Six Browser-Only Writing Tools: What Can Be Counted in a Scene, and What Cannot
Some decisions in a scene can be counted: does the name of an emotion appear, is there a simile, how long are the sentences. Others cannot: how close the narrator stands to the character, whether the first sentence of an opening is an event or backstory.
I built six free tools around that difference. What can be counted, the tool counts. What cannot, the writer marks, and the tool only displays it.
All six run in the browser. The text is never sent to a server or stored, there is no account, and no model is called. Each tool works with English and Turkish text.
Tools page: https://leventbulut.com/objective-projection-tools/ Full walkthrough with examples: https://leventbulut.com/how-to-use-objective-projection-writing-tools/
A two-sentence example
Paste these two lines into the Scene Checker:
Selin was terrified. The hallway stretched on as if it would never end.
The tool gives two flags: one emotion label (terrified) and one simile (as if).
Selin got the key into the lock on the third try. At the end of the hallway the bulb flickered.
The count is zero. That is all it says. It does not say the second version is better.
The six tools
| Tool | What it does | How it works |
|---|---|---|
| Scene Checker | Marks words that name an emotion directly, and similes | Automatic, dictionary-based |
| Text Analyzer | Sentence length distribution, dialogue share, repeated phrases, lexical diversity (moving 50-word windows, Covington and McFall 2010) | Automatic count |
| Iceberg Tool | Sets subtext and surface side by side; flags emotion names, similes and subtext words that reach the surface | Writer writes, tool scans |
| Micro Focus Tracker | Lists planted details and their payoffs (a Chekhov's gun workbook) | Writer enters, tool asks |
| Narrative Distance Tool | Draws the narrator's distance sentence by sentence (adapted from Gardner's psychic distance) | Writer labels 1–5 |
| Story Opening Tool | Shows in which sentence the first event of an opening comes | Writer labels |
Why two tools give no automatic score
In an earlier reliability study (arXiv:2609.13936), a rule-based detector was compared with an independent human rater on 100 scenes. For materialised metaphor, a feature that needs context, agreement was κ = 0.015.
For that reason the Narrative Distance and Story Opening tools do not estimate anything. The writer labels; the tool draws the labels. The study does not tell us whether the low agreement comes from the feature itself or from how the rule was defined.
What the tools cannot do
- They do not measure literary quality or give a score.
- A scene with no flags does not mean the emotion reaches the reader; only readers can tell you that.
- Naming an emotion is sometimes a deliberate choice: in dialogue, in irony, with a first-person unreliable narrator. A flag is not a verdict.
- The Scene Checker is dictionary-based and does not read context; an ambiguous word such as tense can be flagged wrongly.
- How closely any of these flags match human judgement has not yet been tested.
What would help
- False positives and misses from the Scene Checker, with the sentence that caused them (English or Turkish).
- Anyone willing to act as a second, independent annotator on the open dataset (https://huggingface.co/datasets/leventbulut/objective-projection). That is the one test this work still lacks.
Article by Levent Bulut, a former journalist, author and the founder of the Objective Projection method. The tools are part of the Objective Projection framework; the method's rules and dataset are openly published and open to scrutiny.
