Information Extraction

Question 1

Information Extraction

Answer 1

Turn unstructured information into structured data, in order to make the information more accessible to machines and humans

Question 2

Named entity recognition

Definition and Two Approaches

Answer 2

Labeling of categories such as
people, organizations, locations

Handled as a supervised learning task

A sequence labeling task much like part-of-speech tagging (using Hidden Markov Models) to describe connections between tags and words

Question 3

Relation extraction

Answer 3

Extraction of relations between entities

Use textual patterns (as with hypernymy)

Question 4

Semantic drift

Answer 4

Compute a confidence score for each pattern, based on:

• The number of already known tuples it finds (hits & misses)
• Productivity: the overall number of tuples that the pattern produces

Question 5

What information is used by named entity extraction?

Answer 5

• orthographical shape (capitalized?)
• predictive tokens (Mrs.)
• bags of words

Question 6

Bootstrapping in relation extraction

Answer 6

Use bootstrapping
- Start with seed tuples
- Acquire patterns from a corpus
OR
- Start with seed patterns
- Acquire tuples