---
language: en
license: apache-2.0
tags:
- forestry
- lumber-industry
- regulatory-compliance
- text-classification
- distilbert
datasets:
- custom-forestry-corpus
metrics:
- accuracy
- f1
---

# ForestryIntelligence

A DistilBERT-based text classifier fine-tuned to categorize forestry and lumber-industry report excerpts
into operational risk categories.

## Model Description

- **Base model:** `distilbert-base-uncased`
- **Task:** Multi-class text classification
- **Labels:** general_operations, harvest_compliance, pest_disease, sustainability, wildfire_risk
- **Fine-tuned by:** mishreyagupta

## Intended Use

Triaging/classifying short excerpts from forestry inspection reports, compliance documents, and field
notes into risk categories for downstream routing (e.g., flagging wildfire-risk passages for review).

This model performs **classification only**. For open-ended analysis or summarization of forestry reports,
pair it with a retrieval-augmented generation (RAG) pipeline — see the companion notebook section.

## Training Data

Fine-tuned on a labeled forestry text corpus spanning wildfire risk, pest/disease, harvest compliance,
sustainability, and general operations categories.

## Limitations

- Trained on a small, partly synthetic dataset — validate on your organization's real documents before
  production use.
- English-language text only.
- Not designed for generation, summarization, or open-ended Q&A.

## Evaluation

See the training notebook for full metrics (accuracy, weighted F1) on the held-out validation split.