Language on social media is one setting where the computational linguistics toolkit is applied to a specific domain. Internet communication has a history that can be traced through generations of users. The first wave learned the network as computer specialists and invented jargon, the second wave arrived for information or for social reasons and grew up on chat and early social sites, and the third wave has never known life without the internet and often treats email as obsolete. Small features such as capitalised versus lowercase LOL work as markers of belonging, so that even a spelling choice carries social meaning. This is the level of sociolinguistic variation, on which form points to community and to time.
The medium itself can then be defined: social media are characterised by user-generated content that is produced and consumed by the same people, and they vary along several dimensions, whether exchange is synchronous or asynchronous, whether the audience is one-to-many or one-to-one, and whether permissions run in both directions or only one. Each combination creates different conditions for writing, and therefore different norms. A chat between teenagers on a messenger, written fast and short, obeys other norms than a Wikipedia discussion page, where content is negotiated collaboratively.
What linguists do with such material is examine variation and innovation. Hashtags illustrate this well, since they can be sorted by content, by how words are combined, and by word class, so that morphology, syntax and semantics all meet in a single small form. Discourse analysis adds broader layers, namely the situation, the register, the genre and the style of a text. The result is an approach that treats non-standard language as a system.