Arabic Normalization

The arabic_normalization filter is a built-in token filter for Arabic text. It normalizes Arabic-specific letter variants and removes optional marks that can make equivalent Arabic terms appear different during text analysis.

Configuration

Notes

For Arabic text, use the built-in Arabic analyzer in most cases. The built-in analyzer includes this filter together with standard tokenization, lowercasing, decimal digit normalization, Arabic stemming, and Arabic stop-word removal. Use arabic_normalization directly only when you need to build a custom analyzer pipeline.

To use the arabic_normalization filter in a custom analyzer, add it to the filter section in analyzer_params:

Python

python

analyzer_params = {
    "tokenizer": "standard",
    "filter": ["arabic_normalization"],
}

Java

java

import java.util.Collections;
import java.util.HashMap;
import java.util.Map;

Map<String, Object> analyzerParams = new HashMap<>();
analyzerParams.put("tokenizer", "standard");
analyzerParams.put("filter", Collections.singletonList("arabic_normalization"));

Go

go

analyzerParams := map[string]any{"tokenizer": "standard", "filter": []any{"arabic_normalization"}}

Rust

rust

let analyzer_params = serde_json::json!({
    "tokenizer": "standard",
    "filter": ["arabic_normalization"]
});

C++

c++

nlohmann::json analyzer_params = {
    {"tokenizer", "standard"},
    {"filter", {"arabic_normalization"}},
};

NodeJS

javascript

const analyzer_params = {
    "tokenizer": "standard",
    "filter": ["arabic_normalization"],
};

cURL

bash

# restful
analyzerParams='{
  "tokenizer": "standard",
  "filter": [
    "arabic_normalization"
  ]
}'

The arabic_normalization filter has no configurable parameters.

The filter applies the following transformations:

Transformation	From	To
Hamza + Alef variants	آ, أ, إ	ا
Teh Marbuta	ة	ه
Alef Maksura	ى	ي
Harakat	U+064B through U+065F	Removed
Tatweel / Kashida	ـ	Removed

The filter operates on tokens generated by the tokenizer. The configuration above is intentionally a custom analyzer example and does not include the complete Arabic processing pipeline.

Examples

Before applying the analyzer configuration to your collection schema, verify its behavior using the run_analyzer method.

Analyzer configuration

Python

python

analyzer_params = {
    "tokenizer": "standard",
    "filter": ["arabic_normalization"],
}

Java

java

import java.util.Collections;
import java.util.HashMap;
import java.util.Map;

Map<String, Object> analyzerParams = new HashMap<>();
analyzerParams.put("tokenizer", "standard");
analyzerParams.put("filter", Collections.singletonList("arabic_normalization"));

Go

go

analyzerParams = map[string]any{"tokenizer": "standard", "filter": []any{"arabic_normalization"}}

Rust

rust

let analyzer_params = serde_json::json!({
    "tokenizer": "standard",
    "filter": ["arabic_normalization"]
});

C++

c++

nlohmann::json analyzer_params = {
    {"tokenizer", "standard"},
    {"filter", {"arabic_normalization"}},
};

NodeJS

javascript

const analyzer_params = {
    "tokenizer": "standard",
    "filter": ["arabic_normalization"],
};

cURL

bash

# restful
analyzerParams='{
  "tokenizer": "standard",
  "filter": [
    "arabic_normalization"
  ]
}'

Verification using run_analyzer

Python

python

from pymilvus import MilvusClient

client = MilvusClient(uri="YOUR_CLUSTER_ENDPOINT")

sample_text = "آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي"

result = client.run_analyzer(sample_text, analyzer_params)
print(result)

Java

java

import io.milvus.v2.client.ConnectConfig;
import io.milvus.v2.client.MilvusClientV2;
import io.milvus.v2.service.vector.request.RunAnalyzerReq;
import io.milvus.v2.service.vector.response.RunAnalyzerResp;
import java.util.ArrayList;
import java.util.List;

ConnectConfig config = ConnectConfig.builder()
        .uri("YOUR_CLUSTER_ENDPOINT")
        .token("YOUR_CLUSTER_TOKEN")
        .build();
MilvusClientV2 client = new MilvusClientV2(config);

List<String> texts = new ArrayList<>();
texts.add("آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي");

RunAnalyzerResp resp = client.runAnalyzer(RunAnalyzerReq.builder()
        .texts(texts)
        .analyzerParams(analyzerParams)
        .build());
List<RunAnalyzerResp.AnalyzerResult> results = resp.getResults();

Go

go

import (
    "context"
    "encoding/json"
    "fmt"

    "github.com/milvus-io/milvus/client/v3/milvusclient"
)

client, err := milvusclient.New(ctx, &milvusclient.ClientConfig{
    Address: "YOUR_CLUSTER_ENDPOINT",
    APIKey:  "YOUR_CLUSTER_TOKEN",
})
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

bs, _ := json.Marshal(analyzerParams)
texts := []string{"آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي"}
option := milvusclient.NewRunAnalyzerOption(texts...).
    WithAnalyzerParamsStr(string(bs))

result, err := client.RunAnalyzer(ctx, option)
if err != nil {
    fmt.Println(err.Error())
    // handle error
}

Rust

rust

use milvus::v2::error::Result;
use milvus::v2::prelude::*;

#[tokio::main]
async fn main() -> Result<()> {
    let config = ConnectConfig::new().uri("YOUR_CLUSTER_ENDPOINT").token("YOUR_CLUSTER_TOKEN");
    let client = ClientV2::new(&config).await?;

    let texts = ["آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي".to_string()];
    let analyzer_params = serde_json::json!({
        "tokenizer": "standard",
        "filter": ["arabic_normalization"]
    });

    let result = client.run_analyzer(
        RunAnalyzerRequest::builder()
            .texts(texts)
            .analyzer_params(analyzer_params)
            .build()?,
    ).await?;

    println!("{:?}", result.results());
    Ok(())
}

C++

c++

#include "milvus/MilvusClientV2.h"
#include <iostream>
#include <string>

auto client = milvus::MilvusClientV2::Create();

milvus::ConnectParam connect_param{"YOUR_CLUSTER_ENDPOINT", "YOUR_CLUSTER_TOKEN"};
auto status = client->Connect(connect_param);
if (!status.IsOk()) {
    std::cout << status.Message() << std::endl;
}

std::string text = "آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي";
auto request = milvus::RunAnalyzerRequest()
                       .AddText(text)
                       .WithAnalyzerParams(analyzer_params);

milvus::RunAnalyzerResponse response;
status = client->RunAnalyzer(request, response);
if (!status.IsOk()) {
    std::cout << status.Message() << std::endl;
}

NodeJS

javascript

const res = await client.runAnalyzer({
    text: 'آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي',
    analyzer_params: analyzer_params
});

console.log(res);

cURL

bash

# restful
curl --request POST \
  --url "${CLUSTER_ENDPOINT}/v2/vectordb/common/run_analyzer" \
  --header "Authorization: Bearer ${TOKEN}" \
  --header "Content-Type: application/json" \
  --data '{
    "text": ["آدم أحمد إسلام مدرسة كبرى كِتَابٌ عـــربي"],
    "analyzerParams": "{\"tokenizer\":\"standard\",\"filter\":[\"arabic_normalization\"]}"
  }'

Expected output

plaintext

['ادم', 'احمد', 'اسلام', 'مدرسه', 'كبري', 'كتاب', 'عربي']