Figure 3 - uploaded by Tobias Günther
Content may be subject to copyright.
Effect of random subspace learning on classification performance  

Effect of random subspace learning on classification performance  

Source publication
Thesis
Full-text available
In this work we examine the problem of sentiment analysis in microblogs, which has become a popular research topic in the last years. We provide a detailed review of previous work in the field and a survey summarizing common practices and available resources. Furthermore, we conduct a series of machine learning experiments using the largest manuall...

Context in source publication

Context 1
... in- tuition behind this method is to randomly disable features for each training sample, which leads to more redundancy in the models by forcing the classi- fier to not rely too much on very strong features. We try random subspace learning, by training classifiers on 10 concatenated corrupted copies of the ALL 3 dataset, where in each copy each feature of each sample is disabled with a probability of p. Figure 3 shows the result for the Naive Bayes classifier and support vector machine for different values of p. From the results we can see that both classifiers marginally benefit from applying random subspace learning. However, while the Naive Bayes classifier reaches its highest per- formance with a very corrupted dataset (when each feature is disable with a probability of 80%), the support vector machine reaches its best result when each feature is deactivated with a probability of only 20%. ...

Similar publications

Article
Full-text available
Opinion targets identification is an important task of the opinion mining problem. Several approaches have been employed for this task, which can be broadly divided into two major categories: supervised and unsupervised. The supervised approaches require training data, which need manual work and are mostly domain dependent. The unsupervised techniq...
Article
Full-text available
We consider the task of identifying attitudes towards a given set of entities from text. Conventionally, this task is decomposed into two separate subtasks: target detection that identifies whether each entity is mentioned in the text, either explicitly or implicitly, and polarity classification that classifies the exact sentiment towards an identi...
Article
Full-text available
With the rapid growth of user-generated content on the internet, automatic sentiment analysis of online customer reviews has become a hot research topic recently, but due to variety and wide range of products and services being reviewed on the internet, the supervised and domain-specific models are often not practical. As the number of reviews expa...

Citations

... They found out that best results are obtained from interpolating semantic concepts into unigram model of the Naive Bayesian classifier. Also, Tobias and Johansson [14] implemented Multinomial Naive Bayesian algorithm to classify sentiments on several datasets by experimenting and evaluating methods and possible reasons for conflicting results from the previous researches. They also proposed ideas to enhance sentiment classification using Multinomial Naive Bayesian (MNB). ...
Article
Full-text available
Twitter, an online micro-blogging and social networking service, provides registered users the ability to write in 140 characters anything they wish and hence providing them the opportunity to express their opinions and sentiments on events taking place. Politically sentimental tweets are top-trending tweets; whenever election is near, users tweet about their favorite candidates or political parties and at times give their reasons for that. In this study, we hybridize two n-gram [two n-gram models used in this study are unigram and n-gram. Therefore, in this study, where unigram is mentioned that refers to a least-order n-gram (unigram) and where n-gram is mentioned that refers to the highest-order (full sentence or tweet level) n-gram] models and applied Laplace smoothing to Naïve Bayesian classifier and Katz back-off on the model. This was done in order to smoothen and address the limitation of accuracy in terms of precision and recall of n-gram models caused by the ‘zero count problem.’ Result from our baseline model shows an increase of 6.05% in average F-Harmonic accuracy in comparison with the n-gram model and 1.75% increase in comparison with the semantic-topic model proposed from a previous study on the same dataset, i.e., Obama–McCain dataset.
... Finding opinion expressions in text is a hard problem. On the one hand, we have texts from social media where the messages are typically very short and often written in a very informal style, which makes NLP analysis harder than for traditional texts [28]. On the other hand, we have media such as newspapers that often contain opposing opinions from different parties in the same text, sometimes adding an author's viewpoint as well. ...
Article
Full-text available
The concept of culturomics was born out of the availability of massive amounts of textual data and the interest to make sense of cultural and language phenomena over time. Thus far however, culturomics has only made use of, and shown the great potential of, statistical methods. In this paper, we present a vision for a knowledge-based culturomics that complements traditional culturomics. We discuss the possibilities and challenges of combining knowledge-based methods with statistical methods and address major challenges that arise due to the nature of the data; diversity of sources, changes in language over time as well as temporal dynamics of information in general. We address all layers needed for knowledge-based culturomics, from natural language processing and relations to summaries and opinions.
... There has also been work on lexicon-based approaches to sentiment analysis of microblogs, such as O' Connor et al. (2010), Thelwall et al. (2010) and Zhang et al. (2011). For a detailed discussion see Günther (2013). In 2013, the International Workshop on Semantic Evaluation (SemEval) organized a shared task on sentiment analysis in Twitter (Nakov et al., 2013) to enable a better comparison of different approaches for sentiment analysis of microblogs. ...
Conference Paper
Full-text available
This paper describes the enhancements made to our GU-MLT-LT system (Günther and Furrer, 2013) for the SemEval-2014 re-run of the SemEval-2013 shared task on sentiment analysis in Twitter. The changes include the usage of a Twitter-specific to-kenizer, additional features and sentiment lexica, feature weighting and random sub-space learning. The improvements result in an increase of 4.18 F-measure points on this year's Twitter test set, ranking 3rd.
Article
The rapid evolution of microblogging and the emergence of sites such as Twitter have propelled online communities to flourish by enabling people to create, share and disseminate free-flowing messages and information globally. The exponential growth of product-based user reviews has become an ever-increasing resource playing a key role in emerging Twitter-based sentiment analysis (SA) techniques and applications to collect and analyse customer trends and reviews. Existing studies on supervised black-box sentiment analysis systems do not provide adequate information, regarding rules as to why a certain review was classified to a class or classification. The accuracy in some ways is less than our personal judgement. To address these shortcomings, alternative approaches, such as supervised white-box classification algorithms, need to be developed to improve the classification of Twitter-based microblogs. The purpose of this study was to develop a supervised white-box microblogging SA system to analyse user reviews on certain products using rough set theory (RST)-based rule induction algorithms. RST classifies microblogging reviews of products into positive, negative, or neutral class using different rules extracted from training decision tables using RST-centric rule induction algorithms. The primary focus of this study is also to perform sentiment classification of microblogs (i.e. also known as tweets) of product reviews using conventional, and RST-based rule induction algorithms. The proposed RST-centric rule induction algorithm, namely Learning from Examples Module version: 2, and LEM2 ++ Corpus-based rules (LEM2 ++ CBR),which is an extension of the traditional LEM2 algorithm, are used. Corpus-based rules are generated from tweets, which are unclassified using other conventional LEM2 algorithm rules. Experimental results show the proposed method, when compared with baseline methods, is excellent, with regard to accuracy, coverage and the number of rules employed. The approach using this method achieves an average accuracy of 92.57% and an average coverage of 100%, with an average number of rules of 19.14.