Elastic Search Example

This project show how to use elastic search for a autocomplete select using relevance and full text search.

The elasticsearch server is served from docker with docker compose, to start execute:

docker-compose up -d

The example documents are in the sample-data directory, the models.json file include a sample list of models for brazilian car from (ComparaOnline)[comparaonline.com.br], to load this data in the elasticsearch server can be use the python script load_models.py so:

# change dir
cd data_loader
# install requirements
pip install -r requirements.txt
# run script
python load_models.py

Finally to test the results, is provided a express server, run with

# install dependencies
npm install
# run server
node index.js

open the sample on (http://localhost:3000/)

Configure search parameters

To configure the elasticsearch query edit the search.js file, the trade-off between full-text and relevance of every document (model) is controlled with the function_score parameter, the field_value_factor set how to get the score from the count_by_model of every document:

...
function_score: {
    query: {
    multi_match: {
        query: term,
        fields: ['brand', 'model'],
    },
    },
    field_value_factor: {
    field: 'count_by_model',
    modifier: 'log1p',
    factor: 2,
    },
}
...

this configuration set the following final score:

new_score = old_score * log(1 + factor * count_by_model)