Reading Tim Oren's The State of Machine Translation post made me think about practical ways to use machine translation technology on the web despite all of its problems. Since I was wrestling with Hotspot JIT compiler just before, I naturally mixed the two to get: Hotspot Machine Translation.
Hmm. What could that mean? It's a Twinky in need of filling.
Documents have hotspots. Title is a hotspot. Phrase "sign here" is a hotspot.
User interfaces have hotspots. Buttons are hotter than menus, window titles, etc. Words that results in action are hotter than other types of words.
Web pages have hotspots. Hyperlinks are obvious hotspots, but buttons are hotter because they are often used to commit or submit.
I see now. Hotspot Machine Translation is realtime translation of words in hotspots. How are hotspots different?
- Hotspots contain incomplete sentences, typically just a few words.
- Number of words and phrases used in hotspots are potentially much smaller.
What can one do with it though beyond being able to more easily navigate foreign porn sites? Oops. My bus to Dreamland is here.