A semiparametric generative model for efficient structured-output supervised learning

  • Fabrizio Costa
  • Andrea Passerini
  • Marco Lippi
  • Paolo Frasconi


We present a semiparametric generative model for supervised learning with structured outputs. The main algorithmic idea is to replace the parameters of an underlying generative model (such as a stochastic grammars) with input-dependent predictions obtained by (kernel) logistic regression. This method avoids the computational burden associated with the comparison between target and predicted structure during the training phase, but requires as an additional input a vector of sufficient statistics for each training example. The resulting training algorithm is asymptotically more efficient than structured output SVM as the size of the output structure grows. At the same time, by computing parameters of a joint distribution as a function of the full input structure, typical expressiveness limitations of related conditional models (such as maximum entropy Markov models) can be potentially avoided. Empirical results on artificial and real data (in the domains of natural language parsing and RNA secondary structure prediction) show that the method works well in practice and scales up with the size of the output structures.


Semiparametric generative model Supervised learning Structured outputs 

Mathematics Subject Classification (2000)



Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Bosco. C., Lombardo, V., Vassallo, D., Lesmo, L.: Building a treebank for Italian: a data-driven annotation schema. In: Proceedings of the Second International Conference on Language Resources and Evaluation LREC, pp. 99–106, Athens, 31 May–2 June 2000Google Scholar
  2. 2.
    Collins, M.: Parameter estimation for statistical parsing models: theory and practice of distribution-free methods. In: New Developments in Parsing Technology, pp. 19–55. Kluwer Academic, Norwell (previusly IWPT 2001) (2004)Google Scholar
  3. 3.
    Cortes, C., Mohri, M., Weston, J.: A general regression technique for learning transductions. In: Proceedings of the 22nd International Conference on Machine Learning, pp. 153–160, Bonn, 7–11 August 2005Google Scholar
  4. 4.
    Johnson, M.: PCFG models of linguistic tree representations. Comput. Linguist. 24(4), 613–632 (1998)Google Scholar
  5. 5.
    Knudsen, B., Hein, J.: RNA secondary structure prediction using stochastic context-free grammars and evolutionary history. Bioinformatics 15(6), 446–454 (1999)CrossRefGoogle Scholar
  6. 6.
    Lafferty, J.D., McCallum, A., Pereira, F.C.N.: Conditional random fields: probabilistic models for segmenting and labeling sequence data. In: ICML ’01: Proceedings of the Eighteenth International Conference on Machine Learning, pp. 282–289. Morgan Kaufmann, San Francisco (2001)Google Scholar
  7. 7.
    Leslie, C., Eskin, E., Noble, W.: The spectrum kernel: a string kernel for svm protein classification. In: Proc. of the Pacific Symposium on Biocomputing, pp. 564–575, Lihue, 3–7 January 2002Google Scholar
  8. 8.
    Manning, C.D., Schütze, H.: Foundations of Statistical Natural Language Processing. MIT, Cambridge (1999)zbMATHGoogle Scholar
  9. 9.
    McAllester, D.: Generalization bounds and consistency for structured labeling. In: Bakir, G., Hofmann, T., Schölkopf, B., Smola, A., Taskar, B., Vishwanathan, S.V.N. (eds.) Predicting Structured Data. MIT, Cambridge (2007)Google Scholar
  10. 10.
    McCallum, A., Freitag, D., Pereira, F.C.N.: Maximum entropy markov models for information extraction and segmentation. In: Langley, P. (ed.) ICML, pp. 591–598. Morgan Kaufmann, San Francisco (2000)Google Scholar
  11. 11.
    Menchetti, S., Costa, F., Frasconi, P.: Weighted decomposition kernels. In: Proceedings of the Twenty-second International Conference on Machine Learning (ICML’05), pp. 585–592. ACM, New York (2005)Google Scholar
  12. 12.
    Sakakibara, Y., Brown, M., Hughey, R., Mian, I.S., Sjölander, K., Underwood, R.C., Haussler, D.: Stochastic context-free grammars for tRNA modeling. Nucleic Acids Res. 22, 5112–5120 (1994)CrossRefGoogle Scholar
  13. 13.
    Taskar, B., Guestrin, C., Koller, D.: Max-margin markov networks. In: Advances in Neural Information Processing Systems (NIPS 2003), Vancouver, 13–18 December 2004Google Scholar
  14. 14.
    Tsochantaridis, I., Joachims, T., Hofmann, T., Altun, Y.: Large margin methods for structured and interdependent output variables. J. Mach. Learn. Res. 6, 1453–1484 (2005)MathSciNetGoogle Scholar
  15. 15.
    Weston, J., Chapelle, O., Elisseeff, A., Scholkopf, B., Vapnik, V.: Kernel dependency estimation. Adv. Neural Inf. Process. Syst. 15, 873–880 (2003)Google Scholar
  16. 16.
    Zhu, J., Hastie, T.: Kernel logistic regression and the import vector machine. In: Advances in Neural Information Processing Systems (NIPS 2001), pp. 1081–1088, Vancouver, 3–8 December 2001Google Scholar

Copyright information

© Springer Science+Business Media B.V. 2009

Authors and Affiliations

  • Fabrizio Costa
    • 1
  • Andrea Passerini
    • 1
  • Marco Lippi
    • 1
  • Paolo Frasconi
    • 1
  1. 1.Dipartimento di Sistemi e InformaticaUniversità degli Studi di FirenzeFirenzeItaly

Personalised recommendations