Indexes should only contain domain relevant information. This requires some document processing.

• Remove formatting: we only care about the text, not how it's presented
• Remove stopwords: words like "a", "and", and "the" don't contribute much because they're very common
• Stem words: combine morphological and inflectional varients to one entry. "Swim", "Swimming", and "Swimmer" are all indexed under "Swim".
• Match synonyms: combine synonyms into one entry. "Car", "Auto", and "Automobile" all map to "Car"