mahmoudd777/qwen35-realestate-2048

Qwen3.5-4B — Real Estate Call Analysis (Egyptian Market)

Fine-tuned version of Qwen3.5-4B for extracting structured information from real estate call transcripts in the Egyptian market. Supports English and Egyptian Arabic.

What it does

Given a call transcript, the model extracts a structured JSON object with:

• —Client name, sentiment, urgency, timeline
• —Confidence score and transcript quality score
• —Client profile, special requests, action items, call summary
• —requested_units array with: intent, property type, location, currency, budget, payment method, area, bedrooms, finishing, key objection

Training

• —Base model: Qwen/Qwen3.5-4B
• —Method: LoRA fine-tuning (r=16) via Unsloth
• —Dataset: 650 real estate call transcripts (70% English, 30% Egyptian Arabic)
• —Sequence length: 2048 tokens
• —Epochs: 6

Language support

• —English call transcripts
• —Egyptian Arabic call transcripts
• —Mixed Arabic/English conversations