You can lookup in the infoboxes using a simple key:value syntax. For example, “author:Shelley book” will give you Frankenstein among others:<\/p>\n\n\n\n Next, I tried looking at quality classes<\/strong>. Wikipedia pages are judged on various criteria<\/a> and given scores like “Featured Article”, “Good Article”, “A-class”, “B-class” and so on. I had some problem pulling this information from the wikipedia dumps. Most of these annotations are in talk pages which require downloading the complete history of wikipedia revisions (no thanks). If you look on the articles themselves, you only see “Featured Articles” (6885), “Good Articles” (43598), “Stubs” (2318123) and a few others. Only “Stubs” has really high coverage and this data didn’t seem like a good test of the bit array filters I wanted to build.<\/p>\n\n\n\n Finally, I looked at categories<\/strong>. If you go to the bottom of almost any wikipedia page, you’ll find a bunch of categories that it falls into. These categories are essentially collections of pages on a similar topic or having something else in common. For example, Frankenstein is in:<\/p>\n\n\n\n Now, these categories appear as text in the article, so you can already search for them but what you couldn’t do restrict your results to the category. For example, if you search for “Romanic novels”, you’ll get a result for “Romantic fantasy”. This is page has both those words, but is not in the category.<\/p>\n\n\n\n So, I decided to add a category filter.<\/p>\n\n\n\n With a category filter, I wanted exact matches on the category name. You should be able to select a category, see the results in it and filter further using a plain search. The user experience for this is a bit different from our plain searches. I don’t know in advance what categories there are and I need to make an exact match. So, the first step was to index the categories with their full text phrase (lowercase) as a key. In our search index we’ll have “romanic novels” as a key with all the documents in that category as values.<\/p>\n\n\n\n That enforces that we require exact matches. In the user interface, I added an input fields specifically for categories, because it needs to suggest that partial matches aren’t allowed. <\/p>\n\n\n\n Next, we need to fix the discovery problem. How does the user know what categories can be searched? There are 1,836,240 different categories which is too many for a drop-down list. Instead, we need to implement suggestions.<\/p>\n\n\n\n When typing into a search box, you are often given options to complete what you might be typing. Often, your text is seen a prefix and the suggestions merely complete the line of text in some way. Other times, when you’re looking for something exact but you don’t know what it is, your text might appear in the middle of the result. That said, typically we don’t start typing in the middle of a word, so you can treat whatever someone types as a prefix of a word in the text they’re looking for. So maybe typing “sci fi” should match “Science Fiction”. Or even “fi sci” should work.<\/p>\n\n\n\n![\"\"](\"https:\/\/byte64.com\/wp-content\/uploads\/2026\/04\/image-14-1024x632.png\")
<\/figure>\n\n\n\n\n
![\"\"](\"https:\/\/byte64.com\/wp-content\/uploads\/2026\/04\/image-15-1024x480.png\")
<\/figure>\n\n\n\nCategory Filter<\/h2>\n\n\n\n
![\"\"](\"https:\/\/byte64.com\/wp-content\/uploads\/2026\/04\/image-16-1024x481.png\")
<\/figure>\n\n\n\n![\"\"](\"https:\/\/byte64.com\/wp-content\/uploads\/2026\/04\/image-17-1024x633.png\")
<\/figure>\n\n\n\n![\"\"](\"https:\/\/byte64.com\/wp-content\/uploads\/2026\/04\/image-18-1024x480.png\")
<\/figure>\n\n\n\n