A large proportion of World Wide Web (WWW) users treat it as a social medium, i.e. many of them use the WWW to express and communicate their opinions. Economic value or utility can be created if these utterances, reactions, or feedback are extracted from various social media platforms and their content analysed. Some of these benefits are related to e-commerce, marketing, product improvements, improving machine learning algorithms etc. Moreover, establishing links between different social media platforms, based on shared topics and content, could provide access to the comments of users of different platforms. However, studies to date have generally tackled the area of content extraction from each type of social media in isolation. There is a lack of research of some aspects of social media, namely, linking the references from a blog post, for example, to information related to the same issue on Twitter. In addition, while studies have been carried out on various languages, there has been little investigation into social media in the Arabic language. This thesis tackles opinion mining and sentiment analysis of Arabic language social media, particularly in blogs and Twitter. The thesis focuses on Arabic language technology blogs in order to identify the expressed sentiments and then to link an issue within a blog post to relevant tweets in Twitter. This was done by assessing the similarity of content and measuring the sentiments scores. In order to extract the required data, text-mining techniques were used to build up corpora of the raw blog data in Modern Standard Arabic (MSA) and to build tools and lexicons required for this research. The results obtained through this research contribute to the field of computer science by furthering the employment of text-mining techniques, thus improving the process of information retrieval and knowledge accumulation. Moreover, the study developed new approaches to working with Arabic opinion mining and the domain of sentiment analysis.

The goal of my master thesis is to describe the Automated Sentiment Analysis, its methods and Cross-domain problems and to test the already existing model. I have applied this model on the data from the Czech-Slovak film database website CSFD.cz, Czech e-shop MALL.cz and one of the biggest Czech websites about books Databazeknih.cz to contribute to the solution of the Cross-Domain issue by using n-grams and the analytic software RapidMiner.

The Vector Space Model (VSM) of text representation suffers a number of limitations for text classification. Firstly, the VSM is based on the Bag-Of-Words (BOW) assumption where terms from the indexing vocabulary are treated independently of one another. However, the expressiveness of natural language means that lexically different terms often have related or even identical meanings. Thus, failure to take into account the semantic relatedness between terms means that document similarity is not properly captured in the VSM. To address this problem, semantic indexing approaches have been proposed for modelling the semantic relatedness between terms in document representations. Accordingly, in this thesis, we empirically review the impact of semantic indexing on text classification. This empirical review allows us to answer one important question: how beneficial is semantic indexing to text classification performance. We also carry out a detailed analysis of the semantic indexing process which allows us to identify reasons why semantic indexing may lead to poor text classification performance. Based on our findings, we propose a semantic indexing framework called Relevance Weighted Semantic Indexing (RWSI) that addresses the limitations identified in our analysis. RWSI uses relevance weights of terms to improve the semantic indexing of documents. A second problem with the VSM is the lack of supervision in the process of creating document representations. This arises from the fact that the VSM was originally designed for unsupervised document retrieval. An important feature of effective document representations is the ability to discriminate between relevant and non-relevant documents. For text classification, relevance information is explicitly available in the form of document class labels. Thus, more effective document vectors can be derived in a supervised manner by taking advantage of available class knowledge. Accordingly, we investigate approaches for utilising class knowledge for supervised indexing of documents. Firstly, we demonstrate how the RWSI framework can be utilised for assigning supervised weights to terms for supervised document indexing. Secondly, we present an approach called Supervised Sub-Spacing (S3) for supervised semantic indexing of documents. A further limitation of the standard VSM is that an indexing vocabulary that consists only of terms from the document collection is used for document representation. This is based on the assumption that terms alone are sufficient to model the meaning of text documents. However for certain classification tasks, terms are insufficient to adequately model the semantics needed for accurate document classification. A solution is to index documents using semantically rich concepts. Accordingly, we present an event extraction framework called Rule-Based Event Extractor (RUBEE) for identifying and utilising event information for concept-based indexing of incident reports. We also demonstrate how certain attributes of these events e.g. negation, can be taken into consideration to distinguish between documents that describe the occurrence of an event, and those that mention the non-occurrence of that event.

