Getting “hung up on semantics” is the kiss of death in an argument. It means your position lacks substance or is merely a clever construction of words. It’s a pejorative phrase to which few are willing to accede.
Why do I admit to being “hung up on semantics,” then? As the enterprise architect for content at the LDS Church, one of my jobs is to design systems that enable those interested in the Church to find what they need as easily as possible. That means our systems—not just our people—need to understand everything possible about the semantics of the content we produce.
In human discourse, we tolerate a high degree of imprecision. We use interpolation, probability, feedback, experience—even circumlocution—to fill in absent or ambiguous information. Unfortunately, machines are not so tolerant. A single bug in a computer program can destroy a rocket or cripple an Aegis missile cruiser. Keywords in a search engine result in thousands or millions of hits—a nearly useless list of duplicates, derivatives, and detritus. As Marshall McLuhan predicted, “We shape our tools and thereafter our tools shape us.” Our tools—our ways of finding what we seek—are breaking down under the intense pressure caused by the volume of available information and the friction between what we need to know and what our tools allow us to find.So how do we get our machines to understand our words so that they can help us find what we seek? Natural language processing, a subset of artificial intelligence, may someday help us break through the human-machine barrier, but few are holding their breath. A less grandiose, but more promising, approach is the digital equivalent of sticky notes: tags.
Tags are everywhere. They identify our photos on flickr, songs on our iPods, our favorites on Digg or Facebook or reddit. But they can also be used to identify the structure of a conference talk or display Church history events as a timeline rather than a list of hits. Tags are the clues that let machines help us rather than hinder us.
Tags can help us identify the entities (people, places, organizations, events, etc.) within our narratives (scriptures, conference talks, blog posts, lyrics, movies, presentations, etc.) and the relationships (creator/created by, counsels/is counseled by, etc.) between entity and entity and narrative and entity. This enrichment can be automated or refined by the collaborative efforts of an entire community. The end result can be understood by a machine, which uses them to provide relevant, focused, in-context results to queries. To the machine, semantics are not merely a clever construction, but the core of its processing.
With a semantic understanding of the content produced by the Church, our future tools can provide much richer experiences:
- A newly-called bishop can use his mobile device to find the one article in a Church handbook describing how to conduct a funeral because the search engine on a future LDS.org knows he’s a bishop, knows what other bishops have read when they used the keyword “funeral,” and can distinguish articles about funerals for specific individuals from articles on how to conduct a funeral.
- A student of Joseph Smith’s life can see the Prophet’s writings not only as a list of sermons, revelations, and articles, but can plot those writings on a geographical map indicating where they were created, or view them in a scrollable timeline—or all three simultaneously.
- A gospel doctrine teacher can dynamically assemble her lesson online by extracting the specific quotes, video, images, and other digital resources referred to in the manual, and then save it as a document or presentation for playback in class.
Far from lost in semantics—or, more accurately, the lack of semantics—the machine can now build on the meaning inherent in our tags and be part of the solution, not the problem. And that’s a “hang up” worth having.
Stewart Shelline is an enterprise architect for the Church.