Requirements
Tensorflow 1.2
Python 2.7
Numpy
Scipy
Data Preparation
To run PARL, 7 files are required:
1.Training Rating records: file_name=TrainInteraction.out
each training sample is a sequence as:
UserId\tItemId\tRating\tDate
Example: 0\t3\t5.0\t1393545600
2.Validate Rating records: file_name=ValidateInteraction.out
The format is the same as the training data format.
3.Testing Rating records: file_name=TestInteraction.out
The format is the same as the training data format.
4.Word2Id diction: file_name=WordDict.out
Each line follows the format as:
Word\tWord_Id
Example: love\t0
5.User Review Document: file_name=UserReviews.out
each line is the format as:
UserId\tWord1 Word2 Word3 …
Example:0\tI love to eat hamburger …
6.Item Review Document: file_name=ItemReviews.out
The format is the same as the user review doc format.
7.User Auxiliary Review Document: file_name=UserAuxiliaryReviews.out
The format is the same as the user review doc format.
Note that: all files need to be located in the same directory
Configurations
num_factor: the latent dimension of the representation learned from the review documents;
num_filters: the number of filters of CNN network;
cnn_windows_size: the length of the sliding window of CNN;
learn_rate: learning rate;
batch_size: batch size;
epochs: number of training epoch;
max_len: the maximum doc length;
gama: the coefficient to control the constraint between tu & tax;
drop_rate: the keep probability of the drop_out strategy;