Open Information Extraction From The Web

1
Open Information Extraction From The Web

• Rani Qumsiyeh

2
What is Information Extraction

• This article surveys a range of Information
  Extraction methods. (Particularly Open)
• A venerable technology that maps natural language
  text into structured relational data.
• Open Information Extraction is where the
  identities of the relations to be extracted are
  unknown and the billions of documents found on
  the Web necessitate highly scalable processing.

3
Most common Ways to do IE

• Direct knowledge-based encoding.
• A human enters regular expressions or rules.
• Supervised learning.
• A human provides labeled training examples.
• Self-supervised learning.
• The system automatically finds and labels its own
  examples.

4
Direct Knowledge

• Not efficient, has to be altered for different
  domains.
• Class PhysicalTarget space to the term bank in
  the terrorism domain.
• Class Corporation in the joint-ventures domain

5
Example of Supervised Learning
6
Self Supervised Knowledge

• A system that labels its own training examples.
  (Example KnowItAll)
• For a given relation
• Use generic pattern ? instantiate
  relation-specific extraction rules ? learn
  domain-specific extraction rules ? apply rules to
  web pages and assign them probabilities.
• Example X is a Y (X is a country).
• China is a country.
• Garth Brooks is a country singer

7
Open Information Extraction

• The challenge of Web extraction is to be able to
  do Open Information Extraction.
• Unbounded number of relations
• Web corpus contains billions of documents.

8
How open IE systems work

• learn a general model of how relations are
  expressed (in a particular language), based on
  unlexicalized features such as part-of-speech
  tags. (Identify a verb)
• Learn domain-independent regular expressions.
  (Punctuations, Commas).

9
Is there a general model of relationships in
English
10
TextRunner

• Works in two phases.
• Using a conditional random field, the extractor
  learns to assign labels to each of the words in a
  sentence.
• Extracts one or more textual triples that aim to
  capture (some of) the relationships in each
  sentence.

11
Additional Tasks to Accomplish

• Opinion mining in which Open IE can extract
  opinion information about particular objects
  (including products, political candidates, and
  more) that are contained in blog, posts, reviews,
  and other texts.
• Fact checking in which Open IE can identify
  assertions that directly or indirectly conflict
  with the body of knowledge extracted from the Web
  and various other knowledge bases.