TextReveal® API workers
What are TextReveal® API workers?
With the API Analyze endpoints,
you can launch custom NLP pipelines and retrieve timeseries of indicators thanks to TextReveal® API workers
In order to run a custom analysis, you should define the list of workers in the request payload
and send it to the /dataset.
When using the /tql endpoint workers are automatically selected based on fields.
For instance, using the similarity_threshold field will automatically select the entity-similarity worker.
The order of the list does not matter since the workflow will run a rule based on workers’ categories: filter -> text filter -> annotator -> match-annotator
The default list can be left unchanged as it handles all of TextReveal® use cases. Removing one or more of the workers will remove the corresponding data filtering or enrichment.
Some workers are restricted to certain languages, see the Language Support page for more informations
List of workers
| Workers | Category | Description |
|---|---|---|
quality-score | filter | Filters documents. It uses a threshold on which it decides whether or not to let a document through. Ranges from 0 to 100. |
ner-linking | text filter | Filters documents relevant to entities. Documents not mentioning relevant entities are filtered out from the pipeline. This module also returns a list of matched entities. |
embedder-indicators | annotator | Generates sentiment These indicators fall into 3 classes:
|
raw-matcher | annotator | Annotates the mentions of a list of keywords within a document. Detected mentions are stored inside a “matches” key, under the class “mention”. |
concept | annotator | Detects occurrence of concepts/risk within a document using predefined concepts/risk keywords. Detected concepts are stored inside a “matches” key, under the class “concept”. |
entity-similarity | match annotator | Annotates sentences containing a mention or an entity with a similarity score. This score indicates the level of similarity between the text of the sentence and the description of the entity of interest. Similarity score ranges from 0 to 1 |