MIME-Version: 1.0 Server: CERN/3.0 Date: Monday, 06-Jan-97 21:58:42 GMT Content-Type: text/html Content-Length: 8275 Last-Modified: Wednesday, 07-Feb-96 21:14:04 GMT Papers and Abstracts

Papers and Abstracts

To view a paper, click on the open book image.

  1. Lexical Acquisition: A Novel Machine Learning Problem
    Cynthia A. Thompson and Raymond J. Mooney
    Submitted to the Thirteenth International Conference on Machine Learning (ICML-96).

    This paper defines a new machine learning problem to which standard machine learning algorithms cannot easily be applied. The problem occurs in the domain of lexical acquisition. The ambiguous and synonymous nature of words causes the difficulty of using standard induction techniques to learn a lexicon. Additionally, negative examples are typically unavailable or difficult to construct in this domain. One approach to solve the lexical acquisition problem is presented, along with preliminary experimental results on an artificial corpus. Future work includes extending the algorithm and performing tests on a more realistic corpus.

  2. Corpus-Based Lexical Acquisition For Semantic Parsing
    Cynthia A. Thompson
    Ph.D. proposal

    Building accurate and efficient natural language processing (NLP) systems is an important and difficult problem. There has been increasing interest in automating this process. The lexicon, or the mapping from words to meanings, is one component that is typically difficult to update and that changes from one domain to the next. Therefore, automating the acquisition of the lexicon is an important task in automating the acquisition of NLP systems. This proposal describes a system, WOLFIE (WOrd Learning From Interpreted Examples), that learns a lexicon from input consisting of sentences paired with representations of their meanings. Preliminary experimental results show that this system can learn correct and useful mappings. The correctness is evaluated by comparing a known lexicon to one learned from the training input. The usefulness is evaluated by examining the effect of using the lexicon learned by WOLFIE to assist a parser acquisition system, where previously this lexicon had to be hand-built. Future work in the form of extensions to the algorithm, further evaluation, and possible applications is discussed.

  3. Inducing Logic Programs without Explicit Negative Examples
    John M. Zelle, Cynthia A. Thompson, Mary Elaine Califf, and Raymond J. Mooney
    To appear in the Proceedings of the Fifth International Workshop on Inductive Logic Programming.

    This paper presents a method for learning logic programs without explicit negative examples by exploiting an assumption of output completeness. A mode declaration is supplied for the target predicate and each training input is assumed to be accompanied by all of its legal outputs. Any other outputs generated by an incomplete program implicitly represent negative examples; however, large numbers of ground negative examples never need to be generated. This method has been incorporated into two ILP systems, CHILLIN and IFOIL, both of which use intensional background knowledge. Tests on two natural language acquisition tasks, case-role mapping and past-tense learning, illustrate the advantages of the approach.

  4. Acquisition of a Lexicon from Semantic Representations of Sentences
    Cynthia A. Thompson 33rd Annual Meeting of the Association of Computational Linguistics, pp. 335-337, Boston, MA July 1995 (ACL-95).

    A system, WOLFIE, that acquires a mapping of words to their semantic representation is presented and a preliminary evaluation is performed. Tree least general generalizations (TLGGs) of the representations of input sentences are performed to assist in determining the representations of individual words in the sentences. The best guess for a meaning of a word is the TLGG which overlaps with the highest percentage of sentence representations in which that word appears. Some promising experimental results on a non-artificial data set are presented.

  5. Inductive Learning For Abductive Diagnosis
    Cynthia A. Thompson and Raymond J. Mooney
    Proceedings of the Twelfth National Conference on AI, Seattle, WA, July 1994. (AAAI-94)

    A new inductive learning system, LAB (Learning for ABduction), is presented which acquires abductive rules from a set of training examples. The goal is to find a small knowledge base which, when used abductively, diagnoses the training examples correctly and generalizes well to unseen examples. This contrasts with past systems that inductively learn rules that are used deductively. Each training example is associated with potentially multiple categories (disorders), instead of one as with typical learning systems. LAB uses a simple hill-climbing algorithm to efficiently build a rule base for a set-covering abductive system. LAB has been experimentally evaluated and compared to other learning systems and an expert knowledge base in the domain of diagnosing brain damage due to stroke.

  6. Inductive Learning For Abductive Diagnosis
    Cynthia A. Thompson
    M.A. Thesis, Department of Computer Sciences, University of Texas at Austin, July 1993.

    A new system for learning by induction, called LAB, is presented. LAB (Learning for ABduction) learns abductive rules based on a set of training examples. Our goal is to find a small knowledge base which, when used abductively, diagnoses the training examples correctly, in addition to generalizing well to unseen examples. This is in contrast to past systems, which inductively learn rules which are used deductively. Abduction is particularly well suited to diagnosis, in which we are given a set of symptoms (manifestations) and we want our output to be a set of disorders which explain why the manifestations are present. Each training example is associated with potentially multiple categories, instead of one, which is the case with typical learning systems. Building the knowledge base requires a choice between multiple possibilities, and the number of possibilities grows exponentially with the number of training examples. One method of choosing the best knowledge base is described and implemented. The final system is experimentally evaluated, using data from the domain of diagnosing brain damage due to stroke. It is compared to other learning systems and a knowledge base produced by an expert. The results are promising: the rule base learned is simpler than the expert knowledge base and rules learned by one of the other systems, and the accuracy of the learned rule base in predicting which areas are damaged is better than all the other systems as well as the expert knowledge base.