D15-1210 grow-diag-final heuristic to generate symmetric alignments . 5.2 Comparison with GIZA +
D09-1106 cost matrices " for producing symmetric alignments . Kumar et al. ( 2007 ) describe
C04-1032 combination heuristic to obtain a symmetric alignment . The results of these experiments
J10-3002 demonstrates the results of the symmetric alignment model on both tasks . As the
C04-1032 graph algorithms to determine the symmetric alignment with minimal total costs ( i.e.
J10-3002 knowledge source . To model a symmetric alignment , a straightforward way is to
J10-3002 focusing on the features that produce symmetric alignments . In Section 4 , we evaluate
J10-3002 offeatures that could produce symmetric alignments . Our model is easy to extend
C04-1032 approach to the task of producing a symmetric alignment . In the experiments with the
J10-3002 directly . As our goal is to produce symmetric alignments , we calculate the product of
C04-1032 algorithm is not able to produce a symmetric alignment , it operates with symmetrized
D15-1210 grow-diag-final heuristic to generate symmetric alignments . For BERKELEY , we trained joint
C04-1031 lower F-values than other refined symmetric alignment strategies . Their implementation
C04-1032 Figure 1 shows an example of a symmetric alignment ) . Another important advantage
C04-1032 relationship types characterizes a symmetric alignment that can potentially improve
D15-1210 links in the intersection of two symmetric alignments or two symmetric models agree
J10-3002 addition , our approach supports symmetric alignment modeling that allows for an arbitrary
J10-3002 the asymmetric model can produce symmetric alignments via symmetrization heuristics
C04-1032 polynomial time . The task is to find a symmetric alignment A , for which the costsc ( A
J10-3002 add features that characterize symmetric alignments . In the following subsections
hide detail