Two Ways of Looking at a Sex Act

First up, "sex act" is a great tongue-twister. So is "tugboat goat," though it's more contrived.

Second, the use of "act" is a little striking, when you think about it. "Act" has overtones of "acting," of theatricality, performativity, and other fun ivities. So this is sort of an interesting linguistic convergence, certainly one as least as meaningful as decompising "history" into "his story." If something is ritualized enough to be a "sex act," the implication is that it's aquired a certain veneer of superficiality. That sounds about right.

Unfortunately, as much as I'd like to read the musings of gender and cultural theorists on this idea, I believe that this is a concept for which no good search exists. I can't boil it down to Google keywords, a Lexis terms-and-connectors search, or a JSTOR term-based search.

The problem here is that the key element in the search is a semantic relationship between the search terms, not a syntactical one. I don't want papers that mention "sex acts" and "acting." I don't want papers that mention "sex acts" and "acting" in close juxtaposition. I don't even want papers about the "acting" of "sex acts." I want papers about "acting" in the context of the phrase "sex acts." But those documents are blocked from view by the far more common documents that use the same meanings of the same words in a different way.

That's the nasty part. A clustering search engine might well be able to separate news stories about sex workers from academic papers about them. But there's nothing globally distinctive about the academic papers I'm looking for. They just have a distinctive local idea; one that's sufficiently uncommon (if it's in the literature at all) that hand-review of a results list is infeasible.

Now, there are domains that are effectively unsearchable. Languages, like Thai, whose written forms don't indicate word breaks, can be difficult for automated tools to analyze. Pictures and music are difficult even to index. But smart people are working very hard on these problems, with results that are, at the least, encouraging. Semantic screening, though, might well be AI-complete. I'm just having a hard time imagining a search engine capable of making these distinctions without understanging the documents well enough to start playing the Turing game on them.