This is a C++ implementaion of "Topkapi" algorithm from the following work:

Ankush Mandal, Cary Jiang, Anshumali Shrivastava, and Vivek Sarkar. Topkapi: Parallel and Fast Sketches for Finding Top-K Frequent Elements. In Neural Information Processing Systems(NIPS), Montreal, Canada, 2018.

This implementation finds Top-K frequent words from text data. It supports multi-threaded execution using OpenMP and distributed computation using MPI.

• GNU Compiler Collection (GCC)
• OpenMPI

1. Download the code from Github: git clone https://github.com/ankushmandal/topkapi.git
2. Go to src directory and create the executable using make command
3. For details on usage, type ./topkapi -h or ./topkapi --help
4. A sample slurm script for running the program on cluster is given in sample_slurm_script directory under src

Instructions on preprocessing any text data is given in utils directory. The experiments on the paper were carried out using two data sets:

• Gutenberg dataset from Project Gutenberg. Useful instructions on downloading the data set can be found at Nico's Blog.
• Puma datasets under "Wikipedia" section from here.