Temporal Relation Classification in Persian and English contexts

Proceedings of Recent Advances in Natural Language Processing, pages 261–269, Hissar, Bulgaria, 7-13 September 2013. Temporal Relation Classification in Persian and English Contexts Mahbaneh Eshaghzadeh Torbati Department of Computer Engineering, Sharif University of Technology, Tehran, Iran Mahbaneh.eshaghzadeh@gmail.com Gholamreza Ghassem-sani Department of Computer Engineering, Sharif University of Technology, Tehran, Iran sani@sharif.edu Seyed Abolghasem Mirroshandel Faculty of Engineering, University of Guilan, Rasht, Iran mirroshandel@guilan.ac.ir Yadollah Yaghoobzadeh Department of Computer Engineering, Sharif University of Technology, Tehran, Iran yaghoobzadeh@ce.sharif.edu Negin Karimi Hosseini School of Computer Science, Faculty of Information Sci- ence & Technology, Univer- siti Kebangsaan Malaysia Negin62_k@yahoo.com Abstract This paper introduces the first pattern-based Persian Temporal Relation Classifier (PTRC) that finds the type of temporal relations between pairs of events in the Persian texts. The proposed system uses support vector machines (SVMs) equipped by combinations of simple, convolution tree, and string subsequence kernels (SSK). In order to evaluate the algorithm, we have developed a Persian TimeBank (PTB) corpus. PTRC not only increases the performance of the classification by applying new features and SSK, but also alleviates the probable adverse effects of the Free Word Orderness (FWO) of Persian on temporal relation classification. We have also applied our proposed algorithm to two standard corpora on English (i.e., TimeBank and TempEval-2) to measure the efficiency of the new features and SSK. The experiments show the accuracies of 65.6%, 59.53%, 50.2%, and 62.17% on an augmented version of PTB, TimeBank, tasks E and F of TempEval-2, respectively. Consequently, we have achieved the third best result on TimeBank, and the second best result on the task F of TempEval-2. 1 Introduction The goal in temporal relation classification is to find the temporal ordering between temporal entities of the input text. As a result, these relations can be used in applications such as question answering and summarization systems. In general, temporal relation classification is the task of determining when an event/time expression has taken place with respect to some other event/time expressions. In this study, we only try to find these relations between events, not between events and time expressions. In temporal corpora that have been created so far, different temporal relation classes have been considered. In TimeBank (Pustejovsky et al., 2003), the first corpus that has changed the research trend towards machine learning methods, there are six different temporal relations, namely SIMULTANEOUS, INCLUDES, BEFORE, IBEFORE, BEGINS, and ENDS. On the other hand, in TempEval-1 (Verhagen et al., 2007) and TempEval-2 (Verhagen et al., 2010), the temporal relations are BEFORE, OVERLAP, AFTER, BEFORE-OR-OVERLAP, OVERLAP-OR- AFTER, and VAGUE. Despite the multitude of speakers of Persian (Bateni, 1995), there has not existed any corpus tagged with temporal relations in Persian yet. Thus, as the first step, events and its attributes were tagged in the PTB corpus (Yaghoobzadeh et al., 2012). We have continued their work by annotating temporal relations between tagged events and Signals, manually and based on an adapted version of the ISO-TimeML guideline (Pustejovsky et al., 2010). In the second step, our goal has been designing a system …

Content truncated to 500 words