It's partly driven by a hope that if you can model language well enough, you'll then have a model of knowledge. Logical positivism tried that with logical systems, which are much more precise languages of expressing facts, and it still fell on its face.