Foundations

Question 1

Example of a Transactional Query

Answer 1

“Best Budget Headphones”

Question 2

Informational Query

Answer 2

Query where the user does not know what they are looking for, but searches for reading material on the given topic.

Question 3

Example of Informational Query

Answer 3

“Books on space”

Question 4

Transactional Queries

Answer 4

Queries when we know what we are looking for and want to find places where we can get that.

Question 5

Question Answering Queries

Answer 5

Queries where we ask a question we don’t know the answer to, and answers are short, verifiable, and specific.

Question 6

3 Main uses of Search Engines

Answer 6

Personal Search: Finding documents within your own personal computer
Enterprise Search: Finding information within an outside server/organization firewall (e.g., library catalog)
Vertical Search: Finding information across public websites about a specific topic (e.g., Yelp)

Question 7

Documents

Answer 7

Storage and retrieval unit of a search system. The system will always return documents or links to them.

Question 8

Example of Documents in a system

Answer 8

In Gmail, search returns links to emails, so emails are considered documents.

Question 9

Query

Answer 9

Expression of information need that the user has.
The input to a search engine.

Question 10

Tokens

Answer 10

Token: Internal representation of a word in a document or query.
Tokenization: Process of converting words into tokens.

Question 11

Terms

Answer 11

Vocabulary (V): Set of unique tokens across all documents.
Term: Each element in the vocabulary set.

Question 12

Out-of-Vocabulary (OOV)

Answer 12

A token appearing in a query but not in any document. IR systems typically ignore OOV tokens.

Question 13

Tokenization Benefits

Answer 13

Helps find documents when users type words differently.
Removes inconsistencies from documents.

Question 14

Architecture of a Search Engine

Answer 14

Documents added to the index.
Users send queries and receive results.
Matching tokens in the query against terms in the index to find relevant documents.

Question 15

Term-Document Incidence Matrix

Answer 15

Matrix indicating the importance of a term for a document. Common weight choices: Boolean, Counts, tf-idf.

Question 16

Index

Answer 16

Associative data structure on disk, allowing quick retrieval of objects given search keys.

Question 17

Boolean Queries

Answer 17

Each term is a boolean predicate.
True for being in the document, false otherwise.

Question 18

Pros of Boolean Query

Answer 18

Easy to understand, implement, and explain.

Question 19

Cons of Boolean Query

Answer 19

Hard to write, treats all documents and terms the same.

Question 20

Querying by Ranking

Answer 20

Scoring function should prioritize the number of query terms.
More appearances of a term in a document contribute more to the score.

Question 21

Scoring with Vectors

Answer 21

Documents and queries represented as vectors.
Score computed using document and query vectors.

A document’s score is influenced by the terms it shares with the query.

Question 22

Evaluating Quality of Answers

Answer 22

Benchmark with queries and answers by humans.
Precision: Fraction of returned documents that are relevant.
Recall: Fraction of relevant documents that are returned.

Question 23

Precision and Recall Trade-off

Answer 23

Precision and recall are somewhat opposites.
Perfect recall is achieved by returning every document.

Question 24

When should you prioritize Precision and Recall?

Answer 24

When to Prioritize:

Precision: When a bad answer can have consequences.
Recall: When missing a good answer can have consequences.