Evaluation of Classification Algorithms and Features for Collocation Extraction in Croatian

Vanja Mladen Karan, Jan Šnajder, Bojana Dalbelo Bašić


Abstract
Collocations can be defined as words that occur together significantly more often than it would be expected by chance. Many natural language processing applications such as natural language generation, word sense disambiguation and machine translation can benefit from having access to information about collocated words. We approach collocation extraction as a classification problem where the task is to classify a given n-gram as either a collocation (positive) or a non-collocation (negative). Among the features used are word frequencies, classical association measures (Dice, PMI, chi2), and POS tags. In addition, semantic word relatedness modeled by latent semantic analysis is also included. We apply wrapper feature subset selection to determine the best set of features. Performance of various classification algorithms is tested. Experiments are conducted on a manually annotated set of bigrams and trigrams sampled from a Croatian newspaper corpus. Best results obtained are 79.8 F1 measure for bigrams and 67.5 F1 measure for trigrams. The best classifier for bigrams was SVM, while for trigrams the decision tree gave the best performance. Features which contributed the most to overall performance were PMI, semantic relatedness, and POS information.
Anthology ID:
L12-1470
Volume:
Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12)
Month:
May
Year:
2012
Address:
Istanbul, Turkey
Editors:
Nicoletta Calzolari, Khalid Choukri, Thierry Declerck, Mehmet Uğur Doğan, Bente Maegaard, Joseph Mariani, Asuncion Moreno, Jan Odijk, Stelios Piperidis
Venue:
LREC
SIG:
Publisher:
European Language Resources Association (ELRA)
Note:
Pages:
657–662
Language:
URL:
http://www.lrec-conf.org/proceedings/lrec2012/pdf/796_Paper.pdf
DOI:
Bibkey:
Cite (ACL):
Vanja Mladen Karan, Jan Šnajder, and Bojana Dalbelo Bašić. 2012. Evaluation of Classification Algorithms and Features for Collocation Extraction in Croatian. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12), pages 657–662, Istanbul, Turkey. European Language Resources Association (ELRA).
Cite (Informal):
Evaluation of Classification Algorithms and Features for Collocation Extraction in Croatian (Karan et al., LREC 2012)
Copy Citation:
PDF:
http://www.lrec-conf.org/proceedings/lrec2012/pdf/796_Paper.pdf