IR Module 5

According to the material, how many information needs are usually a sufficient minimum for a reasonable test collection?

• 10

• 25

• 50

• 100

Which evaluation series focuses on European languages and cross-language information retrieval?

• NTCIR

• CLEF

• TREC

• 20 Newsgroups

For text classification research, which corpus is the larger, more recent Reuters release consisting of 806,791 documents and offering richer annotation?

• Reuters-21578

• Reuters-RCV1 (Reuters Corpus Volume 1)

• Cranfield

• 20 Newsgroups (18,941 articles)

Which user group typically prefers high precision according to the priorities listed?

• Web users

• Law and intelligence professionals

• Medical researchers

• Personal searches

Which formula represents accuracy as presented?

• (TP + TN)/(TP + FP + FN + TN)

• TP/(TP + FP)

• TP/(TP + FN)

• 2TP/(2TP + FP + FN)

Alpha (α) in the F-measure is best described as which of the following?

• Relative importance of recall compared to precision

• Weight that decides the importance given to precision vs. recall

• Number of positive predictions

• Rate of false negatives

• Scaling factor for accuracy

Beta (β) in the F-measure is defined as what in relation to α?

• β = α

• β = 1 − α

• β² = (1 − α)/α

• β² = α/(1 − α)

• β = √α

Why is the F-measure considered a good single-number summary of performance?

• It uses the arithmetic mean to smooth values

• It gives a weighted harmonic mean and punishes large differences between P and R

• It maximizes accuracy directly

• It ignores extreme values

• It equally weights true negatives

If α increases while P and R are fixed, what is the conceptual effect on the weighting within the F-measure?

• More weight to recall compared to precision

• Equal weight to precision and recall

• More weight to precision compared to recall

• No change in weighting

• Weight shifts to accuracy

Which statement best explains why formal measures like Precision, Recall, and F-measure are considered indirect for evaluating user-centered Information Retrieval?

• They perfectly capture user happiness with search results.

• They only assess system performance, not actual user satisfaction.

• They are qualitative metrics based on user comments.

• They only apply to enterprise search systems.