Latent Semantic Analysis
I am allergic to jargon! It can be used to
- conceal the lack of substance or meaning in an idea
- inflate the commonplace or trivial
- exclude the unititiated
- confer pseudo-scientific authority and claim false superiority
So what about Latent Semantic Analysis and its inevitable contraction to LSA? I came across it in an interesting conversation with Mike Joy about current issues in e-learning and assessment. LSA is a statistical method designed to measure the commonality of meaning in a collection of text passages or documents. It compares the frequency of significant words, numerically conflates their meanings, applies some mathematical jiggery-pokery to the data (viewed as sparse matrices), and comes up with some numbers that may indicate how close the texts are in meaning. It can be used as an alternative to the more familiar comparison of strings (a la Google) in detecting likely plagiarism. I believe that it is used effectively in monitoring plagiarism in program source code submitted for assessment by students in the Department of Computer Science.
Clearly jargon is both necessary and useful to experts, and LSA meets this test. It also has the virtue of meaning what it says: the analysis of hidden meaning. It might be interesting to run LSA on this and other blogs on plagiarism!