Difference between revisions of "Vulcan/TextualEvidenceFinder"

Revision as of 21:21, 20 August 2013

I/O

Input: A Proposition [A natural language sentence + Open IE tuples from the sentence.]

Output: A list of query/score pairs representing evidence for the proposition.

Components

System Architecture: Weak Evidence Finder

Query Generator

The query generator outputs two types of queries for each proposition:

1. Keyword queries -- Extract keywords from the query sentence [TBD: Stemming? Stopword removal?]
2. Template queries -- A template query is simply a tuple (or the sentence) where one or more words in the tuple is replaced with a wild-card operator.

The system will be given a set of rules that specify how to convert a tuple into different template queries. Start with two rules:

1. Keyword queries -- Remove stopwords.
2. Template queries -- Take each tuple. For each field (arg1, rel and arg2), if it is a multi-word query replace each word with a wild-card.

Examples

Input: 

      Sentence: Iron nail is a good conductor of electricity
        Tuples: (iron nail, is a good conductor of, electricity)

Output: 
      Q1: (iron *, is a good conductor of, electricity) //Template query
      Q2: (*  nail, is a good conductor of, electricity)//Template query
      Q3: (iron nail, is a * conductor of, electricity) //Template query
      Q4: iron * conductor * electricity                //Template query 
      Q5: iron or conductor or electricity              //Keyword query

Solr/Lucence Layer

Open IE 4.0

Use Open IE 4.0.