怎样在PHP项目中实现命名实体识别？

wen java案例 2026-07-22 1

本文目录导读：

怎样在PHP项目中实现命名实体识别？

调用外部NER API（推荐）
使用PHP扩展
使用斯坦福CoreNLP PHP包装器
基于词典的简单实现
使用深度学习框架（高级）
推荐方案
注意事项

在PHP项目中实现命名实体识别（NER）主要有以下几种方式，我按推荐程度从高到低介绍：

调用外部NER API（推荐）

使用 spaCy + Python 服务

<?php
class NERService {
    private $pythonScript = 'ner_service.py';
    public function extractEntities($text) {
        $escapedText = escapeshellarg($text);
        $command = "python3 {$this->pythonScript} {$escapedText}";
        $output = shell_exec($command);
        return json_decode($output, true);
    }
}

ner_service.py:

import spacy
import sys
import json
nlp = spacy.load("zh_core_web_sm")  # 中文模型
text = sys.argv[1]
doc = nlp(text)
entities = []
for ent in doc.ents:
    entities.append({
        'text': ent.text,
        'label': ent.label_,
        'start': ent.start_char,
        'end': ent.end_char
    })
print(json.dumps(entities, ensure_ascii=False))

使用腾讯云/阿里云等平台API

<?php
// 腾讯云NLP示例
require_once 'vendor/autoload.php';
use TencentCloud\Common\Credential;
use TencentCloud\Nlp\V20190408\NlpClient;
use TencentCloud\Nlp\V20190408\Models\LexicalAnalysisRequest;
class TencentNER {
    private $client;
    public function __construct($secretId, $secretKey) {
        $cred = new Credential($secretId, $secretKey);
        $this->client = new NlpClient($cred, "ap-guangzhou");
    }
    public function extract($text) {
        $req = new LexicalAnalysisRequest();
        $req->Text = $text;
        $req->Flag = 1; // 开启命名实体识别
        $resp = $this->client->LexicalAnalysis($req);
        return json_decode($resp->toJsonString(), true);
    }
}

使用PHP扩展

PHP-Spacy扩展

# 安装
composer require php-spacy/php-spacy

<?php
require_once 'vendor/autoload.php';
use PhpSpacy\Spacy;
$spacy = new Spacy();
// 加载中文模型
$nlp = $spacy->load('zh_core_web_sm');
// 实体识别
$text = "马云在杭州创办了阿里巴巴集团";
$doc = $nlp($text);
foreach ($doc->getEntities() as $entity) {
    echo "文本: {$entity->getText()}\n";
    echo "标签: {$entity->getLabel()}\n";
    echo "位置: {$entity->getStart()}-{$entity->getEnd()}\n\n";
}

使用斯坦福CoreNLP PHP包装器

<?php
// 安装: composer require jasonroman/php-stanford-corenlp-adapter
$annotator = new \StanfordCoreNLP\StanfordCoreNLP();
$annotator->setAnnotators("tokenize,ssplit,pos,lemma,ner");
$text = "Apple Inc. was founded by Steve Jobs in Cupertino, California.";
$result = $annotator->annotate($text);
$entities = [];
if (isset($result['sentences'])) {
    foreach ($result['sentences'] as $sentence) {
        foreach ($sentence['tokens'] as $token) {
            if ($token['ner'] != 'O') {  // O表示非实体
                $entities[] = [
                    'text' => $token['word'],
                    'type' => $token['ner'],
                    'start' => $token['characterOffsetBegin'],
                    'end' => $token['characterOffsetEnd']
                ];
            }
        }
    }
}

基于词典的简单实现

对于特定场景,可以使用简单词典匹配：

<?php
class SimpleNER {
    private $dictionaries = [];
    public function __construct() {
        // 加载实体词典
        $this->dictionaries = [
            'PERSON' => ['张三', '李四', '王五'],
            'ORG' => ['腾讯', '阿里巴巴', '百度'],
            'LOC' => ['北京', '上海', '深圳']
        ];
    }
    public function extract($text) {
        $entities = [];
        foreach ($this->dictionaries as $type => $terms) {
            foreach ($terms as $term) {
                $pos = mb_strpos($text, $term);
                if ($pos !== false) {
                    $entities[] = [
                        'text' => $term,
                        'label' => $type,
                        'start' => $pos,
                        'end' => $pos + mb_strlen($term)
                    ];
                }
            }
        }
        return $entities;
    }
}
// 使用
$ner = new SimpleNER();
$text = "阿里巴巴在杭州举办了技术大会";
$entities = $ner->extract($text);
print_r($entities);

使用深度学习框架（高级）

使用PHP调用TensorFlow模型：

<?php
// 安装: composer require dml/tensorflow
$model = new TensorFlow\SavedModel('/path/to/ner/model');
// 数据预处理
$tokens = tokenize($text);
$input = encode($tokens);
// 预测
$predictions = $model->predict(['input' => $input]);
// 解码结果
$entities = decode_predictions($predictions, $tokens);