Week 3 Quizhelp Center
Warning:the hard deadline has passed. You can attempt it, but and you won't be. You are are welcome to try it as a learning exercise. In accordance with the Coursera Honor Code, I certify this answers here are I own work. Question 1 Assume you are using a Unigram language model to calculate the probabilities of phrases. Then, the probabilities of generating the phrases "study text mining" and "text mining study" are not equal, i.e., P ("Stu Dy text Mining ") ≠p (" Text Mining Study "). False True question 2 are given a vocabulary composed of only four words: "The", "Computer", "Science", and "Technolog" Y ". Below are the probabilities of three of these four words given by a unigram model.
Word probability
The
0.4
Computer
0.2
Science
0.3
What is the probability of generating the phrase ' the technology ' using this Unigram language model? 0.5 0.04 0.0024 0.1 question 3 You are given the query q= "online courses" and two documents:
D1 = "Online Courses search Engine"
D2 = "Online education is affordable"
Assume you are using the maximum likelihood estimator without smoothing to calculate the probabilities of words in docume NTS (i.e., estimated P (w| D) is the relative frequency of Word w in the document D). Based on the Unigram query likelihood model, which to the following choices is correct? P (q| D1) = 1/16 P (q| D2) = 0 P (q| D1) = 0 P (q| D2) = 1/4 P (q| D1) = 1/16 P (q| D2) = 1/4 P (q| D1) = 1/2 P (q| D2) = 1/2 Question 4 Assume the same scenario as in question 3, but using linear interpolation (jelinek-mercer) smoothing withλ=0.5. Furthermore, you are given the following probabilities of some of the words in the collection model:
Word
P (w| C
Online
1/4
Courses
1/4
Education
1/8
Based on the Unigram query likelihood model, which to the following choices is correct? P (q| D1) = 1/16 P (q| D2) = 1/16 P (q| D1) = 1/16 P (q| D2) = 1/32 P (q| D1) = 1/32 P (q| D2) = 1/32 P (q| D1) = 1/16 P (q| D2) = 0 Question 5 The BM25 has more free parameters to tune than the ranking function of the Dirichlet Prior. True False Question 6 assume you are using Dirichlet Prior smoothing to estimate the probabilities of words in a certain d Ocument. What happens to the smoothed probability of the word when the parameter μ is increased? It becomes closer to the probability of the word in the collection language model it becomes closer to the maximum Ood estimate of the the probability derived from the document it does not change it tends to 1 question 7 it are possible that Pseudo feedback decreases the precision and recall of a CErtain retrieval System. True False question 8 refer to the Rocchio feedback formula in the slides. If you want to eliminate the effect of non-relevant documents when doing feedback, which of the following Eters must is set to zero? Γ and βγαβquestion 9 let q be The original query vector, DR={P1,..., pn} be the set of Positive document vectors, and dn={n1,..., nm} be The set of negative document vectors. let q1 be the expanded query vector after applying Rocchio on dr and dn with par Ameter values α, β, And γ. let q2 be the expanded query vector after applying Rocchio on dr and Dn with the same Val UEs forα, β, but γ being set to zero. Which of the following is correct? Q2 can have greater or equal weights to q1 q1 has strictly greater weights than q2 Q2 has Strictl Y Greater Weights than q1 Q1 can have greater or equal weights to q2 question a Which of the following is not true about the Kl-divergence retrieval model? It cannot be computed as efficiently as the query likelihood model. It represents both queries and documents as language models. It supports relevance feedback.