Stackexchange Siamese Posts Search

LaTeX documentation documentation.pdf

How to run:

• run on meta-center cloud
  • follow tutorial here
  • dont forget to setup MTU on docker

Look at releases

• run docker-compose up

• wait until logstash is finished...

• like this picture above

Create your index with data from archive.org

• there are scripts for this too!
• pip install -r ./data/requirements.txt
• data/main_posts_faiss_indexer.py

usage: main_posts_faiss_indexer.py [-h] -i INPUT_DIR_PATH [-b BATCH_SIZE] [-s STOP_AT] [--offset OFFSET] [-o OUTPUT_DIR_PATH]

What is done?

• extending web application by:
  • MQDD - Multimodal Question Duplicity Detection which searches for similar questions among the first results of a user query. This approach showed that the inference of the MQDD model is not fast enough to run 50-500 times in real time on ordinary CPU.

• creating scripts for file-based index for similarity search Faiss library and its support in app

• docker support of the web app

and more...

Details you can see in documentation.pdf