Suppose you need to find customer reviews about cold food or soggy fries. A dictionary-based method searches text using a prepared lexicon of relevant terms rather than relying only on keywords you invent for the task. The dictionary is the starting point, not a guarantee that every relevant review will be found. section 5.6
Choose the lexicon before trusting the results:
- Check whether it covers the subject and includes enough relevant terms and detail. section 5.6.2
- Check its accuracy, publication date, and fit for current language. section 5.6.2
- Look for cultural, linguistic, or domain bias. Compare candidate lexicons with pilot tests rather than taking the first search result. section 5.6.2
A lexicon can hold more than a list of words: entries may carry usage tags, regional meanings, or emotion intensity. Such detail can support a richer reading of sentiment and context than simple keyword detection, but its usefulness still depends on the dictionary’s quality. section 5.6.1
When review wording does not match the dictionary
A Portuguese sentiment search showed low sensitivity for positive reviews. The lexicon contained lemmas, but the reviews had not been converted to lemmas before matching. In an iterative AI-assisted Python workflow, the follow-up request was to adapt preprocessing; the proposed change used spaCy’s Portuguese model to extract lemmas from review tokens. Inspect matches against the original reviews after making such a change: matching word forms fixes one possible cause of missed reviews, not every cause. section 5.6.5
