Surface Realization by Drew Jacobs What does it do? Derive a human readable sentence from a discourse plan. Discourse plan does not give syntax, only functional.
Download ReportTranscript Surface Realization by Drew Jacobs What does it do? Derive a human readable sentence from a discourse plan. Discourse plan does not give syntax, only functional.
Surface Realization by Drew Jacobs What does it do? Derive a human readable sentence from a discourse plan. Discourse plan does not give syntax, only functional information. The Surface Realizer adds syntactical information and assures that the sentence will comply with lexical and grammatical constraints. What doesn’t it do? Will not verify that the correctness of the data provided by the discourse planner or that the information makes sense. Does not deal with more than one sentence at a time. If the plan calls for many sentences, the surface realizer will be called once for each sentence required. Simple Surface Realization Tools Canned Text Systems - Takes a given input and matches it directly to a pre-made sentence. - Commonly used in simple systems such as error messages or warnings. - Has no flexibility whatsoever. Simple Surface Realization (cont.) Template Systems - The idea of a template is that there are premade sentences with fill in the blank words that are filled in by the input. - These systems work well with Form Letters and Slightly more advanced Error or Warning Messages. - They are still very inflexible, but better than canned text systems. Two More Advanced Approaches (Feature Based Systems) Systemic Grammar – Representation of sentences as collections of functions. Rules allow mapping from functions to grammatical forms. (Halliday, 1985) Functional Unification Grammar – Represents sentences as feature structures that can be combined and altered to produce sentences. (Kay, 1979) Systemic Grammar A systemic grammar comes from branch of linguistics called Systemic-Functional Linguistics Described by M.A.K. Halliday in the book “An Introduction to Functional Grammar” (1985) Systemic Grammars work by utilizing rules on functions in layers to generate actual sentences Mood Layer Also known as the interpersonal meta-function. It describes the interaction between the sentence writer and the reader. Examples would be the whether the writer is telling the reader something or is asking a question. Transitivity Layer Also known as the ideational meta-function. It identifies items such as who the actors are and what the goals are for the sentence. Also identifies the type of process being performed. Theme Layer Also known as the textual meta-function. This layer tries to fit the expression with a given theme and reference. Systemic Grammar Rules Within the layers, there exist realization statements that allow for transition from functional form to syntactic form. These statements contain operators for adding syntactic objects to the sentence. Operators take the form of : +X for inserting a function X=Y for assigning equivalency X > Y for saying X must come before Y X/Y for assigning a syntactic expression Y to a function X. Figure 20-2 From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin, Prentice Hall, 2000. Example Input “The System Will Save the Document” Input to a systemic grammar would take this form : :process save :actor system :goal document :speechact assertion :tense future From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin, Prentice Hall, 2000. Example Sentence Here we can see how each layer corresponds to the text. From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin, Prentice Hall, 2000. Functional Unification Grammar This method of NLG works using the principles of Unification grammars. These grammars were shown in Chapter 11 to describe a parsing method. It works by building a feature structure and a list of potential alternations for the structure. It then matches the feature structure of the input with the feature structures in the grammar. Categories, Elements, and Patterns The Category in a feature structure represents what the feature structure corresponds to. For example S is a sentence, and NP is a noun phrase. Each of the elements in a feature structure are the components that make up the feature that is being represented. So a Process or a Goal would be an element of a Sentence, and so forth. Patterns describe the ordering of elements in a feature structure. So, for example, a Noun Phrase could have the pattern (Det, PrNoun). Figure 20-3 From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin. Prentice Hall, 2000. An Example Unification An Example Input (Functional Description) From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin. Prentice Hall, 2000. Figure 20-4 From “”Speech and Language Processing”, Daniel Jurafsky & James H. Martin. Prentice Hall, 2000. Natural Language Generation Programs KPML – a text generation system based off of the earlier Penman system. Uses Systemic-Functional Linguistics Principles. http://www.fb10.uni-bremen.de/anglistik/langpro/kpml/README.html FUF/SURGE – a text generation system and English Grammar using Functional Unification. http://www.cs.bgu.ac.il/research/projects/surge/index.htm Additional Resources http://registry.dfki.de/ - A comprehensive listing of current NLP tools. http://www.dynamicmultimedia.com.au/siggen/ - The website for the Special Interest Group on Generation. http://www.fb10.uni-bremen.de/anglistik/langpro/NLG-table/NLGtable-root.htm - A website with a full listing of all programs specifically used for Natural Language Generation. References Halliday, M.A.K. “An Introduction to Functional Grammar”. Edward Arnold, London 1985. Hovy, Eduard. “Language Generation.” http://www.lt-world.org/HLT_Survey/ltw-chapter4-all.pdf Jurafsky, Daniel & Martin, James H. “Speech and Language Processing”. Prentice Hall, New York 2000.