LAC Analysis for Elasticsearch and OpenSearch

The LAC Analysis plugin integrates Lucene LAC analyzer. It supports major versions of Elasticsearch and OpenSearch.

The plugin comprises analyzer: lac, and tokenizer: lac.

🛠️ How to Install

Install LAC Service by PaddleHub**

See PaddleHub

Create python environment

conda create -n paddlehub python=3.8

conda activate paddlehub

Install PaddleNLP

pip install paddlepaddle paddlehub

Start LAC Serving

hub serving start --modules lac --port 8866 --use_multiprocess --workers 8

Test

curl -X 'POST' 'http://127.0.0.1:8866/predict/lac' \
-H 'accept: */*' \
-H 'Content-Type: application/json' \
-d '{
    "text": [
        "今天是个好日子", "天气预报说：今天要下雨"
    ]
}'

Install Analysis Plugin

For Elasticsearch:

bin/elasticsearch-plugin install https://github.com/thundax-lyp/analysis-lac/releases/download/8.12.2/elasticsearch-analysis-lac-8.12.2.jar

For OpenSearch:

bin/opensearch-plugin install https://github.com/thundax-lyp/analysis-lac/releases/download/8.12.2/elasticsearch-analysis-lac-8.12.2.jar

⚠️ Tip

Make sure to replace the version number with the one that matches your Elasticsearch or OpenSearch version.

Getting Started

Step 1. create an index

curl -XPUT http://localhost:9200/index

Step 2. create a mapping

curl -XPOST http://localhost:9200/index/_mapping -H 'Content-Type:application/json' -d'
{
    "properties": {
        "content": {
            "type": "text",
            "analyzer": "lac",
            "search_analyzer": "lac"
        }
    }

}'

Step 3. index some docs

curl -XPOST http://localhost:9200/index/_create/1 \
-H 'Content-Type:application/json' \
-d'{
    "content":"站一个制高点看上海,上海的弄堂是壮观的景象。"
}'

Configuration

Config file LacAnalyzer.cfg.xml can be located at {conf}/analysis-lac/config/LacAnalyzer.cfg.xml or {plugins}/elasticsearch-analysis-lac-*/config/LacAnalyzer.cfg.xml

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE properties SYSTEM "http://java.sun.com/dtd/properties.dtd">
<properties>
    <entry key="service_url">http://127.0.0.1:8866/predict/lac/</entry>
</properties>

Thanks

⚠️ We borrow lots of codes from IK Analysis ⚠️