We advocate a relation based approach to Argumentation Mining. Our focus lies on the extraction of argumentative relations instead of the identification of arguments, themselves. By classifying pairs of sentences according to the relation that holds between them we are able to identify sentences that may be factual when considered in isolation, but carry argumentative meaning when read in context. We describe scenarios in which this is useful, as well as a corpus of annotated sentence pairs we are developing to provide a testbed for this approach.

Argumentation has proven successful in a number of domains, including multi-agent systems and decision support in medicine and engineering. We propose its application to a domain yet largely unexplored by argumentation research: computational linguistics. Over the past decade or so advances in this field have commonly relied on data-driven solutions, i.e. machine learning. Recently, however, there appears to be a growing consent that, in order to achieve significant advances in certain areas of artificial intelligence, in general, and computational linguistics, in particular, we may need to consider data-driven approaches in unison with reasoning-, logic-and rule-based solutions. To this end we have developed a novel classification methodology that incorporates reasoning through argumentation with supervised classifiers. We train classifiers and then argue about the validity of their output.To do so we identify arguments that formalise prototypical knowledge of a problem and use them to correct misclassifications. We thus have at our disposal multiple ways of incorporating knowledge in classification and are able to integrate knowledge as it becomes available, without the need to retrain classifiers.We illustrate our methodology on two tasks. On the one hand we address binary cross-domain sentiment polarity classification, where we train classifiers on one corpus, e.g. Tweets, to identify positive/negative polarity, and classify instances from another corpus, e.g. sentences from movie reviews. On the other hand we address a form of argumentation mining that we call Relation-based Argumentation Mining, where we classify pairs of sentences based on whether the first sentence attacks or supports the second, or whether it does neither. Whenever we find that one sentence attacks/supports the other we consider both to be argumentative, irrespective of their stand-alone argumentativeness. For both tasks we improve classification performance when using our methodology, compared to using standard classifiers, only.3

No abstract

No abstract

Abstract-Domain dependence is an issue that most researchers in corpus-based computational linguistics have faced at one time or another. With this paper we describe a method to perform sentiment polarity classification across domains that utilises Argumentation. We train standard supervised classifiers on a corpus and then attempt to classify instances from a separate corpus, whose contents are concerned with different domains (e.g. sentences from film reviews vs. Tweets). As expected the classifiers perform poorly and we improve upon the use of a simple classifier for out-of-domain classification by taking class labels suggested by classifiers and arguing about their validity. Whenever we can find enough arguments suggesting a mistake has been made by the classifier we change the class label according to what the arguments tell us. By arguing about class labels we are able to improve F 1 measures by as much as 14 points, with an average improvement of F 1 = 7.33 across all experiments.

No abstract

No abstract

Sentiment Analysis is concerned with (1) differentiating opinionated text from factual text and, in the case of opinionated text, (2) determine its polarity. With this paper, we address problem (1) and present A-SVM (Argument enhanced Support Vector Machines), a multimodal system that focuses on the discrimination of opinionated text from non-opinionated text with the help of (i) Support Vector Machines (SVM) and (ii) arguments, acquired by means of a user feedback mechanism, and used to improve the SVM classifications. We have used a prototype to investigate the validity of approaching Sentiment Analysis in this multi faceted manner by comparing straightforward Machine Learning techniques with our multimodal system architecture. All evaluations were executed using a purpose-built corpus of annotated text and A-SVM's classification performance was compared to that of SVM. The classification of a test set of approximately 4,500 n-grams yielded an increase in classification precision of 5.6%.