FCA, one of the most important steps is the evaluation of the yield organization, which is performed considering that all analyzed topics are adequately identified by cluster descriptors. In general, cluster descriptors are extracted using some heuristic over a small number of documents. The adequate extraction and evaluation of cluster descriptors is important because they are terms that represent the collection and identify the topics of the documents. Therefore, an adequate description of the obtained clusters is as important as a good clustering, since the same descriptor might identify one or more clusters. Hence, the development of methods to extract descriptors from fuzzy clusters obtained for soft organization of documents motivated this thesis. Aiming at investigating such methods, we developed: i) the SoftO-FDCL (Soft Organization - Fuzzy Description Comes Last) method, in which descriptors of fuzzy clusters are extracted after clustering documents, identifying topics regardless the adopted fuzzy clustering algorithm; ii) the SoftO-wFDCL (Soft Organization - weighted Fuzzy Description Comes Last) method, in which descriptors of fuzzy clusters are also extracted after the fuzzy clustering process using the membership degrees of the documents as a weighted factor for the candidate descriptors; iii) the HSoftO-FDCL (Hierarchical Soft Organization - Fuzzy Description Comes Last) method, in which descriptors of hierarchical fuzzy clusters are extracted after the hierarchical fuzzy clustering process, identifying topics by means of a soft hierarchical organization of documents. Besides presenting these new methods, this thesis also discusses the application of the SoftO-FDCL method on documents produced by the Canadian continuing medical education program, presenting the utility and applicability of the soft organization of documents in real-world scenario

Atualmente, o meio mais comum de busca de informações é a WEB. Assim, é importante procurar métodos eficientes para recuperar essa informação. As máquinas de busca na WEB usualmente utilizam palavras-chaves para expressar uma busca. Porém, não é trivial caracterizar a informação desejada. Usuários diferentes com necessidades diferentes podem estar interessados em informações relacionadas, mas distintas, ao realizar a mesma busca. O processo de realimentação de relevância torna possível a participação ativa do usuário no processo de busca. A idéia geral desse processo consiste em, após o usuário realizar uma busca na WEB permitir que indique, dentre os sites encontrados, quais deles considera relevantes e não relevantes. A opinião do usuário pode então ser considerada para reordenar os dados, de forma que os sites relevantes para o usuário sejam retornados mais facilmente. Nesse contexto, e considerando que, na grande maioria dos casos, uma consulta retorna um número muito grande de sites WEB que a satisfazem, das quais o usuário é responsável por indicar um pequeno número de sites relevantes e não relevantes, tem-se o cenário ideal para utilizar aprendizado parcialmente supervisionado, pois essa classe de algoritmos de aprendizado requer um número pequeno de exemplos rotulados e um grande número de exemplos não-rotulados. Assim, partindo da hipótese que a utilização de aprendizado parcialmente supervisionado é apropriada para induzir um classificador que pode ser utilizado como um filtro de realimentação de relevância para buscas na WEB, o objetivo deste trabalho consiste em explorar algoritmos de aprendizado parcialmente supervisionado, mais especificamente, aqueles que utilizam multidescrição de dados, para auxiliar na recuperação de sites na WEB. Para avaliar esta hipótese foi projetada e desenvolvida uma ferramenta denominada C-SEARCH que realiza esta reordenação dos sites a partir da indicação do usuário. Experimentos mostram que, em casos que buscas genéricas, que o resultado possui um bom diferencial entre sites relevantes e irrelevantes, o sistema consegue obter melhores resultados para o usuário / As nowadays the WEB is the most common source of information, it is very important to find reliable and efficient methods to retrieve this information. However, the WEB is a highly volatile and heterogeneous information source, thus keyword based querying may not be the best approach when few information is given. This is due to the fact that different users with different needs may want distinct information, although related to the same keyword query. The process of relevance feedback makes it possible for the user to interact actively with the search engine. The main idea is that after performing an initial search in the WEB, the process enables the user to indicate, among the retrieved sites, a small number of the ones considered relevant or irrelevant according with his/her required information. The users preferences can then be used to rearrange sites returned in the initial search, so that relevant sites are ranked first. As in most cases a search returns a large amount of WEB sites which fits the keyword query, this is an ideal situation to use partially supervised machine learning algorithms. This kind of learning algorithms require a small number of labeled examples, and a large number of unlabeled examples. Thus, based on the assumption that the use of partially supervised learning is appropriate to induce a classifier that can be used as a filter for relevance feedback in WEB information retrieval, the aim of this work is to explore the use of a partially supervised machine learning algorithm, more specifically, one that uses multi-description data, in order to assist the WEB search. To this end, a computational tool called C-SEARCH, which performs the reordering of the searched results using the users feedback, has been implemented. Experimental results show that in cases where the keyword query is generic and there is a clear distinction between relevant and irrelevant sites, which is recognized by the user, the system can achieve good results

