CODEWITHSUNDEEP
nlpwordnet
B bonita
B bonita

Reputation: 181

How to extract words based on wordnet event synset?

I want to implement "Template-based information extraction without the templates" paper and for the first step, I have to do "clustering Events". I have the MUC dataset and have parsed and tokenized it. But there is something that I cannot understand. In the paper, it says that : "We cluster event patterns to create templates. An event pattern is either (1) a verb, (2) a noun in WordNet under the Event synset, or (3) a verb and the head word of its syntactic object. Examples of each include (1) ‘explode’, (2) ‘explosion’, and (3) ‘explode:bomb’." I was wondering what are these three conditions. How can I apply them to cluster my data? I mean should I first go search in wordnet and remove some words? I just do not know how to achieve the goal. Any help will be appreciated.

Upvotes: 0

Views: 242

Answers (1)

Adeeb
Adeeb

Reputation: 1261

I am writing this reply to help future readers.

You've posed 4 questions. I will try and address all of them. All your questions are surrounding a singular portion of the paper: the approximation of template topics using what the authors term "event patterns".

  1. I was wondering what are these three conditions

The three "conditions" are the definition of event patterns (This is explained in the paper --- its also in the text you've quoted).

The authors define an event pattern as (a) a verb, (b) a noun you can extract from an event's synset on WordNet (recall that there are six types of events that have templates in the corpus i.e. bombing, kidnap, attack, and others), and (c) the concatenation of the verb with the head of its object.

  1. How can I apply them to cluster my data

You're not using them to cluster your "data" -- they are the data to be clustered. You're supposed to extract these so-called event patterns from the documents and cluster them using the two algorithms discussed in [2, p978-979]. The documents are from the terrorism corpus [1].

  1. I mean should I first go search in wordnet and remove some words?

No. You're supposed to determine what nouns belong to an event's synset on Wordnet.

  1. I just do not know how to achieve the goal

You need to improve the way you read a paper. Read a paper and make notes. After that, re-read the paper and revisit the questions + points you jotted down. See [3] for tips. If you read it in that manner, you should be able to get sufficient understanding to achieve your goal.

Refs

  1. Beth Sundheim, 1991. Third message understanding evaluation and conference (MUC-3): Phase 1 status report. In proceedings of the message understanding conference
  2. Nathaneal Chambers, Dan Jurafsky, 2011. Template-based Information extraction without the templates. In proceedings of the 49th Meeting of the Association for Computational Linguistics
  3. https://www.eecs.harvard.edu/~michaelm/postscripts/ReadPaper.pdf

Upvotes: 1

Related Questions