Convert JSON for TransRec

A script json2txt.py converts a JSON file into a text file which can be used as an input for TransRec.

Usage

For Amazon datasets,

python3 ./json2txt.py reviews_Office_Products_5.json.gz | gzip > reviews_Office_Products_5.txt.gz

For other datasets, use -d option.

python3 ./json2txt.py -d googlelocal reviews.clean.json.gz | gzip > reviews.clean.txt

Datasets for TransRec

Paper: http://cseweb.ucsd.edu/~jmcauley/pdfs/recsys17.pdf

Compilation

Code: https://drive.google.com/file/d/0B9Ck8jw-TZUEVmdROWZKTy1fcEE/view?usp=sharing

`sequential_rec`

Edit src/main.cpp to use go_TransRec.
make clean
make

You can use the result executable train as the following:

./train reviews_Automotive.txt.gz 5 5 10 0.1 0.1 0.01 0 10000 my_model_path_blah_blah

Amazon

Data: http://jmcauley.ucsd.edu/data/amazon/

We can use json2txt.py to convert Amazon's 5-core JSON datasets to txt.

Google Local

Data: http://jmcauley.ucsd.edu/data/googlelocal/googlelocal.tar.gz

Actually, this "JSON" file is not a JSON. So I wrote a converter for dealing with this problem. You can convert reviews.clean.json by the following command.

go run ./toJSON.go < reviews.clean.json > reviews.clean.real.json

If you think this conversion is slow, you can compile the Go source by using go build ./toJSON.go and run ./toJSON.

We can use json2txt.py to convert JSON to txt.

Epinions

Data: http://jmcauley.ucsd.edu/data/epinions/

Foursquare

Data: https://archive.org/details/201309_foursquare_dataset_umn

Flixter

Data: ~~http://www.cs.ubc.ca/~jamalim/datasets/~~ --> http://socialcomputing.asu.edu/datasets/Flixster