02.04.2013 Views

Statistical Language Models based on Neural Networks - Faculty of ...

Statistical Language Models based on Neural Networks - Faculty of ...

Statistical Language Models based on Neural Networks - Faculty of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

C<strong>on</strong>tents<br />

1 Introducti<strong>on</strong> 4<br />

1.1 Motivati<strong>on</strong> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4<br />

1.2 Structure <strong>of</strong> the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />

1.3 Claims <strong>of</strong> the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7<br />

2 Overview <strong>of</strong> <str<strong>on</strong>g>Statistical</str<strong>on</strong>g> <str<strong>on</strong>g>Language</str<strong>on</strong>g> Modeling 9<br />

2.1 Evaluati<strong>on</strong> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />

2.1.1 Perplexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />

2.1.2 Word Error Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br />

2.2 N-gram <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16<br />

2.3 Advanced <str<strong>on</strong>g>Language</str<strong>on</strong>g> Modeling Techniques . . . . . . . . . . . . . . . . . . . 17<br />

2.3.1 Cache <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . . . . . . . . . . 18<br />

2.3.2 Class Based <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . . . . . . . . . . . . 19<br />

2.3.3 Structured <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . . . . . . . 20<br />

2.3.4 Decisi<strong>on</strong> Trees and Random Forest <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . 22<br />

2.3.5 Maximum Entropy <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . . . 22<br />

2.3.6 <strong>Neural</strong> Network Based <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> . . . . . . . . . . . . . . . . 23<br />

2.4 Introducti<strong>on</strong> to Data Sets and Experimental Setups . . . . . . . . . . . . . 24<br />

3 <strong>Neural</strong> Network <str<strong>on</strong>g>Language</str<strong>on</strong>g> <str<strong>on</strong>g>Models</str<strong>on</strong>g> 26<br />

3.1 Feedforward <strong>Neural</strong> Network Based <str<strong>on</strong>g>Language</str<strong>on</strong>g> Model . . . . . . . . . . . . . 27<br />

3.2 Recurrent <strong>Neural</strong> Network Based <str<strong>on</strong>g>Language</str<strong>on</strong>g> Model . . . . . . . . . . . . . . 28<br />

3.3 Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

3.3.1 Backpropagati<strong>on</strong> Through Time . . . . . . . . . . . . . . . . . . . . 33<br />

3.3.2 Practical Advices for the Training . . . . . . . . . . . . . . . . . . . 35<br />

3.4 Extensi<strong>on</strong>s <strong>of</strong> NNLMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

1

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!