Entity isn't provided in the answer text

[mhallxyz]

The word "Clive" should make its way into the answer to replace the {{firstName}} text.

To reproduce just call the nlpListen() function.

const { containerBootstrap } = require('@nlpjs/core');
const { Nlp } = require('@nlpjs/nlp');
const { Ner } = require('@nlpjs/ner');
const { LangEn } = require('@nlpjs/lang-en-min');

const nlpListen = async () => {
  const container = await containerBootstrap();
  container.use(Nlp);
  container.use(Ner);
  container.use(LangEn);
  const nlp = container.get('nlp');
  nlp.settings.autoSave = false;
  nlp.addLanguage('en');
  nlp.addDocument('en', 'call @firstName', 'call.callFirstName');
  nlp.addNerRuleOptionTexts('en', 'firstName', 'Clive');
  nlp.addAnswer('en', 'call.callFirstName', 'Calling {{firstName}}');

  await nlp.train();
  const response = await nlp.process('en', 'call Clive');
  console.log(response);
  return response;
};

export default nlpListen;

Output "Calling {{firstName}}"

Expected Output "Calling Clive"

[markgibaud]

Hi @jesus-seijas-sp I also am struggling to understand how named entities are now put into answers in v4+.

It seems so are some others like https://github.com/axa-group/nlp.js/issues/440 and https://github.com/axa-group/nlp.js/issues/442

In v3 it looks like Handlebars was doing the work, but Handlebars is not present in v4+.

It looks like Evaluator might be the think we are looking for, with some context management? https://github.com/axa-group/nlp.js/tree/master/packages/evaluator

The example works with numerals and doesn't show us how to make the default evaluator part of the pipeline, ie. how tokens like {{ firstName }} in predefined answers are substituted with named entities from context when using nlpManager.process()

Or is it compile we need to use from @nlpjs/evaluator??

Any help appreciated.

[jesus-seijas-sp]

Hi, I updated @nlpjs/nlp to include entities in the context, and also by alias...

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

async function main() {
    const manager = new NlpManager({ languages: ['en'], nlu: { useNoneFeature: false, log:false }});
     manager.addDocument(
        'en',
        'I saw %hero% together with %hero%, they were eating %food%',
        'saw_heroes_eating'
    );
    manager.addAnswer('en', 'saw_heroes_eating', 'You saw {{ hero_0 }} with {{ entities.hero.items[1].sourceText }} eating {{ food }} {{ 5*2 }}')
    manager.addNamedEntityText('hero', 'spiderman', ['en'], ['Spider-man']);
    manager.addNamedEntityText('hero', 'iron man', ['en'], ['iron man']);
    manager.addNamedEntityText('hero', 'thor', ['en'], ['Thor']);
    manager.addNamedEntityText('food', 'burger', ['en'], ['Burger']);
    manager.addNamedEntityText('food', 'pizza', ['en'], ['pizza']);
    manager.addNamedEntityText('food', 'pasta', ['en'], ['Pasta', 'spaghetti']);

    await manager.train();

    const result = await manager.process('en','I saw spiderman together with ironman, they were eating spaghetti');
    console.log(result.answer);
}

main();

This returns "You saw Spider-man with iron man eating spaghetti". You can see it working here: https://runkit.com/jesus-seijas-sp/5f60a0d59e05de001b1cd637

Things to take into account:

• It allows entities to be "list", as we can see in the example. When entities are "list" they receive aliases with entity name and the position of the entity, thats why "hero_0" returns the entity at position 0 on the array of entities "hero".
• In the "entities" property of the context you have access to the entities object, that's why "entities.hero.items[1].sourceText" return the sourceText of the second hero. This is usefull
• Handlebars is not longer used and instead we have the evaluator because that way it can be used in the browser and also because it allows to have javascript expressions inside the {{ }}, one example is the "{{ 5*2 }}" of the example being resolved to 10.

[jesus-seijas-sp]

Closing as is solved