Skip to content

Latest commit

 

History

History
25 lines (20 loc) · 1.43 KB

Named-entities.md

File metadata and controls

25 lines (20 loc) · 1.43 KB
Raw

Named Entities

I-Analyzer has the capacity to display named entities.

Prerequisites

In order to display a corpus enriched with named entities, install the Annotated Text plugin of Elasticsearch, following the instructions here.

Named entity fields

To determine whether named entities are available for a given corpus, the application checks if a given corpus contains fields ending with :ner.

If the main content field is called speech, the field containing named entity annotations should be called speech:ner. This field should have the following Elasticsearch mapping:

{
    'type': 'annotated_text'
}

Moreover, an enriched corpus should contain the following keyword fields:

  • person:ner-kw
  • location:ner-kw
  • organization:ner-kw
  • miscellaneous:ner-kw These can be used to search or filter (to be implemented).

Enriching a corpus with named entities

To enrich a corpus with named entities, we recommend using the TextMiNER library. This library will read from an existing index and a specified field name. The content of the field is analyzed with the BERT-based models for named entity recognition provided by flair. The library then adds named entities to the annotated_text field and the keyword fields, as outlined above.