News Site Numerical Data Coding

I have a stream of numerical information being pulled automatically from several news sites, and I need that raw output cleaned, coded, and delivered in a tidy, analysis-ready format. The web-scraping side is already set up; what’s missing is a solid routine that parses each scrape, identifies the key figures buried in the articles, applies consistent codes to them, and drops everything into a structured file (CSV or JSON works for me).

You’ll receive:
• the current scraping script and a sample of the raw dump
• a field dictionary showing how each number should be labeled or categorised

I’m expecting your returned script (Python preferred—BeautifulSoup/Scrapy plus pandas is perfect, but use what you like) along with the final processed dataset and a brief read-me so I can rerun or extend the pipeline later. Accuracy of the coding and reproducibility of results will be the benchmarks for sign-off.