Extracting Services from Free Text

A Hugging Face user asks for the best method to extract services from free text, citing a rule-based approach with synonym dictionaries and negation detection as a starting point, with NLP or zero-shot models as a potential second validation step for varied reports.

ulajci https://discuss.huggingface.co/u/ulajci 1 What is the best method to find out which services are reported in a text? Should I use a ML approach or Rule-based Approach? What are the most accurate approaches? Example Catalog: 14 home services window cleaning, roof inspection, pest control, etc. Input short report : Did a roof inspection , need gutter repair , window cleaning , tree pruning , pressure washing , sprinkler repair . Owner declined pest control , said no fence install , driveway sealing not needed . Output present : roof inspection, gutter repair, window cleaning, tree trimming, power washing, sprinkler repair Excluded: pest control, fence installation, driveway sealing, others not mentioned I agree that a rule-based approach can work well when you have a fixed catalog of services. In your example, the bigger challenge seems to be handling negations like “declined pest control”, “not needed”, or “no fence install”, along with different ways people describe the same service e.g., “tree pruning” vs. “tree trimming”, “pressure washing” vs. “power washing” . I’d probably start with rule-based matching using a synonym dictionary and negation detection. If the reports become more varied over time, adding an NLP or zero-shot model as a second validation step could improve recall without making the system overly complex.