Course Overview

1
Course Overview

• What is AI?
• What are the Major Challenges?
• What are the Main Techniques?
• Where are we failing, and why?
• Step back and look at the Science
• Step back and look at the History of AI
• What are the Major Schools of Thought?
• What of the Future?

• What are we trying to do? How far have we got?
• Natural language (text speech) (continued)
• Robotics
• Computer vision
• Problem solving
• Learning
• Board games
• Applied areas Video games, healthcare,
• What has been achieved, and not achieved, and
  why is it hard?

2
Language Technology
Interlingua open1 open2
Meaning
Hard!
Text
Text
Cheaters shortcut
Speech
Speech
3
Modern Machine Translation

• Prevalent approach uses statistics, following an
  idea by Warren Weaver (conceived as early as
  1947)
• View translation as a form of decoding Dutch is
  just coded English (or the other way round)
• i.e. look at the problem from the computers
  point of view
• Deciphering coded text, which replaces each
  English word with a coded word
• Suppose you have a large English text, and an
  even larger corpus of English
• You guess the correct version of a coded word by
  comparing the frequency of that word in the
  corpus with the frequency of all the words in
  your text
• E.g., most frequent word must be the, so the
  most frequent word in the corpus may be code for
  the. (Just a guess!)
• Check whether this combination of guesses is a
  proper English text change where necessary
• Can try with Google wound will cure/heal
  served his sorrow/sentence

4
Modern Machine Translation

• But of course, Dutch is not just coded English.
  (For example, the right translation for open
  may depend on the words surrounding open.)
• How do we find out how sentences in the two
  languages are related?
• To get a good starting point, Machine Translation
  uses huge bilingual corpora (usually based on
  human translation)
• Example Canadian Hansard corpus, bilingual
  French/English parliament proceedings, also Hong
  Kong
• (But Ill use Dutch as an example)

5
Modern Machine Translation

• Here we will not explain the statistical
  techniques used
• Just observe Guess how expressions line up
  across two languages
• Based on pure pattern matching. No knowledge of
  Dutch or English is required
• NB in statistical translation program, no longer
  easy to see understanding followed by generation

6
Modern Machine Translation

• perform various preparatory operations (e.g.,
  match corresponding sentences with each other)
• hypothesise ways of matching smaller expressions
  with each other. Example 1
• E that Blair responded
• N dat Blair antwoordde
• E whether Kennedy responded
• N of Kennedy antwoordde

7
Modern Machine Translation

• Here is a more interesting example, involving
  differences in word order between the two
  languages
• Example 2
• E that Blair responded to the question
• N dat Blair op de vraag antwoordde
• E whether Kennedy responded to the challenge
• N of Kennedy op de uitdaging antwoordde
• Need offsets in translation model

8
Models, Unigrams, Bigrams, Trigrams

• Need a translation model and a language model
• Translation model tells us likely translations
  (roughly)
• Language model tells us how good those sentences
  are in the target language
• Language model
• Ideally we would like to know how common any
  sentence is
• We will settle for pairs (bigrams)
• Translation model
• Often use something quite crude, like word by
  word
• Correct positions with offsets
• Good language model can save bad translation model

9
How far has Machine Translation advanced?

• National Institute of Standards and Technology
  (NIST)
• Regular competitions between MT systems
• (Source K.Knight, Statistical MT Tutorial,
  Aberdeen 2005)

10
winner 2002

• insistent Wednesday may recurred her trips to
  Libya tomorrow for flying
• Cairo 6-4 (AFP) an official announced today in
  the Egyptian lines company for flying Tuesday is
  a company insistent for flying may resumed a
  consideration of a day wednesday tomorrow her
  trips to Libya of security council decision trace
  international the imposed ban comment

winner 2003
Egyptair Has Tomorrow to Resume its flights to
Libya Cairo 4-6 (AFP) Said an official at the
Egyptian Aviation Company today that the company
egyptair may resume as of tomorrow, Wednesday its
flights to Libya after the International Security
Council resolution to the suspension of the
embargo imposed on Libya.
11
Conclusion on Statistical MT

• This approach to MT relies on massive parallel
  corpora these are not yet available for all
  language pairs
• The MT system does not understand the content
  of the sentences
• Perhaps progress using statistical methods will
  flatten in future
• but they are starting to be combined with
  higher-level information

12
Practical Machine Translation

• Types of translation
• Rough translation
• Could perhaps be post-edited by a monolingual
  human (cheaper)
• Restricted source translation
• Subject and form restricted, e.g. weather
  forecast
• Pre-edited translation
• Human pre-edits, e.g. Caterpillar English
• Can improve original too
• Literary

13
Summing up Modern Translation

• Deep vs. Shallow?
• Deep - comprehensive knowledge of the word.
• Shallow - no knowledge.
• So far, shallow approaches more successful.
• Deep can be better on a particular domain if a
  lot of expert effort is put into building models
• Shallow approach is much easier
• Similar story in other areas of AI
• Each of these programs on its own is highly
  specialised (i.e., limited)

14
On the other hand

• Humans dont always get it right either!
• French hotel Please leave your values at the
  front desk.
• Athens hotel We expect our visitors to complain
  daily at the office between the hours of 9 and 11
  a.m.
• Tokyo hotel room The flattening of underwear is
  the job of the chambermaid - get it done, turn
  her on.
• Hong Kong tailor shop Order your summer suit
  now. Because of big rush we execute customers in
  strict rotation.
• Men's room at Mexican golf course/resort
  Guests are requested not to wash their balls in
  the hand basins.
• Budapest elevator due to out of order we
  regret that you are unbearable
• Bangkok Dry Cleaner's Drop Trousers Here for
  Best Results
• Tokyo hotel room Please take advantage of our
  chambermaids.
• They do understand, but they may make the wrong
  choices in the target language

15
Speech Recognition

• Signal processing to recognise features
• Coarticulation model how each sound (phone)
  depends on neighbours
• Dialect different possible pronunciations
• To recognise isolated words use unigram language
  model again
• Continuous speech use bigram or trigram model
• Try
• eat I scream vs. eat ice cream
• eat a banana vs. eat a bandana

16
Speech Recognition

• Humans are remarkably good because of high level
  knowledge
• Computers
• No background noise, single speaker, vocabulary
  few thousand words
• gt99
• In general with good acoustics
• 60-80
• On noisy phone
• terrible

17
Natural Language Generation (NLG)

• Natural Language Generation is better than having
  people write texts when
• There are many potential documents to be written,
  differing according to the context (user,
  situation, language)
• There are some general principles behind document
  design

18
Example Noun Phrase design

• A noun phrase can convey an arbitrary amount of
  information
• Someone vs.
• a designer vs.
• an old designer vs.
• an old designer with red hair
• How much information should we pack into a
  given Noun Phrase?
• This is normally considered part of the
  aggregation task.

19
Some Issues to Consider

• Preferred ordering within the text (e.g. most
  important first)
• Readability of the Noun Phrase,
• Flow of focus,
• Successful use of pronouns and abbreviated
  references

20
Example Content

• (NB we assume that words, basic syntax etc have
  been chosen)
• This T-shirt was made by James Sportler .
• Sportler is a famous British designer.
• He drives an ancient pink Jaguar.
• He works in London with Thomas Wendsop.
• Wendsop won the first prize in the FWJG awards.
• Can/should we add more to the Noun Phrase?

21
One possible addition

• This T-shirt was made by James Sportler, who
  works in London with Thomas Wendsop .
• Sportler is a famous British designer. He drives
  an ancient pink Jaguar.
• Wendsop won the first prize in the FWJG awards.
• Facts about Wendsop are now separated from one
  another (focus).
• Wendsop now has greater prominence in the text
  (ordering)

22
Another possible addition

• This T-shirt was made by James Sportler, a famous
  British designer who works in London with Thomas
  Wendsop, who won the first prize in the FWJG
  awards .
• Sportler drives an ancient pink Jaguar.
• The Noun Phrase is now very complex (readability)
• He now doesnt seem to work in the second
  sentence (pronouns)

23
Another possible addition

• This T-shirt was made by James Sportler, a famous
  British designer .
• He drives an ancient pink Jaguar.
• He works in London with Thomas Wendsop.
• Wendsop won the first prize in the FWJG awards.
• Possibly the best solution, but is this better
  than the original text?

24
Why is Natural Language Generation hard?

• Natural Language Generation involves making many
  choices, e.g. which content to include, what
  order to say it in, what words and syntactic
  constructions to use.
• Linguistics does not provide us with a
  ready-made, precise theory about how to make such
  choices to produce coherent text
• The choices to be made interact with one another
  in complex ways
• Many results of choices (e.g. text length) are
  only visible at the end of the process
• There doesnt seem to be any simple and reliable
  way to order the choices