Filters

2020 - 20232

More filters

2021 - 20242
Reset filters

Settings

search.filters.applied.f.access: openAccess × Subject: Statistical methods ×

Search Results

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Ranking facts for explaining answers to elementary science questions
    (Cambridge : Cambridge University Press, 2023) D’Souza, Jennifer; Mulang, Isaiah Onando; Auer, Sören
    In multiple-choice exams, students select one answer from among typically four choices and can explain why they made that particular choice. Students are good at understanding natural language questions and based on their domain knowledge can easily infer the question's answer by “connecting the dots” across various pertinent facts. Considering automated reasoning for elementary science question answering, we address the novel task of generating explanations for answers from human-authored facts. For this, we examine the practically scalable framework of feature-rich support vector machines leveraging domain-targeted, hand-crafted features. Explanations are created from a human-annotated set of nearly 5000 candidate facts in the WorldTree corpus. Our aim is to obtain better matches for valid facts of an explanation for the correct answer of a question over the available fact candidates. To this end, our features offer a comprehensive linguistic and semantic unification paradigm. The machine learning problem is the preference ordering of facts, for which we test pointwise regression versus pairwise learning-to-rank. Our contributions, originating from comprehensive evaluations against nine existing systems, are (1) a case study in which two preference ordering approaches are systematically compared, and where the pointwise approach is shown to outperform the pairwise approach, thus adding to the existing survey of observations on this topic; (2) since our system outperforms a highly-effective TF-IDF-based IR technique by 3.5 and 4.9 points on the development and test sets, respectively, it demonstrates some of the further task improvement possibilities (e.g., in terms of an efficient learning algorithm, semantic features) on this task; (3) it is a practically competent approach that can outperform some variants of BERT-based reranking models; and (4) the human-engineered features make it an interpretable machine learning model for the task.
  • Thumbnail Image
    Item
    Two-Dimensional Partial-Covariance Mass Spectrometry of Large Molecules Based on Fragment Correlations
    (College Park, Md. : APS, 2020) Driver, Taran; Cooper, Bridgette; Ayers, Ruth; Pipkorn, Rüdiger; Patchkovskii, Serguei; Averbukh, Vitali; Klug, David R.; Marangos, Jon P.; Frasinski, Leszek J.; Edelson-Averbukh, Marina
    Covariance mapping [L. J. Frasinski, K. Codling, and P. A. Hatherly, Science 246, 1029 (1989)] is a well-established technique used for the study of mechanisms of laser-induced molecular ionization and decomposition. It measures statistical correlations between fluctuating signals of pairs of detected species (ions, fragments, electrons). A positive correlation identifies pairs of products originating from the same dissociation or ionization event. A major challenge for covariance-mapping spectroscopy is accessing decompositions of large polyatomic molecules, where true physical correlations are overwhelmed by spurious signals of no physical significance induced by fluctuations in experimental parameters. As a result, successful applications of covariance mapping have so far been restricted to low-mass systems, e.g., organic molecules of around 50 daltons (Da). Partial-covariance mapping was suggested to tackle the problem of spurious correlations by taking into account the independently measured fluctuations in the experimental conditions. However, its potential has never been realized for the decomposition of large molecules, because in these complex situations, determining and continuously monitoring multiple experimental parameters affecting all the measured signals simultaneously becomes unfeasible. We introduce, through deriving theoretically and confirming experimentally, a conceptually new type of partial-covariance mapping—self-correcting partial-covariance spectroscopy—based on a parameter extracted from the measured spectrum itself. We use the readily available total ion count as the self-correcting partial-covariance parameter, thus eliminating the challenge of determining experimental parameter fluctuations in covariance measurements of large complex systems. The introduced self-correcting partial covariance enables us to successfully resolve correlations of molecules as large as 103–104  Da, 2 orders of magnitude above the state of the art. This opens new opportunities for mechanistic studies of large molecule decompositions through revealing their fragment-fragment correlations. Moreover, we demonstrate that self-correcting partial covariance is applicable to solving the inverse problem: reconstruction of a molecular structure from its fragment spectrum, within two-dimensional partial-covariance mass spectrometry.