API NER with url

[Thomas Pfuhl]

Hello,

when onvoking the API with a URL which points to a plain text file, I get an empty output. The API key is valid, though.

Using the text parameter instead of the url parameter, I get an output.

What's wrong? Thanks for hints!

th

[Giacomo Berardi]

Hi Thomas,

can you please show us an example of API call you are making?

Giacomo Berardi 

Dandelion team

[Thomas Pfuhl]

Buon giorno Giacomo,

buon giorno Giacomo,

molto grazie per la sua aiuta !

This is the URL pointing to plain text: http://www.proportiodivina.eu/flucht/src/Flucht-und-Vertreibung-der-Holder-Eggers-aus-Stettin.txt

I presume you do not need my API key (token).

invoking :  https://api.dandelion.eu/datatxt/nex/v1/?lang=de&url=http://www.proportiodivina.eu/flucht/src/Flucht-und-Vertreibung-der-Holder-Eggers-aus-Stettin.txt&token=...

gives a jason with empty results.

Even shorter text (one sentence) submitted by this URL gives the same empty result.

what's wrong?

greetings fom Berlin,

Thomas

[Giacomo Berardi]

Hi Thomas,

the problem is about the `url` parameter, which has a specific use for web pages, in fact, from the documentation:

use "url" when you have an URL and you want the Entity Extraction API to work on its main content; it will fetch the URL for you, and use an AI algorithm to extract the relevant part of the document to work on; in this case, the main content will also be returned by the API to allow you to properly use the annotation offsets;

The content extraction algorithm in this case does not find the html, so it immediately drops the text file content

I strongly suggest you to retrieve the text content first, and then use the `text` parameter of the Dandelion API

Cheers

Giacomo Berardi

Dandelion team

[Thomas Pfuhl]

Buon giorno Giuacomo,

as you outlined, the NER works fine when embedding the text in a html file and involing that Url.

Thank you for your precious hint! Thomas