Relevance Feedback and Deep Neural Network-Based Semantic Method for Query Expansion

Algorithm 1

Skip-gram and relevance feedback-based query expansion RESKQ(Q[],D[]).

1. for each q in Q, do

  1.1. Retrieve top “k” documents d_k from initial retrieved documents.

2. for each d in d_k, do

  2.1. Retrieve user assisted relevant document d_u and store them in D_u.

3. for each t in q, do

  3.1. Create a hot vector x form dataset D_u.

4. iter←epoch

5. Initialize weight matrix w and w’ with random weights.

6. for i←1 to itr, do

  6.1. Compute h←w^T.x

  6.2. Compute v←.h

  6.3. Update ←- e^’

  6.4. Update w_new←w - e

7. mer←[]

8. for each t in q, do

  8.1. Retrieve top 15 context words w_cons for w_t such that y is maximum.

  8.2. mer←mer+ w_cons

9. qexp←q+mer

10. return qexp