Dada a grande quantidade e diversidade de dados textuais sendo criados diariamente, as aplicações do processo de Mineração de Textos são inúmeras e variadas. Nesse processo, a qualidade da solução final depende, em parte, do modelo de representação de textos adotado. Por se tratar de textos em língua natural, relações sintáticas e semânticas influenciam o seu significado. No entanto, modelos tradicionais de representação de textos se limitam às palavras, não sendo possível diferenciar documentos que possuem o mesmo vocabulário, mas que apresentam visões diferentes sobre um mesmo assunto. Nesse contexto, este trabalho foi motivado pela diversidade das aplicações da tarefa de classificação automática de textos, pelo potencial das representações no modelo espaço-vetorial e pela lacuna referente ao tratamento da semântica inerente aos dados em língua natural. O seu desenvolvimento teve o propósito geral de avançar as pesquisas da área de Mineração de Textos em relação à incorporação de aspectos semânticos na representação de coleções de documentos. Um mapeamento sistemático da literatura da área foi realizado e os problemas de classificação foram categorizados em relação à complexidade semântica envolvida. Aspectos semânticos foram abordados com a proposta, bem como o desenvolvimento e a avaliação de sete modelos de representação de textos: (i) gBoED, modelo que incorpora a semântica obtida por meio de conhecimento do domínio; (ii) Uni-based, modelo que incorpora a semântica por meio da desambiguação lexical de sentidos e hiperônimos de conceitos; (iii) SR-based Terms e SR-based Sentences, modelos que incorporam a semântica por meio de anotações de papéis semânticos; (iv) NASARIdocs, Babel2Vec e NASARI+Babel2Vec, modelos que incorporam a semântica por meio de desambiguação lexical de sentidos e embeddings de palavras e conceitos. Representações de coleções de documentos geradas com os modelos propostos e outros da literatura foram analisadas e avaliadas na classificação automática de textos, considerando datasets de diferentes níveis de complexidade semântica. As propostas gBoED, Uni-based, SR-based Terms e SR-based Sentences apresentam atributos mais expressivos e possibilitam uma melhor interpretação da representação dos documentos. Já as propostas NASARIdocs, Babel2Vec e NASARI+Babel2Vec incorporam, de maneira latente, a semântica obtida de embeddings geradas a partir de uma grande quantidade de documentos externos. Essa propriedade tem um impacto positivo na performance de classificação. / Text Mining applications are numerous and varied since a huge amount of textual data are created daily. The quality of the final solution of a Text Mining process depends, among other factors, on the adopted text representation model. Despite the fact that syntactic and semantic relations influence natural language meaning, traditional text representation models are limited to words. The use of such models does not allow the differentiation of documents that use the same vocabulary but present different ideas about the same subject. The motivation of this work relies on the diversity of text classification applications, the potential of vector space model representations and the challenge of dealing with text semantics. Having the general purpose of advance the field of semantic representation of documents, we first conducted a systematic mapping study of semantics-concerned Text Mining studies and we categorized classification problems according to their semantic complexity. Then, we approached semantic aspects of texts through the proposal, analysis, and evaluation of seven text representation models: (i) gBoED, which incorporates text semantics by the use of domain expressions; (ii) Uni-based, which takes advantage of word sense disambiguation and hypernym relations; (iii) SR-based Terms and SR-based Sentences, which make use of semantic role labels; (iv) NASARIdocs, Babel2Vec and NASARI+Babel2Vec, which take advantage of word sense disambiguation and embeddings of words and senses.We analyzed the expressiveness and interpretability of the proposed text representation models and evaluated their classification performance against different literature models. While the proposed models gBoED, Uni-based, SR-based Terms and SR-based Sentences have improved expressiveness, the proposals NASARIdocs, Babel2Vec and NASARI+Babel2Vec are latently enriched by the embeddings semantics, obtained from the large training corpus. This property has a positive impact on text classification performance.