The spelling in the captions can be hard to predict, and it's not good at fuzzy matching:

https://frinkiac.com/?p=search&q=I+was+saying+buu-urns

They may or may not (be allowed to) have a sense of humor about it. Our 24 Hours of LeMons car's publicity was sent to Matt Groening by a friend, and he passed it around the office. Apparently he asked their publicity folks if they could invite us up to show off the car about the same time that legal asked about sending us a cease and desist.

In the case of the car, it's probably fair use and the only issue was likely that we have non-Fox-approved sponsorship on it, but they probably decided their advertisers wouldn't complain about it because it's not exactly big bucks changing hands here.

So yeah, we got to meet Matt Groening and David X Cohen and Al Jean and a lot of the writers. It was definitely a cool experience.

http://www.thehomercar.com

Or even with just one quote:

https://frinkiac.com/?p=search&q=%22moon

It looks like it uses really naive word breaking, so it considers the punctuation part of the word.

They could definitely use some better text indexing/relevancy ranking implementations. I had mixed success. I'd recommend lucene or something based on lucene (Solr, ElasticSearch).