Early Release

This evaluator reflects early-stage work. We’re continuously improving its accuracy and reliability.
Take a look at the evaluator files.

​
What it is

This evaluator was developed specifically for AI-generated informational texts aimed at Grades 3-4. The evaluation looks at a text and returns a complexity score of:
  • Slightly complex
  • Moderately complex
  • Very complex
  • Exceedingly complex
You can use the complexity score to help decide if the text meets your needs and expectations. The evaluator automates the time-consuming analysis of sentence structure as part of defining text complexity. It can be used to help ensure AI-generated texts are sufficiently complex for the grade level, and that sentence structure complexity progresses appropriately across the year and grades.

​
Why it matters

Sentence structure is one of the most important—and overlooked—drivers of text complexity. Long, multi-clause sentences, embedded phrases, and varied syntax place unique demands on readers that simple vocabulary or readability scores cannot capture. For edtech companies, a sentence structure evaluator provides critical insight they can’t get from existing tools: whether the texts they generate or curate reflect the kinds of syntax students need to practice to grow as readers. For educators and students, this matters because grappling with increasingly complex sentence structures is essential for building comprehension skills and preparing students to access grade-level texts across disciplines.

​
How it works

This evaluator works by breaking the evaluation of sentence structure into 2 distinct stages:
  1. Sentence analysis: Identifying sentence features in the text, including percentage composition of sentence types (e.g., simple, compound, complex, compound-complex), average words per sentence, ratio of subordinate and multiple subordinate clauses, number of concepts per sentence, and so forth.
  2. Final complexity rating: Assigning a complexity rating to the passage.
It uses an LLM and statistical methods to boost accuracy and cut down on result variability.