Overview

Multi-domain NER system to recognize and classify any entity according to your custom classification.

We provide a high-performance, multi-domain named entities recognition (NER) system that cannot only return a list of text chunks and their classes but also the exact location of an entity with a predicted label and its score. It gives you more control and is especially helpful for high-precision projects.

We provide several types of models and algorithms to perform zero-shot named entity recognition. Depending on requirements such as performance, recall, precision or interpretability, you can select the following endpoints:

Endpoints

• Fast - high-performance but accurate probabilistic zero-shot named entity recognition model;
  • When to use: if it’s essential high-performance and you need to perform analysis of large massive of data;
  • Pros: high speed, scoring of outputs, returning location of an entity in a text;
  • Cons: not the best recall, can’t logically infer entity categories, can’t group entities;
• Deterministic - rule-based entities recognition and its classification according to user-defined labels;
  • When to use: you need high recall and deterministic outputs;
  • Pros: high recall, scoring of outputs, return location of an entity in a text;
  • Cons: bad with cases, where it’s needed to recognize sub-parts of phrases;
• Advanced - performs recognition and classification of entities according to the user description. It can logically infer an entity and its class. Moreover, you can group entities into a single cluster using this endpoint.
  • When to use: if it’s required, logically infer entities categories of group entities by some rule;
  • Pros: good accuracy, can logically infer entity categories, can group entities;
  • Cons: not the best speed, can’t provide locations of an entity, no scoring;

What you need:

• Put text;
• Set names of entities categories;
• Optionally put entities spans (‘Fast’ and ‘Deterministic’ endpoints);
• Optionally set rules for grouping entities (‘Advanced’ endpoints);

What you get:

• List of entities with the following data:
  • text of an entity;
  • start of an entity (for Fast and Deterministic endpoints);
  • end of an entity (for Fast and Deterministic endpoints);
  • label class of an entity;
  • probability score of prediction (for Fast and Deterministic endpoints);

Key features:

• Multi-domain data inputs;
• Zero-shot capabilities;
• Scoring of outputs;
• Defining the exact position of an entity in a text;
• Grouping of entities by some rule;
• Logical inference of entities categories;

Check our other APIs

• Comprehend-it - fast zero-shot text classification service to categorize texts at a scale;
• Text2Table - extract any table from the text just putting column names and text itself;
• Web2Meaning - comprehensive web scraper that extracts text from the selected web pages;

Support

1. Documentation: for in-depth information on our API, functionalities, and integration guidelines, please refer to our extensive documentation.
2. Discord: encounter an issue or have a question? Join our Discord channel for rapid assistance and engage with our community and support team.
3. Email: feel more comfortable communicating via email? Reach out to us at info@knowledgator.com, and we’ll be happy to help with any queries you might have.
4. Our website