Melanie Neunerdt

Part-Of-Speech Tagging for Social Media Texts

Work on Part-of-Speech (POS) tagging has mainly concentrated on standardized texts for many years. However, the interest in automatic evaluation of social media texts is growing considerably. As the nature of social media texts is clearly different from standardized texts, Natural Language Processing methods need to be adapted for reliable processing. The basis for such an adaption is a reliably tagged social media text training corpus. In this paper, we introduce a new social media text corpus and evaluate different state-of-the-art POS taggers that are retrained on that corpus. In particular, the applicability of a tagger trained on a specific social media text type to other types, such as chat messages or blog comments, is studied. We show that retraining the taggers on in-domain training data increases the tagging accuracies by more than five percentage points.

Focused crawling for building Web comment corpora

Web 2.0 provides various types of social media applications, e.g., blogs, forums and news sites that allow users to post Web comments. This kind of communication plays an important role in acceptance research. To extract different opinions from such data, it is necessary to build Web comment corpora. Building such corpora requires focused crawling. Many focused Web crawling algorithms are known to build topic-specific Web collections. However, the type of Web pages is typically not considered. In this paper, we introduce a new type-specific focused crawler, which uses a classifier based on HTML meta information. Its application allows for collecting only Web pages that cover Web comments from various domains.

A POS Tagger for Social Media Texts Trained on Web Comments

Using social media tools such as blogs and forums have become more and more popular in recent years. Hence, a huge collection of social media texts from different communities is available for accessing user opinions, e.g., for marketing studies or acceptance research. Typically, methods from Natural Language Processing are applied to social media texts to automatically recognize user opinions. A fundamental component of the linguistic pipeline in Natural Language Processing is Part-of-Speech tagging. Most state-of-the-art Part-of-Speech taggers are trained on newspaper corpora, which differ in many ways from non-standardized social media text. Hence, applying common taggers to such texts results in performance degradation. In this paper, we present extensions to a basic Markov model tagger for the annotation of social media texts. Considering the German standard Stuttgart/T¨ ubinger TagSet (STTS), we distinguish 54 tag classes. Applying our approach improves the tagging accuracy for social media texts considerably, when we train our model on a combination of annotated texts from newspapers and Web comments. standardized text, since they are characterized by a spoken language, a dialogic and an informal writing style. This poses some special challenges to deal with in developing methods for automatic POS tagging of Web comments. These are particularly, the treatment of unknown (out-of-vocabulary) words and the different grammatical structure of social media texts in contrast to newspaper text. Furthermore, text genre specific manually annotated corpora, i.e., Web comments are required for training and testing. To the best of our knowledge all large manually annotated corpora are exclusively newspaper texts. In this work, we propose a Markov model tagger with parameter estimation enhancements for the POS annotation of social media texts. We apply and evaluate the tagger for German social media texts exemplarily. In order to make our method usable for NLP methods requiring POS information, e.g., syntactical parsing, we use the 54 Stuttgart/T¨ ubinger

Acceptance as a success factor for planning wireless network infrastructure

Availability of ubiquitous wireless services is taken for granted by most of the people. Therefore, network operators have to deploy and to operate large radio networks that, particularly, support broadband services and applications. One effect is a growing cell density, i.e., the number of required base stations increases. On the other hand, there are people that are afraid of getting harmed by electromagnetic radiation emitted by base stations or there are people that just dislike the prevalence of base station antennas. Both types of peoples attitude refer to the field of user acceptance. A high deficiency in acceptance might lead to public disputes, political disputes, and negative economical consequences. To counteract such trends preventively, we propose a mathematical optimization model that allows for planning wireless network infrastructure with respect to user acceptance. While keeping technical and economical aspects as primarily considered planning criteria, we propose a multi-objective optimization model that takes into account user acceptance as an objective. Numerical results demonstrate effects of this approach compared to application of a conventional planning model.

Land use classification as a key component for path loss prediction in rural areas

Path loss prediction is an essential building block for planning and optimization of cellular radio networks. Semi-empirical land use based models yield accurate and efficient path loss prediction results in rural areas. Consequently, land use information serves as a key input for those models. In this paper, we present a new C × K-Nearest-Mean classification method operating on landscape images to provide the required input data for land use based path loss prediction models. With respect to this purpose our approach exceeds conventional land use classification using a neural network particularly in terms of total error rate. Utilization of our classification method by a specific path loss prediction model leads to prediction results with a mean square error of less than 6dB.

Enhanced Web Page Cleaning for Constructing Social Media Text Corpora

Web page cleaning is one of the most essential tasks in Web corpus construction. The intention is to separate the main content from navigational elements, templates, and advertisements, often referred to as boilerplate. In this paper, we particularly enhance Web page cleaning applied to pages containing comments and introduce a new training corpus for that purpose. Beside extending an existing boilerplate detection algorithm by means of a comment classifier, we train and test different classifiers on extended feature sets solving a two-class problem (content vs. boilerplate) on our and an existing benchmark corpus. Results show that the proposed approach outperforms existing methods, particularly on comment pages from different domains. Finally, we point out that our trained classifiers are domain independent and with small adjustments only transferable to other languages.