Lecture 2 Finite State Machines

1
Lecture 2Finite State Machines Morphology
CSCE 771 Natural Language Processing

• Topics
• Finite State Machines
• Morphology
• Readings Chapter 3 (skim Chapter 2)

June 4, 2009
2

• Last Time
• Challenge of 2001s HAL
• Areas of Research
• Examples of Language Processing
• Formal Languages
• Alphabets, strings,
• regular expressions denote languages
• finite automata (DFA, NFA) accept languages
• Today
• Slides from Lecture 1 30-
• Regular expressions in Perl, grep, vi, emacs,
  word?
• Eliza
• Morphology

3
Python ??

• Windows

4
Concepts on Regular Languages

• If r is a regular expression then there is a
  construction from r that will yield an NFA Mr
  such that L(r) L(Mr)
• Every NFA can be converted into an equivalent
  DFA?
• There are languages that cannot be regular.
• Every regular language can be generated by a
  regular grammar.

5
Grep family revisited

• Global match Regular Expression and Print (GREP)
• grep uUnix f1 f2 fn
• egrep pat files // efficient NFA?DFA, then
  execute
• fgrep pat files // fixed grep for fixed strings
• Find for searching directories (not really reg
  expr)
• find dir name pat // search for files with name
  matching pat
• find dir -exec grep pat //search in
  files for the pattern pat

6
Editing scripts

• Create a script of editing commands then execute
  with
• ex file1 lt edScript
• Example
• 1,s/uUnix/UNIX/g
• 1,s/langauge/language/g
• g//d // delete empty lines start of line
  end
• w
• q

7
Other Unix regular expression Based Tools

• sed (stream editor)
• awk
• Perl scripting language
• Python, Ruby
• Nltk now in Python

8
Perl Regular Expressions
9
Eliza Substitutions

• Eliza took the input and performed several
  transformations (substitutions) to produce the
  output.
• First
• s/mMy/YOUR/g my ? YOUR
• s/Im/YOU ARE/g Im ? YOU ARE

10
Eliza in Perl

• print "Welcome to Elizalike. Talk to me! (Or type
  \"bye\" to quit.)\n"
• Start an infinite loop
• while (1 1)
• This line reads in user input, and stores it in
  the special
• variable _, which makes the regular expression
• statements below more succint.
• _ ltSTDINgt
• Allow users to quit
• if (_ /byeByeBYE/)
• print "Elizalike Well, it was nice talking
  with you!\n" exit (0)

http//en.wikipedia.org/wiki/ELIZA
11

• Insert a tag at the beginning of the line to
  identify
• it as Eliza's reponse, and to make finding word
• boundaries easier.
• s//Elizalike /
• Replace all instances of "you are" with "Eliza
  is"
• Note how (\W) and \1, etc are used to mark word
• boundaries and keep whatever non-word character
  was
• in the input in the output.
• s/(\W)(youYou) are(\W)/\1Eliza is\3/g

http//en.wikipedia.org/wiki/ELIZA
12
Transformations

• x

http//en.wikipedia.org/wiki/ELIZA
13

• Print the result to STDOUT.
• print STDOUT

http//en.wikipedia.org/wiki/ELIZA
14
Morphology -

• A writer is someone who writes, a stinger is
  something that stings. But fingers dont fing,
  grocers dont groc, hammers dont ham and
  humdingers dont humding.
• Richard Lederer, Crazy English
• Consider mapping singular nouns to plurals
• Duck ? ducks, cat ? cats, book ? books
• Fox ? foxes, bush ? bushes
• Goose ? geese
• Fish ? fish
• Morphological parsing recognizing plurals
• Spelling rules
• Morphological rules

15
Morphology -

• Morphemes minimal meaning bearing unit
• Stems and affixes
• Cat -s
• Affixes
• Prefixes
• Suffixes
• Infixes
• Circumfixes
• Inflection combination of word stem with affix
  reulting in a word of the same class (noun ?
  noun)
• Dervivation combination of word stem with
  grammatical morpheme resulting in another class
  verb?noun (e.g. walk ing ? walking)

16
Turkish Example of a complex language

• The following is a Turkish word and its
  decomposition.

17
English Regular Nouns (Number)

• In other languages you typically indicate other
  features such as gender.
• Plurality
• possesive

SLP fig from 3.1.1
18
Regular Verbs

• X

SLP fig from 3.1.1
19
Irregular Verbs

• X

SLP fig from 3.1.1
20
To Love in Spanish

• X

OLD SLP fig from 3.1.1
21
Nominalization

• Nominalization formation of nouns from verbs or
  adjectives

SLP fig from 3.1.2
22
Adjectives from Nouns and Verbs

• X

23
Morphological Parsed Output

• N Noun, V Verb
• PL, SG, Present (default not shown), Past,
  Pres-Participle, Past-Participle

SLP fig 3.2
24
Figure 3.3 FSA for nominal inflection

• X

25
Figure 3.4 FSA for verbal inflection
26
Verb stem classes

• X

27
Fig 3.5 FSA for Adjective morphology

• Conisder
• Cool, happy, natural
• Big, equal

28

• X

29
Fig 3.6 FSA for another Fragment of English
derivational morphology

• X

30
3.7 English nouns with inflection

• X

31

• X

32

• X

33

• X

34
Fig 3.14 FSA for English nominilazation

• X

35

• X

36

• X

37

• X

38

• X

39

• x

40

• X