Can't interpret the text information and ratings matrix imported to NN

Question

I have a Recommender system which uses a Collaborative bayesian approach using pSDAE for recommending scientific articles from the Citeulike Dataset

The text information (as input to pSDAE) is in the file mult.dat and the rating matrix (as input for the MF part) is in the file cf-train-1-users.dat and is loaded using the following code:

def get_mult():
    X = read_mult('mult.dat',8000).astype(np.float32)
    return X

def read_user(f_in='cf-train-1-users.dat',num_u=5551,num_v=16980):
    fp = open(f_in)
    R = np.mat(np.zeros((num_u,num_v)))
    for i,line in enumerate(fp):
        segs = line.strip().split(' ')[1:]
        for seg in segs:
            R[i,int(seg)] = 1
    return R

The raw data is in proper Excel format with citations as doc-id, title, citeulike-id, raw-title, raw-abstract.

The mult.dat file containing hte text information looks like:

63 1:2 1666:1 132:1 901:1 1537:2 8:1 9:1 912:1

The trainusers.dat file looks like:

10 1631 3591 10272 14851 4662 13172 12684 5324 3595 3404

Here is the link to th ipynb for the whole Recommender system:

https://github.com/js05212/MXNet-for-CDL/blob/master/collaborative-dl.ipynb

Answer 1

I am the author of the CDL paper.

For the mult.data file, in

63 1:2 1666:1 132:1 901:1 1537:2 8:1 9:1 912:1

63 is the number of words for this documents, 1:2 means word 1 appears twice in the document, 1666:1 means word 1666 appears once in the document, etc.

For the trainuser.dat, in

10 1631 3591 10272 14851 4662 13172 12684 5324 3595 3404

10 is the number of positive samples for this user, the rest is a list of 10 items that are related to (liked by) this user.

You can check the README file in the Datasets collection for more details on the datasets.