Writing Assessment Is Now Part of Speechace. Here’s Why We Built It.

We’ve Always Been About Spoken Language. So Why Writing?

 

When we started Speechace, the problem we were solving was specific: spoken language assessment was broken. Teachers and administrators were drowning in manual pronunciation scoring, and learners had no reliable way to get objective feedback on their speech. We built a comprehensive spoken language evaluation engine — one that could score pronunciation, fluency, prosody, and more at scale, in real time.

Then our customers started asking: “Can you do the same for writing?”

Honestly, for a while we pointed them elsewhere. Writing assessment wasn’t our thing. But the more we looked at what was actually available, the more we realized the problem was the same one we’d solved for speech: there was no equivalent — no purpose-built, rubric-aligned, automatically scored writing assessment that institutions and business could actually trust and deploy at scale.

So we built one.

What “Purpose-Built” Actually Means

The obvious question is: why not just use a general-purpose LLM to score writing? ChatGPT can give feedback on an essay, right?

Here’s the difference: General LLMs are generalists. They’re trained to be helpful across an enormous range of tasks, and they’ll produce plausible-sounding feedback on writing. But “plausible” isn’t the same as “calibrated,” and that gap matters enormously when you’re making hiring and placement decisions or tracking learner progress over time.

The Speechace writing model is different in a few concrete ways:

•       Specific trained models for each rubric component. We don’t have a single model that does everything. We have separate models trained to assess grammatical accuracy, vocabulary range and diversity, sentence structure, coherence, and task achievement — the components that actually appear in standard assessment rubrics like IELTS and PTE.

 

•       Independently evaluated for purpose. Our models aren’t just trained and shipped. They’re evaluated against human rater benchmarks specifically for the task of writing assessment. We measure alignment with expert scorers, not just general text quality.

•       Scores mapped to recognized frameworks. The output isn’t a proprietary black box. Scores are aligned to CEFR, IELTS, PTE, TOEFL, and TOEIC, so results are interpretable by anyone who works with language standards.

 

•       No prompt-sensitivity. With a generic LLM, slightly changing how you ask the question changes the score. Our scoring pipeline doesn’t work that way — same writing, same score, every time.

Two Ways to Use It

We’ve made writing assessment available through our Web SaaS or through our REST API.

1. Speechace Speaking Test: — our SaaS solution for building and deploying language assessments — now includes Writing Assessments. If you’re already building assessments in Speechace Workspaces, you can add writing assessments directly alongside speaking tasks. Documentation for writing question configuration is here.
2. Speechace API: If you’re a developer integrating assessment into your own platform, the Speechace API now includes a Score Writing endpoint. You send us the text, we send back structured scores across rubric dimensions. Full API reference is here.

What Gets Scored

A writing submission — whether it’s a long essay, a short task, or a chat response — gets analyzed across:

•       Grammatical range and accuracy

•       Lexical diversity and precision

•       Coherence and cohesion

•       Task Response (how well the response addresses the prompt)

 

Results come back in seconds, with component-level scores and an overall performance mapped to the major international scales, and optional detailed feedback with classified suggested improvements. No manual grading. No waiting.

Complete Language Assessment

Spoken and written language are both part of how people communicate. We’ve always believed that if you want to assess language ability properly, you need to look at both. Now you can.

If you’re using the Speechace Speaking Test, writing is available to add to your assessments now. If you’re building on our API, check out the Score Writing endpoint and reach out if you have questions about integration.

More soon.

— The Speechace Team