Detect diagrams draft

[lhaibach]

This draft builds on #46 and experiments with extending the diagram detection logic in src/identifiers/diagram.py.

The identify_diagram function uses a voting system based on keywords, units, and axis progression checks, where the axis detection is includes a new axis_checks function, which checks for both monotonicity and numeric progression in clusters of entries.

However it introduces more code without improving the F1 scores compared to the latest version in the other branch. We probably do not want to merge this? But the arithmetic checks might come in handy as additional features for the treebased model training?

Introduces an axis_checks helper that evaluates clusters for both monotonicity and numeric progression. It updates identify_diagram to use a voting system across:

• diagram keywords
• units
• y-axis and x-axis monotonicity
• y-axis and x-axis numeric progression (new)

However, F1 scores are not improved compared to the latest implementation in the other branch and the code increases complexity. We likely do not want to merge this, however the progression logic might still be valuable as features for tree-based models.

For diagram

Branch	F1 Score	Precision	Recall
detect-diagram	62.22	60.87	63.64
detect-diagram-draft	60.47	61.90	59.09

Metric	text	boreprofile	map	geo_profile	title_page	diagram	table	unknown	Macro Avg
F1 Score	65.00	77.77	73.07	0.00	48.78	62.22	0.00	26.47	44.16
F1 Score draft	65.00	77.77	73.07	0.00	48.78	60.47	0.00	25.71	43.85

[Copilot]

Pull Request Overview

This PR experiments with extending diagram detection logic by implementing a voting system that incorporates new axis progression checks alongside existing keyword and unit detection. The changes aim to improve diagram identification by analyzing both monotonicity and numeric progression patterns in clustered data points.

Key changes:

• Introduces a voting system combining keyword detection, unit detection, and axis analysis
• Adds new axis_checks function to evaluate monotonicity and numeric progression in data clusters
• Implements arithmetic and logarithmic progression detection for axis validation

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}