06 Vector Space Model

Question 1

what is an N-Gram

Answer 1

a sequence of any consecutive chars
eg. john is quicker than marry
n =1&raquo_space; “john”, “is”, “quicker”, “than”, “marry”&raquo_space; doest not imply who is quicker
n =2&raquo_space; “john is”, “is quicker”, “quicker than”, “than marry”…

used to estimate probability of sequence of word

Question 2

what is TF-IDF

Answer 2

matrix of term weights

Question 3

what is an IR model

Answer 3

model for
- document representation
- query representation
- estimating the relevance given the query

Question 4

what is the vector space model

Answer 4

representation of documents and queries as vectors
relevance can be calculated by comparing the similarity of the vectors

documents that have similar vectors talk about the same thing

Question 5

what does it mean if the cosine similarity is negative

Answer 5

some algorithms to penalise non matching terms to a negative number

Question 6

what are the advantages for vector space model

Answer 6

• simple geometric interpretation
• easy to compute and measure
• easy to adapt to various weighting schemes
• provides ranked output

Question 7

what are the disadvantages of vector space model

Answer 7

• high dimensionality
• term independence assumption
• which similarity metric to use?
• no guidance on when to stop ranking