Introducing New NLP Platform Updates: Smarter Search, Validation, and Reporting
We're excited to roll out a series of updates to our NLP platform, designed to streamline your annotation process and improve data accuracy. Whether you’re searching for specific annotations, validating responses, or generating dynamic questions, these updates provide greater flexibility and efficiency.
From advanced search capabilities to automated validation scripts, improved reporting filters, and Markdown support for question hints, each feature is built to make managing large datasets smoother and more intuitive.
Read on to explore how these updates can optimize your labeling workflow!
Advanced Search
Finding specific annotations in Span Labeling is now more powerful. Previously, search was limited to text and labels. Advanced Search expands this by including metadata and supporting multiple conditions, allowing for precise filtering. You can now combine text, labels, and metadata to refine searches and configure queries with specific parameters, giving full control over search logic. This ensures you can quickly retrieve exactly what you need.
Answer Validation Script
Ensuring accurate annotations is crucial for high-quality training data. The Answer Validation Script automates error detection in Row Labeling by enforcing custom validation rules. It ensures answers follow specific formats, detects inconsistencies, and flags invalid entries before submission. This reduces back-and-forth corrections and improves the overall efficiency of the labeling workflow.
Script-Generated Questions
Previously, all rows had the same predefined questions. Now, Script-Generated Questions allow dynamic question generation based on data attributes. This brings greater flexibility, ensuring relevant questions for each row. It’s especially useful for evolving datasets with varying structures.
Additional Filters for Custom Report Builder
You can now generate reports that are filtered by project, tags, members, and labeling types. Combined with the existing date filter, you can configure more tailored and accurate reports based on your needs, eliminating the need for manual work.
Question Hints Now Support Markdown
Question Hints now support Markdown and are available for all question types. You can format instructions using bold, bullet points, and underline to structure content, among other options. This makes it easier to highlight key details, improve readability, and reference external resources for clearer guidance.
These updates make it easier to manage large datasets, maintain data quality, and guide annotators effectively. Try them out and let us know what you think!
If you would like a free consultation to discuss your NLP project, schedule time with our Customer Success team: we’re here to help!
