CSCE 470 Lecture 34

« previous | Sunday, November 17, 2013 | next »

Exam Preparation

Big-8 Topics:

1. Vector Space Retrieval (TF-IDF + Cosine)
2. Evaluation (Precision, Recall, F-Measure, NDCG)
3. Statistical Properties of Text (Heaps, Zipfs)
4. Link Analysis (PageRank, Hubs and Authorities)
5. Clustering (K-Means, maybe hierarchical)
6. Classification (Rocchio, KNN, Naive Bayes)
7. Recommenders (Content-Based, Collaborative Filtering)
8. Extra (Learning to Rank, Location and Geo)

Question Answering

Dan Jurafsky

Google gives you 10 blue links to nagivate through. Most times, we just want a single answer.

Example:

• Siri
• WolframAlpha

Complex Questions

• In children with an acute febrile illnese, what is the efficacy of acetaminophen in reducing fever?
• What do scholars think about Jefferson's position on dealing with pirates?

"Factoid" Questions

• Who wrote "The Universal Declaration of Human Rights"?
• How many calories are there in two slices of apple pie?
• What is the average age of the onset of autism?
• Where is Apple Computer based?

two approaches

1. IR (relevant to our class)
   • TREC
   • IBM Watson
   • Google
2. Knowledge-based (more of an AI topic)
   • Siri
   • Evi

IR Factoid Q/A

1. Find a bunch of relevant documents
2. pull out passage windows containing query terms
3. filter top words that are close by

Further analysis:

• coarse topics
  1. description
  2. location
  3. abbreviation
  4. entity
  5. numeric
  6. human
• finer topics underneath each coarse topic