O sucesso da mineração de opiniões para processar automaticamente grandes quantidades de conteúdo opinativo disponíveis na Internet tem sido demonstrado como uma solução de baixa latência e mais barata para a análise de opinião pública. No presente trabalho foi investigado se é possível prever variações de intenção de voto com base em séries temporais de sentimento extraídas de comentários de notícias, utilizando três eleições brasileiras como estudo de caso. As contribuições deste estudo de caso são: a) a comparação de duas abordagens para a mineração de opiniões em conteúdo gerado por usuários em português do Brasil; b) a proposta de dois tipos de atributos discriminantes para representar o sentimento em relação a candidatos políticos a serem usados para a previsão, c) uma abordagem para prever variações de intenção de voto que é adequada para cenários de dados esparsos. Foram desenvolvidos experimentos para avaliar a influência dos atributos discriminantes propostos em relação a acurácia da previsão, e suas respectivas preparações. Os resultados mostraram uma acurácia de 70% na previsão de variações de intenção de voto positivas e negativas. Estas contribuições são importantes passos em direção a um framework que é capaz de combinar opiniões de diversas fontes para encontrar a representatividade de uma população alvo, de modo que se possa obter previsões mais confiáveis. / The success of opinion mining for automatically processing vast amounts of opinionated content available on the Internet has been demonstrated as a less expensive and lower latency solution for gathering public opinion. In this work, we investigate whether it is possible to predict variations in vote intention based on sentiment time series extracted from news comments, using three Brazilian elections as case study. The contributions of this case study are: a) the comparison of two approaches for opinion mining in user-generated content in Brazilian Portuguese; b) the proposition of two types of features to represent sentiment behavior towards political candidates that can be used for prediction, c) an approach to predict polls vote intention variations that is adequate for scenarios of sparse data. We developed experiments to assess the influence on the forecasting accuracy of the proposed features, and their respective preparation. Our results display an accuracy of 70% in predicting positive and negative variations. These are important contributions towards a more general framework that is able to blend opinions from several different sources to find representativeness of the target population, and make more reliable predictions.

O sucesso da mineração de opiniões para processar automaticamente grandes quantidades de conteúdo opinativo disponíveis na Internet tem sido demonstrado como uma solução de baixa latência e mais barata para a análise de opinião pública. No presente trabalho foi investigado se é possível prever variações de intenção de voto com base em séries temporais de sentimento extraídas de comentários de notícias, utilizando três eleições brasileiras como estudo de caso. As contribuições deste estudo de caso são: a) a comparação de duas abordagens para a mineração de opiniões em conteúdo gerado por usuários em português do Brasil; b) a proposta de dois tipos de atributos discriminantes para representar o sentimento em relação a candidatos políticos a serem usados para a previsão, c) uma abordagem para prever variações de intenção de voto que é adequada para cenários de dados esparsos. Foram desenvolvidos experimentos para avaliar a influência dos atributos discriminantes propostos em relação a acurácia da previsão, e suas respectivas preparações. Os resultados mostraram uma acurácia de 70% na previsão de variações de intenção de voto positivas e negativas. Estas contribuições são importantes passos em direção a um framework que é capaz de combinar opiniões de diversas fontes para encontrar a representatividade de uma população alvo, de modo que se possa obter previsões mais confiáveis. / The success of opinion mining for automatically processing vast amounts of opinionated content available on the Internet has been demonstrated as a less expensive and lower latency solution for gathering public opinion. In this work, we investigate whether it is possible to predict variations in vote intention based on sentiment time series extracted from news comments, using three Brazilian elections as case study. The contributions of this case study are: a) the comparison of two approaches for opinion mining in user-generated content in Brazilian Portuguese; b) the proposition of two types of features to represent sentiment behavior towards political candidates that can be used for prediction, c) an approach to predict polls vote intention variations that is adequate for scenarios of sparse data. We developed experiments to assess the influence on the forecasting accuracy of the proposed features, and their respective preparation. Our results display an accuracy of 70% in predicting positive and negative variations. These are important contributions towards a more general framework that is able to blend opinions from several different sources to find representativeness of the target population, and make more reliable predictions.

O sucesso da mineração de opiniões para processar automaticamente grandes quantidades de conteúdo opinativo disponíveis na Internet tem sido demonstrado como uma solução de baixa latência e mais barata para a análise de opinião pública. No presente trabalho foi investigado se é possível prever variações de intenção de voto com base em séries temporais de sentimento extraídas de comentários de notícias, utilizando três eleições brasileiras como estudo de caso. As contribuições deste estudo de caso são: a) a comparação de duas abordagens para a mineração de opiniões em conteúdo gerado por usuários em português do Brasil; b) a proposta de dois tipos de atributos discriminantes para representar o sentimento em relação a candidatos políticos a serem usados para a previsão, c) uma abordagem para prever variações de intenção de voto que é adequada para cenários de dados esparsos. Foram desenvolvidos experimentos para avaliar a influência dos atributos discriminantes propostos em relação a acurácia da previsão, e suas respectivas preparações. Os resultados mostraram uma acurácia de 70% na previsão de variações de intenção de voto positivas e negativas. Estas contribuições são importantes passos em direção a um framework que é capaz de combinar opiniões de diversas fontes para encontrar a representatividade de uma população alvo, de modo que se possa obter previsões mais confiáveis. / The success of opinion mining for automatically processing vast amounts of opinionated content available on the Internet has been demonstrated as a less expensive and lower latency solution for gathering public opinion. In this work, we investigate whether it is possible to predict variations in vote intention based on sentiment time series extracted from news comments, using three Brazilian elections as case study. The contributions of this case study are: a) the comparison of two approaches for opinion mining in user-generated content in Brazilian Portuguese; b) the proposition of two types of features to represent sentiment behavior towards political candidates that can be used for prediction, c) an approach to predict polls vote intention variations that is adequate for scenarios of sparse data. We developed experiments to assess the influence on the forecasting accuracy of the proposed features, and their respective preparation. Our results display an accuracy of 70% in predicting positive and negative variations. These are important contributions towards a more general framework that is able to blend opinions from several different sources to find representativeness of the target population, and make more reliable predictions.

