User @faujidoc1 has made a Google search for SuttaCentral.
I’ve had it in the back of my mind to add a google search box to SC, just wondering what it takes. Obviously the recent outages of search make this more important.
For such a long time we have gotten used to the idea that search is something simple, a solved problem, and we forget that Google worked out a massive set of technical issues and by doing so became one of the world’s biggest companies. Nothing about search is easy; as Chade Meng-Tan, one of the early Google engineers and an SC sponsor, told me, “search is more dark arts than science”.
Anyway, we have a long term commitment to improve the search on SC, but possibly we may add a third-party search as well. The obvious candidate is Google.
- It’s Google, you get to leverage Google’s ancient hoodoo.
- Google site search is used on lots of major sites, eg. Guardian, Boing Boing.
- Resilience: there’s a backup search if SC’s search fails
- Every approach to search will give somewhat different results.
- You have to pay or else you have ads
- risk creepy tracking
- Being in thrall of dark overlords
- lack of customization or results
- can’t guarantee indexing of all texts
- interface is optimized for the wrong things, eg. “date”
Quality of results: mixed
Here’s a few sample searches side by side.
Dhamma: a common Pali word
SC has a dictionary result which is great. Both give relevant results, but Google is better at making core suttas prominent. Also, Google results are more focused on the EBTs, SCs are more random. We don’t weight the searches according to past searches, which would explain this.
winner: Google (narrowly)
Atammayata: a rare Pali word
SC gives some relevant results, although it doesn’t give them all, and misses the main canonical reference in MN. Google doesn’t get it at all.
Savatthi: a place name
SC gives a dictionary result and a map. Awesome! Both give relevant results.
Cat: a moderately common English word
Both give relevant results, with similar content near the top. Google gives multiple results for dictionary entries that include the word cat. We don’t give dictionary results for English words.
So the outcome is that both approaches have their uses. Of course, using Google site search is always possible for a user, so the question is whether it’s worth exposing it more prominently.
- Use another service such as Duck Duck Go. (I’ve checked the results, they are similar, in some cases better, than Googles.)
- Can’t have results on page, you gert redirected to DDG.
- Which then exposes you to ads.
- Expose Voice search on SC (only for a small subset of texts).
Anyway, having not used Google CSE I’m interested if anyone has any feedback, or if there are other alternatives.