节点nlp如何提取电子邮件、电话、url?

我将向您提供一个示例代码:

const { NlpManager } = require('node-nlp');

const manager = new NlpManager({ languages: ['en'] });

async function mainExtractEntities() {
  const result = await manager.extractEntities('en', 'Are you able to identify that meh@meh.com is an email and moh@moh.com is another email so there are 2 emails?');
  console.log(result);
}

async function mainFullExample() {
  manager.addDocument('en', 'My mail is %email%', 'email');
  manager.addDocument('en', 'My email is %email%', 'email');
  manager.addDocument('en', 'Here you have my email: %email%', 'email');
  manager.addDocument('en', 'Hello', 'greet');
  manager.addDocument('en', 'Good morning', 'greet');
  manager.addDocument('en', 'good afternoon', 'greet');
  manager.addDocument('en', 'good evening', 'greet');
  manager.addAnswer('en', 'email', 'Your email is {{email}}');
  manager.addAnswer('en', 'greet', 'Hi!');
  await manager.train();
  let result = await manager.process('en', 'I think that my mail is meh@meh.com');
  console.log(result);
  result = await manager.process('en', 'Hello bot!');
  console.log(result);
}

mainExtractEntities();
mainFullExample();

这将显示在控制台中:

[ { start: 30,
    end: 40,
    len: 11,
    accuracy: 0.95,
    sourceText: 'meh@meh.com',
    utteranceText: 'meh@meh.com',
    entity: 'email',
    resolution: { value: 'meh@meh.com' } },
  { start: 58,
    end: 68,
    len: 11,
    accuracy: 0.95,
    sourceText: 'moh@moh.com',
    utteranceText: 'moh@moh.com',
    entity: 'email',
    resolution: { value: 'moh@moh.com' } },
  { start: 100,
    end: 100,
    len: 1,
    accuracy: 0.95,
    sourceText: '2',
    utteranceText: '2',
    entity: 'number',
    resolution: { strValue: '2', value: 2, subtype: 'integer' } } ]
{ locale: 'en',
  localeIso2: 'en',
  language: 'English',
  utterance: 'I think that my mail is meh@meh.com',
  classification:
   [ { label: 'email', value: 0.9994852170204532 },
     { label: 'greet', value: 0.0005147829795467752 } ],
  intent: 'email',
  domain: 'default',
  score: 0.9994852170204532,
  entities:
   [ { start: 24,
       end: 34,
       len: 11,
       accuracy: 0.95,
       sourceText: 'meh@meh.com',
       utteranceText: 'meh@meh.com',
       entity: 'email',
       resolution: [Object] } ],
  sentiment:
   { score: 0.25,
     comparative: 0.027777777777777776,
     vote: 'positive',
     numWords: 9,
     numHits: 1,
     type: 'senticon',
     language: 'en' },
  srcAnswer: 'Your email is {{email}}',
  answer: 'Your email is meh@meh.com' }
{ locale: 'en',
  localeIso2: 'en',
  language: 'English',
  utterance: 'Hello bot!',
  classification:
   [ { label: 'greet', value: 0.8826839762075465 },
     { label: 'email', value: 0.1173160237924536 } ],
  intent: 'greet',
  domain: 'default',
  score: 0.8826839762075465,
  entities: [],
  sentiment:
   { score: 0,
     comparative: 0,
     vote: 'neutral',
     numWords: 2,
     numHits: 0,
     type: 'senticon',
     language: 'en' },
  srcAnswer: 'Hi!',
  answer: 'Hi!' }

重要事项:

• 您可以省略ExtractEntities和Process中的语言,而传递未定义的语言,这样就可以从您的句子中猜测语言以适合您的nlpmanger的最佳语言。

• 电子邮件提取适用于任何语言。您有其他更复杂的实体,如文本编号,这些实体将只针对某些语言提取。

• 实体提取只是一个部分,其他有趣的部分是NLU分类器和自然语言生成,您将看到答案“您的电子邮件是电子邮件”是一个模板,并将电子邮件替换为从会话中提取的电子邮件。