After a long hiatus, I think I will start posting on research papers again. I’ve been reading still, I just have been pretty busy with other stuff so I haven’t written about them. Here goes.

Frame Semantics, by Charles J. Fillmore. SICOL 1981.

This is (one of?) the introductory paper about frame semantics. There are no experiments or anything; Fillmore just describes the theory, its development, and why he thinks it’s a good model of semantics.

The gist of the theory is that words evoke “frames,” or types of events that happen in human experience. For instance, the word “buy” evokes a frame that involves the offering of money in exchange for goods or services; there are several participants in this frame that are all brought to mind with the use of any word that evokes the frame (other words might be “sell,” “buyer,” or “price”).

Fillmore gives some pretty good examples and explanations of what he’s talking about, and I think it’s a pretty good theory of semantics. There have been some data sets created where people have manually specified a number of frames and labeled text with instances of those frames, including which word evokes the frame and what parts of the sentence fill each role in the frame. I’m currently building a model that maps noun phrases in a document into real world entities in a knowledge base (pretty similar to Google’s new “knowledge graph” stuff). When I have that working, the next step is to put it inside of a generative model of frame semantics, which I think would be pretty cool.