Describing Languages with Regular Expressions

Transcription

1 University of Oslo : Department of Informatics Describing Languages with Regular Expressions Jonathon Read 25 September 2012 INF4820: Algorithms for AI and NLP

2 Outlook How can we write programs that handle sentences?

3 Outlook How can we write programs that handle sentences? Describing languages with regular expressions Representing and implementing regular expressions using finite state automata Estimating the probability of unobserved strings of words with language models Sequence-labelling part-of-speech using Hidden Markov models

4 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x =

5 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox.

6 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox. The hungry fox.

7 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox. The hungry fox. The hungry fox ate.

8 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox. The hungry fox. The hungry fox ate. The hungry fox ate the chicken.

9 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox. The hungry fox. The hungry fox ate. The hungry fox ate the chicken. The hungry fox quickly ate the chicken.

10 Productivity of languages Even simple formal languages are infinite: x =1 + 2 x = x = With natural languages there are so many more choices: The fox. The hungry fox. The hungry fox ate. The hungry fox ate the chicken. The hungry fox quickly ate the chicken. The hungry brown fox quickly ate the delicious roast chicken and washed it down with a pint of beer.

11 Characterising language Simplifying assumption A language is a set of utterances utterances inside this set are well-formed utterances not in this set are ill-formed

12 Characterising language Simplifying assumption A language is a set of utterances utterances inside this set are well-formed utterances not in this set are ill-formed How do we represent sets of utterances, if the set is infinite?

13 Regular expressions Regular expressions (RE, RegEx, RegExp): Algebraic notation for characterising sets of strings They consist of constants and operators Example /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ Note: an implementation is supplied in many programming languages and text editors for instance, try C-M-s in emacs, or grep on the command line.

14 Matching Sequences of character constants specify how to match strings. Further expressiveness is added by metacharacters, including: Example. any single character (except new lines) ˆ the start of a line $ the end of a line /ˆChapter.$/ { Chapter 1, Chapter 2,..., Chapter & } Note: When the literal of an operator or metacharacter i.e. one of {}[]()ˆ$. *+?\ should be matched, it must be escaped using a back slash, e.g. match a full-stop with /\./

15 Disjunction The operator expresses a logical or Example /ˆa (fox wolf)$/ { a fox, a wolf } Note: The operator has low precedence brackets ensure that it does not specify the set { a fox, wolf }

16 Character classes Character classes can also be used to specify disjunction they are expressed using square brackets, [ and ]: Examples /ˆ[Ff]ox$/ { Fox, fox } /ˆf[aio]x]$/ { fax, fix, fox } /ˆ[a-z]$/ { a, b, c,..., z } /ˆChapter [1-9]$/ { Chapter 1, Chapter 2,..., Chapter 9 }

17 Character classes Used inside a character class, ˆ negates the class: Example /[ˆA-Za-z]/ matches any non-alphabetic character /[ˆ ]/ matches anything that is not a space Many implementations provide named character classes: Examples /\d/ /[[:digit:]]/ /[0-9]/ /\w/ /[[:alnum:]]/ /[a-za-z0-9 ]/ /\D/ /[ˆ0-9]/ /[[:punct:]]/ matches punctuation characters

18 Quantification Quantification can be specified in a number of ways: Example? zero or one of the preceeding element * zero or more of the preceeding element + one or more of the preceeding element {n} exactly n of the preceeding element {n,m} from n to m of the preceeding element {n,} n or more of the preceeding element {,m} less than m of the preceeding element /ˆChapter [1-9]\d*$/ { Chapter 1, Chapter 2,..., Chapter 99999,... }

19 Lazy quantification How to match quoted items? Yes, he said, but why? Normal quantification operators are greedy they will the match the largest possible sequence in the input: /.+ / { Yes, he said, but why? } This can be overridden with?, which becomes the lazy operator when used next to a quantification operator: /.+? / { Yes, but why? }

20 Capturing groups Brackets are used to specify matching groups, which (a) enforce precedence and (b) indicate groups for later reference, using an escaped number (1-9). Example <[bi]>.+?</[bi]> { <b> </b>, <i> </i>, <b> </i>, <b> </i> } <([bi])>.+?</\1> { <b> </b>, <i> </i> }

21 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/

22 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / a word with an initial capital, followed by a space

23 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / /\d+[a-z]?/ a word with an initial capital, followed by a space one or more digits, optionally followed by a capital letter

24 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / a word with an initial capital, followed by a space /\d+[a-z]?/ one or more digits, optionally followed by a capital letter /, / a comma and a space

25 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / a word with an initial capital, followed by a space /\d+[a-z]?/ one or more digits, optionally followed by a capital letter /, / a comma and a space /\d{4} / four digits, followed by a space

26 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / a word with an initial capital, followed by a space /\d+[a-z]?/ one or more digits, optionally followed by a capital letter /, / a comma and a space /\d{4} / four digits, followed by a space /[A-Z][a-z]*/ a word with an initial capital

27 Putting it all together What does this match? /[A-Z][a-z]* \d+[a-z]?, \d{4} [A-Z][a-z]*/ /[A-Z][a-z]* / a word with an initial capital, followed by a space /\d+[a-z]?/ one or more digits, optionally followed by a capital letter /, / a comma and a space /\d{4} / four digits, followed by a space /[A-Z][a-z]*/ a word with an initial capital Gaustadalleén 23B, 0373 Oslo

28 Some exercises Write regular expressions for the following: 1. all alphabetic strings; 2. all lower case alphabetic strings ending in a b; 3. all strings of two repeated words; 4. all strings from the alphabet a,b such that a is immediately preceeded by and immediately followed by a b. 5. capturing the first word of an English sentence (making sure to deal with punctuation)

29 Some exercises 1. all alphabetic strings; /[a-za-z]+/

30 Some exercises 1. all alphabetic strings; /[a-za-z]+/ 2. all lower case alphabetic strings ending in a b; /[a-z]*b/

31 Some exercises 1. all alphabetic strings; /[a-za-z]+/ 2. all lower case alphabetic strings ending in a b; /[a-z]*b/ 3. all strings of two repeated words, separated by a space; /([a-za-z]+) \1/

32 Some exercises 1. all alphabetic strings; /[a-za-z]+/ 2. all lower case alphabetic strings ending in a b; /[a-z]*b/ 3. all strings of two repeated words, separated by a space; /([a-za-z]+) \1/ 4. all strings from the alphabet a,b such that a is immediately preceeded by and immediately followed by a b. /b+(ab+)+/

33 Some exercises 1. all alphabetic strings; /[a-za-z]+/ 2. all lower case alphabetic strings ending in a b; /[a-z]*b/ 3. all strings of two repeated words, separated by a space; /([a-za-z]+) \1/ 4. all strings from the alphabet a,b such that a is immediately preceeded by and immediately followed by a b. /b+(ab+)+/ 5. capturing the first word of an English sentence (making sure to deal with punctuation) /ˆ[ˆa-zA-Z]*([a-zA-Z]+)/

34 Applications in AI and NLP Weizenbaum 1966 User: Men are all alike. Eliza: In what way? User: They re always bugging us about something or other. Eliza: Can you think of a specific example? User: Well, my boyfriend made me come here. Eliza: Your boyfriend made you come here? User: He says I am depressed much of the time. Eliza: I am sorry to hear you are depressed.

35 Applications in AI and NLP Weizenbaum 1966 User: Eliza: User: Eliza: User: Eliza: User: Eliza: Men are all alike. In what way? They re always bugging us about something or other. Can you think of a specific example? Well, my boyfriend made me come here. Your boyfriend made you come here? He says I am depressed much of the time. I am sorry to hear you are depressed. Can be reproduced with a cascade of regular expression substitutions, e.g. using sed: s/.* all.*/in what way/ s/.* always.*/can you think of a specific example/ s/.* I am (depressed sad).*/i am sorry to hear you are \1/

36 Applications in AI and NLP Lexical morphology s/mouse/mice/ s/(bush fox house)/\1es/ s/(.)/\1s/

37 Applications in AI and NLP Lexical morphology s/mouse/mice/ s/(bush fox house)/\1es/ s/(.)/\1s/ Concise expressions of genetic sequences: finding codons, e.g. s/cg. AG[AG]/arginine specifying patterns e.g. /CG. AG[AG].{,100}GG./

38 Summary Regular expressions A finite way of specifying infinite sets Character constants, metacharacters and operators The fundamental operations are: Matching characters, wildcards (.) and anchors (ˆ and $) Disjunction ( and [ ]) Quantification (?, *, + and {n, m}) Precedence can be enforced with brackets (( and )) More complex operations include capturing groups Next week: Finite state automata Searching state spaces