README.md

mathmex-functions

This is a python package that allows you to easily work with our OpenSearch interface and our fine-tuned sentence transformer.

Installation

To install the package run this command.

pip install git+https://github.com/jgore077/mathmex-functions.git

opensearch

This our personal interface to the OpenSearch search engine. It provides many functions that can manipulate the OpenSearch instance and fulfill queries.

from MathmexFunctions import ElasticSearchInterFace 
dir(ElasticSearchInterFace)

It provides all these functions

['autocomplete', 'create_index', 'delete_index', 'format_results', 'get_all_indices', 'insert_data', 'insert_session', 'perform_autocomplete', 'perform_autocorrect', 'read_top_k_index', 'retrieve_entry', 'single_index_search', 'vector_search']

Instantiate a ElasticSearchInterFace object like this.

CONFIG_PATH="/path/to/config.ini"

# setting model=True will attach a `ArqMathSentenceTransformer` object to interface, this is required for vector querys
# if you use the config_path variable all the default parameters will be set to the contents of the config file

interface=ElasticSearchInterFace(host='localhost',port=9200,username='myOpenSearchUsername',password='myOpenSearchPassword',model=True,config_path=CONFIG_PATH)

Here is an example of what your config file should look like

[OpenSearch]
host = yourhost
port = yourport
username = yourusername
password = yourpassword

ArqMathSentenceTransformer

This is our fine-tuned sentence transformer which is fine-tuned for mathematical querys and designed to deal with LaTeX.

from MathmexFunctions import ArqMathSentenceTransformer

transformer=ArqMathSentenceTransformer()

# The only function provided by this class is encode
# It returns a vector which is the encoded string
transformer.encode("What is an affine transform")