Chapter 13 Word Associating Mining

Question 1

paradigmatic relation

Answer 1

Word wa and wb have a paradigmatic relation if they can be substituted for each other.

Question 2

syntagmatic relation

Answer 2

the two words that have this relation can be combined with each other in a grammatical sentence - meaning that these two words are semantically related

Question 3

In general, there are two type of word relations

Answer 3

one is called paradigmatic

the other is called syntagmatic

Question 4

Examining context is a general way of discovering paradigmatic words.

Answer 4

similar left context
similar right context
similar general context

How similar are context (“cat”) and context (“dog”)?
How similar are context (“cat”) and context (“computer”)?

Question 5

we ask the question, what words tend to occur to the left of eats? What words tend to occur to the right of eats?

Answer 5

This is the intuition we would like to capture. In other words, if we see eats occur in the sentence, that should increase the chance that meat would also occur.
This is syntagmatic relation.

How helpful is the occurrence of “eats” for predicting occurrence of “meat”?
How helpful is the occurrence of “eats” for predicting occurrence of “text”?

Question 6

how to discover paradigmatic relations

Answer 6

the general idea of discovering paradigmatic relations is to compute the similarity of context of two words. For

By viewing context in the vector space model, we convert the problem of paradigmatic relation discovery into the problem of computing the vectors and their similarity.

Question 7

EOW (expected overlap of words)

Answer 7

Fig 13.5

Question 8

BM25 Weighting

Answer 8

In order to achieve this desired weighting, we will use BM25 weighting, which is of course based on the BM25 retrieval function. It is able to solve the above two problems by sublinearly transforming the count of wi in d1 and including the IDF weighting heuristic in the similarity measure.