310 likes | 464 Views
Prototype-Driven Grammar Induction. Aria Haghighi and Dan Klein Computer Science Division University of California Berkeley. Grammar Induction. DT NN VBD DT NN IN NN The screen was a sea of red. First Attempt. DT NN VBD DT NN IN NN
E N D
Prototype-Driven Grammar Induction Aria Haghighi and Dan Klein Computer Science Division University of California Berkeley
Grammar Induction DTNNVBD DT NN IN NN The screen was a sea of red
First Attempt DTNNVBD DT NN IN NN The screen was a sea of red
Central Questions • How do we specify what we want to learn? • How do we fix observed errors? What’s an NP? That’s not quite it!
Experimental Set-up • Binary Grammar { X1, X2, … Xn} plus POS tags • Data • WSJ-10 [7k sentences] • Evaluate on Labeled F1 • Grammar Upper Bound: 86.1 Xi Xj Xk
Experiment Roadmap • Unconstrained Induction • Need bracket constraint! • Gold Bracket Induction • Prototypes and Similarity • CCM Bracket Induction
Unconstrained PCFG Induction (Outside) 0 i j n (Inside) • Learn PCFG with EM • Inside-Outside Algorithm • Lari & Young [93] • Results
Constrained PCFG Induction • Gold Brackets • Periera & Schables [93] • Result
Encoding Knowledge What’s an NP? Semi-Supervised Learning
Encoding Knowledge What’s an NP? For instance, DT NN JJ NNS NNP NNP Prototype Learning
Grammar Induction Experiments • Add Prototypes • Manually constructed
How to use prototypes? S ? VP ? PP ? NP ? NP ? NN koala VBD sat IN in DT the NN tree DT The ¦ ¦
How to use prototypes? S VP NP ? PP NP JJ hungry NN koala VBD sat IN in DT the NN tree DT The ¦ ¦
Distributional Similarity • Context Distribution (DT JJ NN) = { ¦ __ VBD : 0.3, VBD __ ¦ : 0.2, IN __ VBD: 0.1, ….. } • Similarity (DT NN) (NNP NNP) (DT JJ NN) (JJ NNS) NP
Distributional Similarity • Prototype Approximation (NP) ¼ Uniform ( (DT NN), (JJ NNS), (NNP NNP) ) • Prototype Similarity Feature • span(DT JJ NN) emits proto=NP • span(MD NNS) emits proto=NONE
Prototype CFG+ Model S P (DT NP | NP) P (proto=NP | NP) VP NP PP NP NP JJ hungry NN koala VBD sat IN in DT the NN tree DT The ¦ ¦
Prototype CFG+ Induction • Experimental Set-Up • BLIPP corpus • Gold Brackets • Results
Summary So Far • Bracket constraint and prototypes give good performance!
Product Model • Different Aspects of Syntax • CCM : Yield and Context properties • CFG: Hierarchical properties • Intersected EM [Klein 2005] • Encourages mass on trees compatible with CCM and CFG
Grammar Induction Experiments • Intersected CFG and CCM • No prototypes • Results
Grammar Induction Experiments • Intersected CFG+ and CCM • Add Prototypes • Results
Reacting to Errors • Possessive NPs Our Tree Correct Tree
Reacting to Errors • Add Prototype: NP-POS NN POS New Analysis
Error Analysis • Modal VPs Our Tree Correct Tree
Reacting to Errors • Add Prototype: VP-INF VB NN New Analysis
Fixing Errors • Supplement Prototypes • NP-POS and VP-INF • Results
Conclusion • Prototype-Driven Learning Flexible Weakly Supervised Framework • Merged distributional clustering techniques with supervised structured models
Thank You! http://www.cs.berkeley.edu/~aria42
Unconstrained PCFG Induction (Outside) 0 i j n (Inside) Xi Xi Xi Xi • Binary Grammar { X1, X2, … Xn} • Learn PCFG with EM • Inside-Outside Algorithm • Lari & Young [93] Xj Xk N Xk Xj V N V