Browsing by Author "Zajic, David"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Citation Handling for Improved Summarization of Scientific Documents
    (2011-07-25) Whidby, Michael; Zajic, David; Dorr, Bonnie
    In this paper we present the first steps toward improving summarization of scientific documents through citation analysis and parsing. Prior work (Mohammad et al., 2009) argues that citation texts (sentences that cite other papers) play a crucial role in automatic summarization of a topical area, but did not take into account the noise introduced by the citations themselves. We demonstrate that it is possible to improve summarization output through careful handling of these citations. We base our experiments on the application of an improved trimming approach to summarization of citation texts extracted from Question-Answering and Dependency-Parsing documents. We demonstrate that confidence scores from the Stanford NLP Parser (Klein and Manning, 2003) are significantly improved, and that Trimmer (Zajic et al., 2007), a sentence-compression tool, is able to generate higher-quality candidates. Our summarization output is currently used as part of a larger system, Action Science Explorer (ASE) (Gove, 2011).
  • Thumbnail Image
    Item
    Correcting Errors in Digital Lexicographic Resources Using a Dictionary Manipulation Language
    (Trojina Institute for Applied Slovene Studies, 2011-11) Zajic, David; Maxwell, Michael; Doermann, David; Rodrigues, Paul; Bloodgood, Michael
    We describe a paradigm for combining manual and automatic error correction of noisy structured lexicographic data. Modifications to the structure and underlying text of the lexicographic data are expressed in a simple, interpreted programming language. Dictionary Manipulation Language (DML) commands identify nodes by unique identifiers, and manipulations are performed using simple commands such as create, move, set text, etc. Corrected lexicons are produced by applying sequences of DML commands to the source version of the lexicon. DML commands can be written manually to repair one-off errors or generated automatically to correct recurring problems. We discuss advantages of the paradigm for the task of editing digital bilingual dictionaries.
  • Thumbnail Image
    Item
    Detecting Structural Irregularity in Electronic Dictionaries Using Language Modeling
    (Trojina Institute for Applied Slovene Studies, 2011-11) Rodrigues, Paul; Zajic, David; Doermann, David; Bloodgood, Michael; Ye, Peng
    Dictionaries are often developed using tools that save to Extensible Markup Language (XML)-based standards. These standards often allow high-level repeating elements to represent lexical entries, and utilize descendants of these repeating elements to represent the structure within each lexical entry, in the form of an XML tree. In many cases, dictionaries are published that have errors and inconsistencies that are expensive to find manually. This paper discusses a method for dictionary writers to quickly audit structural regularity across entries in a dictionary by using statistical language modeling. The approach learns the patterns of XML nodes that could occur within an XML tree, and then calculates the probability of each XML tree in the dictionary against these patterns to look for entries that diverge from the norm.
  • Thumbnail Image
    Item
    A random forest system combination approach for error detection in digital dictionaries
    (Association for Computational Linguistics, 2012-04-23) Bloodgood, Michael; Ye, Peng; Rodrigues, Paul; Zajic, David; Doermann, David
    When digitizing a print bilingual dictionary, whether via optical character recognition or manual entry, it is inevitable that errors are introduced into the electronic version that is created. We investigate automating the process of detecting errors in an XML representation of a digitized print dictionary using a hybrid approach that combines rule-based, feature-based, and language model-based methods. We investigate combining methods and show that using random forests is a promising approach. We find that in isolation, unsupervised methods rival the performance of supervised methods. Random forests typically require training data so we investigate how we can apply random forests to combine individual base methods that are themselves unsupervised without requiring large amounts of training data. Experiments reveal empirically that a relatively small amount of data is sufficient and can potentially be further reduced through specific selection criteria.