1. Simple Maths for Keywords Adam Kilgarriff Lexical Computing Ltd

2. Liverpool, July 2009Kilgarriff: Simple Maths2 “This word is twice as common here as there”

3. Liverpool, July 2009Kilgarriff: Simple Maths3 “This word is twice as common here as there”  What does it mean? For word wubble Ratio=2: wubble is twice as common in fc as rc Freq (f)Corp SizePer million Focus corp (fc) 4010m4 Reference corp (rc) 5025m2

4. Liverpool, July 2009Kilgarriff: Simple Maths4 “This word is twice as common here as there”  Not just words Grammatical constructions Suffixes …  Keyword list Calculate ratio for all words Sort Keywords: at top of list

5. Liverpool, July 2009Kilgarriff: Simple Maths5 Good enough for keywords?  Almost, but 1.Are corpora well matched? 2.Burstiness 3.You can’t divide by zero 4.High ratios more common for rare words

6. Liverpool, July 2009Kilgarriff: Simple Maths6 1Are corpora well matched?  Proportionality If fiction contains more American, newspaper more British… genre compromised by region  Usual problem  Issue in corpus design  Not here

7. Liverpool, July 2009Kilgarriff: Simple Maths7 2Burstiness WordBNC freqBNC files mucosa10319 theology1032230 unfortunate1031648 Discount frequency for bursty words Gries, CL 2007, also CL journal We use ARF (average reduced frequency) Not here

8. Liverpool, July 2009Kilgarriff: Simple Maths8 3You can’t divide by zero  Standard solution: add one  Problem solved fc rcratio buggle100? stort1000? nammikin10000? fc rcratio buggle111 stort1011 nammikin10011

9. Liverpool, July 2009Kilgarriff: Simple Maths9 4High ratios more common for rarer words fc rc ratiointeresting? spug101 no grod100010010yes some researchers: grammar, grammar words some researchers: lexis content words No right answer Slider?

10. Liverpool, July 2009Kilgarriff: Simple Maths10 Solution  Don’t just add 1, add n: n=1  n=100 word fc rc fc+n rc+nRatioRank obscurish10011111.001 middling2001002011011.992 common120001000012001100011.203 word fc rc fc+n rc+nRatioRank obscurish1001101001.103 middling2001003002001.501 common120001000012100101001.202

11. Liverpool, July 2009Kilgarriff: Simple Maths11 Solution  n=1000 Summary word fc rc fc+n rc+nRatioRank obscurish100101010001.013 middling200100120011001.092 common120001000013000110001.181 word fc rc n=1 n=100n=1000 obscurish1001st2nd3rd middling2001002nd1st2nd common12000100003rd 1st

12. Liverpool, July 2009Kilgarriff: Simple Maths12 But what about  Mutual information  Log-likelihood  Chi-square  Fisher’s test  …  Don’t they use cleverer maths?

13. Liverpool, July 2009Kilgarriff: Simple Maths13 Yes but  Clever maths is for hypothesis testing Can you defeat null hypothesis?  Language is not random, so  … you always can  Null hypothesis never true  Hypothesis-testing not informative  Clever maths irrelevant Kilgarriff 2006, CLLT

14. Liverpool, July 2009Kilgarriff: Simple Maths14 Moreover…  just one answer grammar words vs content words? does not help  confuses and obscures

15. Liverpool, July 2009Kilgarriff: Simple Maths15 you should understand the maths you use

16. Liverpool, July 2009Kilgarriff: Simple Maths16 The Sketch Engine  Leading corpus query tool  Widely used by dictionary publishers, at universities  Large corpora for many lgs available  Word sketches  Web service  Since last week: Implements SimpleMaths

17. Liverpool, July 2009Kilgarriff: Simple Maths17 Example  BAWE British Academic Written English  Nesi and Thompson, completed last year Student essays  Arts/Humanities, Social Sciences, Life Sciences, Physical Sciences fc: ArtsHum, rc: SocSci With n=10 and n=1000

18. Liverpool, July 2009Kilgarriff: Simple Maths18

19. Liverpool, July 2009Kilgarriff: Simple Maths19

20. Liverpool, July 2009Kilgarriff: Simple Maths20 Thank you http://www.sketchengine.co.uk

21. Liverpool, July 2009Kilgarriff: Simple Maths21 Language is never ever ever random

22. Liverpool, July 2009Kilgarriff: Simple Maths22 Language

23. Liverpool, July 2009Kilgarriff: Simple Maths23 is

24. Liverpool, July 2009Kilgarriff: Simple Maths24 never

25. Liverpool, July 2009Kilgarriff: Simple Maths25 ever

26. Liverpool, July 2009Kilgarriff: Simple Maths26 ever

27. Liverpool, July 2009Kilgarriff: Simple Maths27 random