With the rise of social media and publicly available data, opinion mining is more accessible than ever. It is valuable for content creators, companies and advertisers to gain insights into what users think and feel. This work examines comments on YouTube videos, and builds a deep learning classifier to automatically determine their sentiment. Four Long Short-Term Memory-based models are trained and evaluated. Experiments are performed to determine which deep learning model performs with the best accuracy, recall, precision, F1 score and ROC curve on a labelled YouTube Comment dataset. The results indicate that a BiLSTM-based model has the overall best performance, with the accuracy of 89%. Furthermore, the four LSTM-based models are evaluated on an IMDB movie review dataset, achieving an average accuracy of 87%, showing that the models can predict the sentiment of different textual data. Finally, a statistical analysis is performed on the YouTube videos, revealing that videos with positive sentiment have a statistically higher number of upvotes and views. However, the number of downvotes is not significantly higher in videos with negative sentiment.

No description available.

One critical aspect that remains in the application of state-of-the-art neural networks to text analysis in applied research is the continued requirement for manual data annotation. In computer science research, there is a strong focus on maximizing the data efficiency of fine-tuning language models. This has led to the development of zero-shot text classification methods, which promise to work effectively without requiring fine-tuning for the specific task at hand. In this paper, we conduct an in-depth analysis of aspect-based sentiment analysis in historic German stock market reports to evaluate the reliability of this promise. We present a comparison of a zero-shot approach with a meticulously fine-tuned three-step process of training and applying text classification models. This study aims to empirically assess the reliability of zero-shot text classification and provide justification for the potential benefits it offers in terms of reducing the burden of data labeling and training for analysis purposes. The findings of our study demonstrate a strong correlation between the sentiment time series generated through aspect-based sentiment analysis using the zero-shot approach and those derived from the fine-tuned supervised pipeline, validating the viability of the zero-shot approach. While the zero-shot pipeline exhibits a tendency to underestimate negative examples, the overall trend remains discernible. Additionally, a qualitative analysis of the linguistic patterns reveals no explicit error patterns. Nevertheless, we acknowledge and discuss the practical and epistemological obstacles associated with employing zero-shot algorithms in untested domains.

Identifying differing opinions on a given topic as expressed by multiple people (as in a set of written reviews for a given product, for example) presents challenges. Opinions about a particular subject are often nuanced: a person may have both negative and positive opinions about different aspects of the subject of interest, and these aspect-specific opinions can be independent of the overall opinion on the subject. Being able to identify, collect, and count these nuanced opinions in a large set of data offers more insight into the strengths and weaknesses of competing products and services than does aggregating the overall ratings of such products and services. I make two useful and useable contributions in working with opinionated text. First, I present my implementation of a semi-supervised co-training machine classification method for identifying both product aspects (features of products) and sentiments expressed about such aspects. It offers better precision than fully-supervised methods while requiring much less text to be manually tagged (a time-consuming process). This algorithm can also be run in a fully supervised manner when more data is available. Second, I apply this co-training approach to reviews of restaurants and various electronic devices; such text contains both factual statements and opinions about features/aspects of products. The algorithm automatically identifies the product aspects and the words that indicate aspect-specific opinion polarity, while largely avoiding the problem of misclassifying the products themselves as inherently positive or negative. This method performs well compared to other approaches. When run on a set of reviews of five technology products collected from Amazon, the system performed with some demonstrated competence (with an average precision of 0.83) at the difficult task of simultaneously identifying aspects and sentiments, though comparison to contemporaries' simpler rules-based approaches was difficult. When run on a set of opinionated sentences about laptops and restaurants that formed the basis of a shared challenge in the SemEval-2014 Task 4 competition, it was able to classify the sentiments expressed about aspects of laptops better than any team that competed in the task (achieving 0.72 accuracy). It was above the mean in its ability to identify the aspects of restaurants about which people expressed opinions, even when co-training using only half of the labelled training data at the outset. While the SemEval-2014 aspect-based sentiment extraction task considered only separately the tasks of identifying product aspects and determining their polarities, I take an extra step and evaluate sentences as a whole, inferring aspects and the aspect-specific sentiments expressed simultaneously, a more difficult task that seems more applicable to real-world tasks. I present first results of this sentence-level task. The algorithm uses both lexical and syntactic information in a manner that is shown to be able to handle new words that it has never before seen. It offers some demonstrated ability to adapt to new subject domains for which it has no training data. The system is characterizable by very high precision and weak-to-average recall and it estimates its own confidence in its predictions; this characteristic should make the algorithm suitable for use on its own or for combination in a confidence-based voting ensemble. The software created for and described in the course of this dissertation is made available online.