Accepted papers

  • Context-Sensitive Spelling Correction with Trigrams: Permitting Multiple Corrections per Window
    Seyed Mohammad Sadegh Dashti,Shiraz University, Iran
    ABSTRACT
    Spelling correction is one of the basic tasks in the area of Natural Language Processing. Unlike common spelling errors, real-word errors may not be detected by conventional spelling correction techniques. The trigram model of real-word error correction that was proposed by Mays, Damerau, and Mercer has already demonstrated great performance in different evaluations. In this work we analyze several drawbacks of this model, and present a new variation, which focuses on correcting multiple real-word errors in a sentence and manipulates the Constrain Grammar to discriminate among the members of the search space. We tested the model on the Corpus of Wall Street Journal; in comparison to fixed windows size method of Wilcox-O’Hearn and Hirst and WordNet-based method of Hirst and Budanitsky the proposed method demonstrated significantly better performance.
  • The Appropriateness of the Factual Density as an Informativeness Measure for Online News
    Filipo Sharevski,Purdue University, USA
    ABSTRACT
    Automatically extracting useful semantic information and generating summary from video to facilitate user’s usage requirements is an important problem, especially in the context of mobile computing and the need for users to access their multimedia information of interest from anywhere at any time. This paper covers the study and comparison of various video summarization techniques for mobile devices available in the literature from 2001 until the first half of 2014. The comparison criteria used her have been extracted from analysis of literature and existing work in the domain of video summarization. The advantage of this research is making a state of the art of the related works which helps researchers in this context. Also, researcher can classify his works based on the proposed criteria.
  • Semantical Considerations on Non-monotonic Description Logics
    Anees Mehdi,Festo AG & Co. KG,Germany
    ABSTRACT
    Description logics of minimal knowledge and negation as failure (MKNF-DLs) are formalisms which augment description logics (DLs) with the modal operators K representing ‘knowledge’ and A representing ‘default assumption’. Such hybrid formalisms are useful in characterizing many nonmonotonic features which otherwise can not be captured in pure DLs. The traditional semantics employed for MKNF-DLs is based on the possible world approach where each world corresponds to a DL interpretation. Further, the semantics requires the interpretations (worlds) to share a common domain and to interpret constants rigidly across the worlds. In this paper we argue that these restrictions lead to unintended effects when an expressive MKNF-DL like SROIQKNF is considered. We thus propose employing the extended semantics (introduced recently) for SROIQKNF. We then provide a comparison between the traditional and the extended semantics including a comparison from a first-order modal logic perspective. Finally, we present a methodology for performing reasoning tasks in SROIQKNF along with the complexity results.