jurafsky&martin_3rdEd_17 (1).pdf

Finally for simplicity we have illustrated this

Info icon This preview shows pages 254–256. Sign up to view the full content.

Finally, for simplicity, we have illustrated this example without the labels on the dependency relations. To produce labeled trees, we can parameterize the LEFT - A RC and RIGHT A RC operators with dependency labels, as in LEFT A RC ( NSUBJ ) or RIGHT A RC ( DOBJ ). This is equivalent to expanding the set of transition operators from our original set of three to a set that includes LEFT A RC and RIGHT A RC opera- tors for each relation in the set of dependency relations being used, plus an additional one for the SHIFT operator. This, of course, makes the job of the oracle more difficult since it now has a much larger set of operators from which to choose.
Image of page 254

Info icon This preview has intentionally blurred sections. Sign up to view the full version.

14.4 T RANSITION -B ASED D EPENDENCY P ARSING 255 14.4.1 Creating an Oracle State-of-the-art transition-based systems use supervised machine learning methods to train classifiers that play the role of the oracle. Given appropriate training data, these methods learn a function that maps from configurations to transition operators. As with all supervised machine learning methods, we will need access to appro- priate training data and we will need to extract features useful for characterizing the decisions to be made. The source for this training data will be representative tree- banks containing dependency trees. The features will consist of many of the same features we encountered in Chapter 10 for part-of-speech tagging, as well as those used in Chapter 13 for statistical parsing models. Generating Training Data Let’s revisit the oracle from the algorithm in Fig. 14.6 to fully understand the learn- ing problem. The oracle takes as input a configuration and returns as output a tran- sition operator. Therefore, to train a classifier, we will need configurations paired with transition operators (i.e., LEFT A RC , RIGHT A RC , or SHIFT ). Unfortunately, treebanks pair entire sentences with their corresponding trees, and therefore they don’t directly provide what we need. To generate the required training data, we will employ the oracle-based parsing algorithm in a clever way. We will supply our oracle with the training sentences to be parsed along with their corresponding reference parses from the treebank. To produce training instances, we will then simulate the operation of the parser by run- ning the algorithm and relying on a new training oracle to give us correct transition Training oracle operators for each successive configuration. To see how this works, let’s first review the operation of our parser. It begins with a default initial configuration where the stack contains the ROOT , the input list is just the list of words, and the set of relations is empty. The LEFT A RC and RIGHT A RC operators each add relations between the words at the top of the stack to the set of relations being accumulated for a given sentence. Since we have a gold-standard reference parse for each training sentence, we know which dependency relations are valid for a given sentence. Therefore, we can use the reference parse to guide the selection of operators as the parser steps through a sequence of configurations.
Image of page 255
Image of page 256
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern