Natural Language and Artificial Intelligence

HSSP Summer 2009, Instructor: Gregory Marton

July 12 -- First Steps

We will introduce the concepts of syntax, semantics, and meaning, in both human languages and computer languages. We will take our first steps in writing a Scheme program, and start to think deeply about English.

Computing: numbers, strings, booleans, procedures, conditionals, naming
Linguistics: nouns, verbs, conditionals
Lab exercise: get set up with subversion and PLT; Generate simple sentences that say hello to someone and incorporate an adjective about them.

July 19 -- Languages, Sounds, Alphabets

We'll look at the world's most common and least common languages, talk about language change and language families, writing systems and sounds, and all that makes one language different from another.

Computing: lists: map and filter, recursion, machine learning, probability
Linguistics: Languages of the world, language families, sounds and alphabets
Lab exercise: Automatically identify whether a text is in English or Dutch.

July 26 -- Morphology

What are words and what is the significance of spaces? Languages like Turkish, Hungarian, and to some extent German, can express entire sentences in one word. Languages like Chinese have words, but do not use spaces in writing. Even in English, we can see regularities of meaning that are smaller than words.

Computing: associations, map over more than one list, ormap
Linguistics: morphemes are the smallest units of meaning; regular and irregular word endings; where words come from
Lab exercise: Predict regular word endings and thus find the irregular ones.

Aug 2 -- Words and Keyword Search

When we put morphemes together into words, they still don't convey just one meaning:

Computing: mappings: list->mapping, mapping-reduce, indexing
Linguistics: homographs and homophones, synonyms, hyponyms, meronyms, stemming
Lab Exercise: Search Twitter-feeds case-sensitively and case-insensitively, and with collapsing morphologically different forms of the same word.

Aug 9 -- Multi-word meanings

Computing: nested data structures, counting bigrams
Linguistics: verb-particle constructions, selectional restrictions
Lab exercise: Extend Twitter search to search for sequences, rather than sets, of words. Weight common words less than rare words.

Aug 16 -- Syntax and Development

You've heard of subjects and predicates -- where do these come from? Is there really structure to our sentences, or are they just n-grams, like the multi-word entities. If we're careful, we can unambiguously hear that there is structure in a sentence, but it's not easy, so how to do kids figure it out? What else do kids invent that they never hear?

Computing: visualization -- turning mappings into scatter plots and bar graphs
Linguistics: syntax, wanna contraction, clefting for constituency, coordination for constituent type, overregularization
Lab exercise: We will use our search engine to search for incorrect irregular verb endings in actual transcripts of children's speech at various ages. We'll make graphs that show children's very odd learning patterns.

Aug 23 -- Semantics and Grounding

Guest Lecturer: Stefanie Tellex

In order for computers to really understand anything of our language, we and they must share some of the same kinds of experiences. Computers are learning to see, to hear, to move around, and interact with the world around them, so there is hope. But they need us to teach them what things mean.

Computing: automatic evaluation; review: features and machine learning
Linguistics: semantics of spatial prepositions
Lab exercise: annotate time and spatial expression examples

Aug 30 -- Meanings as Programs

Many meanings have to be grounded in the real world, and connected to cameras and motors and experiences, but surely there are some meanings that computers can more easily relate to, like numbers and time. Can we get computers to talk and reason about numbers and about time in a human language? How do we put together what we've learned about morphemes, words, and structure, and teach a computer to communicate coherently about a few things?

Computing: procedures as return values, currying
Linguistics: Eliza, procedural semantics, CCG, discourse and coherence, Gricean implicature, brains.
Lab exercise: write meanings for some simple sentences, course evaluation