Reading this paper I came across term "surface expression". What does it mean?


Surface expression basically means the form you see. It is the result of applying a rule to some (hypothetical) underlying form of a word (or phrase). So the deep structure would be something like 'dog+plural' and the surface form of that would be 'dogs' or 'I+direct object' and 'me'.

It is largely used in generative contexts and even then only historically since the original deep vs. surface structure distinction has been mostly abandoned. It can be a useful shortcut when talking about the results of rules but few people believe that there is actually any such thing as deep structure.


Surface expression = surface form = input form = the token as it appeared in the text (no stemming, no lemmatization, etc.).

For example, the tokens sang, sung, and sings are all surface expressions of the verb sing.

enter image description here

(Figure from Martin, James H., and Daniel Jurafsky. "Speech and language processing." 2nd Edition (2008))

  • I don't think the term 'surface expression' is common in non-computational linguistics? Oct 8 '15 at 3:14
  • @GastonÜmlaut unsure I mostly read nlp literature, not linguistics. Oct 8 '15 at 3:24
  • @GastonÜmlaut how would a linguist call a surface expression? Oct 8 '15 at 4:22
  • I think your example is a bit misleading. 'sang' is not the surface form or 'sing' but rather of 'sing+past'. That's what makes it different from just lemma - because it is not limited just to lexical items. The attached figure explains it better. Oct 8 '15 at 8:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.