-
Notifications
You must be signed in to change notification settings - Fork 2
Project Description
doodyparizada edited this page Aug 8, 2011
·
4 revisions
- receive RTE6
- parse structure and create an object/list
- this might be done effortlessly with json.loads
- enumerate on hypothesis and given sentences
- for each input hypothesis and candidate sentence.
- for each word in the sentence (or hypo, or both?) create a vector with the entailing words (hyponyms)
- using wordnet and another resource
- try to match each word in the hypo to a word in the sentence
- this should be done by POS, ie. a noun in the hypo will only be matched to a noun in the sentence
- the previous step creates a feature vector, which will be used to decide whether the sentence entails the hypo
- this step can be done by guessing (with manual limits)
- or, using a machine learning approche. this depends on Ido Dagan's answer...
- mark V or X for each sentence
- at the end grade ourselves against the given answers to determine recall and precision
- language of choice
- python
- java
- who does what
- system design