Quickly build custom categorizers to optimize your NLP pipeline

Nearly every business has a specific set of categories, themes, or topics that fits their tasks, and no out-of-the-box taxonomy will satisfy. That’s why Rosette® offers a Classification Field Training Kit so that users can train their own categories, or engage the Basis Technology solutions team to do so. The Classification Field Training Kit builds upon the fundamental linguistic analysis of Rosette Base Linguistics, allowing users to extend categorization to process documents in 30+ languages.

There are two training methods:

1. Keyword-based training uses keywords that are representative of each category. It leverages Wikipedia pages that are representative of the keywords to use as the training set.
2. Machine learning from a training set involves handpicking documents that best represent each category. Rosette uses the training documents to machine learn the new taxonomy, so that future documents can be categorized with accuracy.

The first step in triaging data for any application

Categorization is a powerful tool for taming large datasets. Depending on the project, it might make sense to categorize documents before performing tasks like named entity recognition or topic extraction. For example:

• Categorize social media posts by topic to highlight trends and segment audience
• Identify positive or negative sentiment in a statement
• Filter out spam in email streams
• Tag webpages by content for targeted advertising and displaying related content
• Detect responsive documents for eDiscovery
• Identify the topic of customer service inquiries to properly prioritize or route questions.

Product highlights

• Field training kit to easily add categories or language support
• Pretrained on IAB Tech Lab Content Taxonomy for English
• Cloud or enterprise deployments
• Fast and scalable
• Industrial-strength support

Categories

Supported languages

* Languages must be supported by Rosette Base Linguistics.

Sample output:

{
  "categories": [
    {
      "label": "ARTS_AND_ENTERTAINMENT",
      "confidence": 0.06416648,
      "score": -0.01447566
    },
    {
      "label": "SPORTS",
      "confidence": 0.05782175,
      "score": -0.11859164
    },
    {
      "label": "TRAVEL",
      "confidence": 0.05627946,
      "score": -0.14562697
    },
    {
      "label": "TECHNOLOGY_AND_COMPUTING",
      "confidence": 0.05617463,
      "score": -0.14749148
    },
    {
      "label": "HEALTH_AND_FITNESS",
      "confidence": 0.05582167,
      "score": -0.15379449
    }
  ]
}