0

Taking the following sentence as an exmaple (gotten from GATE official tutorial slide:module 11 https://gate.ac.uk/sale/talks/gate-course-may10/track-3/module-11-ml-adv/):

I was told the item was in stock and next day delivery. After a couple of days i chased them to be told there was an error, none in stock. Chased again after a few days, still no stock then found out that they had billed me right from the beginning. So they have had my money for over a week - have no idea when they will have stock- and i have done all the chasing. Despite what others say in eight years of Internet shopping worst experinece to date bar none.

When the whole sentence is tagged as an anntation and treated as an instance in GATE machine learning PR(the Batch Learning), how the GATE would process the learning process?

I have two guesses. One is the GATE automatically tags every words in the sentence with human-language meanings and collects these features to build a classification model. The other is GATE simply transoforms the sentence into a mathematically variable, like vector and trains the model according to linguist features, like how many nouns, adj, adv are used.

I am not sure which one is right or whether there will be another explanation for the process. Hope someone could give a confirmation or any releated information.

Thank you!

Fan Yang
  • 23
  • 3

1 Answers1

0

GATE build a model based on your definitions in configuration xml file. Except instance type there could be annotations like Token(with features root,POS,..) or N-grams or your own annotations (with features) created by JAPE rules. Gate will use thee definitions and will build a model.

ashingel
  • 494
  • 3
  • 11
  • Thanks ashingel. But if the instance or my own annotation are phrase or sentence, will them be split into single words to conduct feature extraction for machine learning? – Fan Yang Mar 28 '16 at 13:36