Presentation Title

Author

How do computers help with language?

Language in the World

Media

https://www.youtube.com/watch?v=ARJ8cAGm6JE

Recall

  • Writing is learned
    • But certain aspects of computer-mediated communication (CMC) seem like they are acquired
    • Emojis as extralinguistic material (like facial expressions and body language)
  • What makes the digital environment different from face-to-face?
  • Computational linguistics is the area of linguistics that applies the methods of computer science to the analysis of language and speech
    • Speech synthesis and speech recognition
    • Not easy problems!!
    • Accuracy and speed

But also…

Computers are stupid.

  • Our focus before was on speech recognition , and we spent our time discussing mostly how sounds were processed and turned into text that a computer could understand
    • But that “understanding” boiled down to 1s and 0s
    • And it relied on the devices being talked to having sets of standard commands (“call Mom”) and built-in probabilities that it could use to decipher what you intended
  • So, is it really “understanding” language?
    • We’re talking about natural language processing
    • The answer to this lies in understanding the complexity of human grammar
    • And, in fact, computers are a still not great on the morphosyntactic properties of language!

Beyond recognition

Media

https://www.youtube.com/watch?v=fOvTtapxa9c

A little bit of a rewind!

The Determiner
Cat Noun
Chased Verb
The Determiner
Dog Noun
  • What is syntax?
    • “The way in which words are put together to form phrases and sentences.” (American Heritage Dictionary)
    • It’s all about structure and rules!
  • Remember, in order to talk about sentence structure, we need to know a little about parts of speech , or categories of words based on similar meanings, forms, and functions
  • You know these! But it’s not always unambiguous…

Q slide

What is the part of speech of “run”? *Verb, *noun, pronoun, preposition

The man bit the dog.

The dog bit the man.

What’s the difference?

What a model of syntax must do…

1. Words belong to lexical categories

We did this already, so know them!

If you are having trouble, please find a friend who is a master of parts of speech.

You will need to know these…

2. Word order is important

  • It affects grammaticality and meaning.
    • Mary dumped John. ≠ John dumped Mary.
    • *John Mary dumped.
    • *Dumped Mary John.
  • Languages differ in word order ::: {.notes}

Word order is more flexible in some languages compared to English, particularly languages that heavily inflect for case.

:::

Word Order

  • We usually talk about word order in terms of grammatical roles like subject (S), object (O), and verb (V)
  • These pieces are combined in different ways
    • SVO, SOV, VSO, VOS, OSV, OVS
    • For example, Irish and Welsh are VSO
    • Siaradodd Aled y Gymraeg.
      • spoke Aled DEF Welsh
      • Verb Subject Object
      • “Aled spoke Welsh.”
  • Almost no languages use OSV or OVS
    • However, every language has some rules about word order.

Activity

What is the word order of English?

3. Words form constituents

Constituency

  • We are trying to show that certain words clump together into groups – and even these groups clump together into bigger and bigger groups.
  • So, we have to think about grammar as being hierarchical
    • Like a tree rather than like a line

4. Constituents form a hierarchical structure

  • Each node is equivalent to a constituent!
    • Everything that hangs below a node (directly or indirectly) is inside that particular constituent, and nothing else.
    • One word can belong to several different (bigger and bigger) constituents at different levels.

Why hierarchical?

  • All utterances are said or written in a linear mode.
  • This can lead to ambiguity…
    • Yoko Ono will talk about her husband John Lennon who was killed in an interview with Barbara Walters.
    • How is this ambiguous?

Activity

  • Which picture could depict this sentence?
    • Sarah saw a man with a telescope.
      • A. This one: B. This one:
      • C. Both

Ambiguity

  • The sentence has 2 meanings:
    • a) Sarah saw [a man with a telescope]. = the man was holding a telescope
    • b) Sarah saw [a man] [with a telescope]. = she used a telescope to see him
  • Diagramming the sentences’ underlying hierarchical structure can disambiguate the meaning.
  • In the scene from 2001: A Space Odyssey, HAL says, “I’m sorry, Dave. I’m afraid I can’t do that”
    • This requires a level of understanding of “Open the pod bay doors, HAL.” that computers just don’t have!
    • All the stuff we’ve said about syntax today is stuff you do automatically!
    • To be able to parse, or break down a sentence into its component parts so that the meaning of the sentence can be understood, a computer would need to (at least) know:
      • Meanings of the words and word order
      • What makes a sentence (specifically an imperative, or command, one)
      • How to respond to imperatives (that they have never encountered and are not likely in the database)

How is this about computers?

Help me, HAL!

  • And, in addition to that, to be able respond in this way, a computer would need to:
    • Recognize that “I’m afraid” can mean something like “I’m sorry” not just “I have fear”
    • That using something like “I’m afraid” would be acceptable/appropriate here
    • Be able to not just execute the command (=free will?!)
    • Refer to the asker by name
    • And so much more!
  • We’ve done better with speech recognition than with true linguistic processing!
    • For a computer to understand meaning, we have to give it a lot more than algorithms and probabilities!
    • “Simple-minded definitions will give the wrong results.” (p. 107)

What’s next?

  • Come Wednesday for “Are chatbots real?” (read Bender)
  • Quiz 10 in recitation on Friday
  • Observation notes #6 (last one!) due Friday!
    • No late work, no AI
    • If you are happy with your grade in this category, you can skip this one, as we drop the lowest one