CS404 - AI and NLP

Getting computers to act like they have intelligence and understand natural language.

A very gentle overview/introduction to the topic.


 

Comparing Irish and English gives us insight into many linguistic phenomena.

1) SVO - Subject-Verb-Object structure
Irish is among 10% of the worlds languages where the Verb is followed by Subject, followed by Object. Most other European languages begin with the subject, then the verb then the object.

SVO

Sean went out. (English)

John est alle au magasin (French).

VSO (Irish)

Chuaigh Sean amach.

Chuaig Sean go dti an siopa.

 2) Head-First vs. Head-Last ordering (branching)
Let’s look at an adjectival phrase. In English, the adjective(s) is placed before the noun. Whereas in Irish the adjective(s) come after the noun.

Head-Last

...big red car.

...run quickly.

Head-First

...gluaistean mor dearg.

...rith go tapaigh.

 3) Ordinal-Person agreement
Irish requires different number terms for people and things, but English treats people and things alike.

People

Beirt

Triuir

Things

Dha sciain.

Tri platai.

People/Things

Two people

Two things

  

    1. Subject Verb Object
    2. "John went to the shop".
    3. "It is John who went to the shop".
    4. "It is the shop that John went to".
    5. "John it was who went to the shop".

Clearly, the structure changes a lot. "John" seems to appear almost anywhere in the sentence, but note that "went" is always next to "to" in these examples.
Automatic Machine Translation is available, free, from BabelFish, though it's less than perfect. Try translating any sentence into a foreign language and then back again. Did you get the same sentence?

-          Given that the technology exists to convert speech into text, text into speech, and machine translation, why can’t I combine these technologies to engage in a computer-assisted-conversation in a foreign language?

 

Topics: (Lecture notes available from Moodle).

  1. Introduction to artificial intelligence.
  2. Heuristic Search and problem solving, A*,
  3. Adversary Search, MiniMax, a-b pruning.
  4. Genetic Algorithms and biologically inspired computation.
  5. Cellular Automata.
  6. Document Retrieval, Ranking algorithms, Citation Ranking, Google’s PageRank algorithm.
  7. Syntax of English, Parsing, Chart parser, Shift-reduce parser, parsing
  8. Semantic processing, categories, WordNet, thematic roles.
  9. Interpreting analogical comparisons, Algorithms for identifying analogical counterparts.
  10. Statistical language processing and n-grams.


Books:

          Artificial Intelligence

1.        “Artificial Intelligence: A Modern Approach”, by Russell and Norvig. Or

2.       “Artificial Intelligence: Structures and Strategies for Complex Problem Solving”, by George F. Luger (2004).

Natural Language Processing

  1. "Natural Language Understanding", by James Allen, Second Edition, 1994. Or
  2. "Speech and Language Processing: An Introduction to Natural Language Processing", Speech Recognition, and Computational Linguistics. Prentice-Hall by Dan Jurafsky, and James H. Martin, 2000.  

See also, the free online book How language works, by Mike Gasser. ”Why Can't Computers Use English?” by the LSA.

 

Past Exam papers:

-          Note: due to a modification of the course, the following are the only really relevant past exam papers.

-          06-sample.pdf. Artificial Intelligence & NLP: 06-AI and NLP.

 

© Diarmuid O'Donoghue, NUI Maynooth. Dec., 2007.