Contrasting Polish and English Derivational Groups

1
Contrasting Polish and English Derivational Groups

• Karolina Tymowicz

• based on
• Jadacka, H. Rzeczeownik polski jako baza
  derywacyjna,WN-PWN 1995
• independent contrastive study of 540
  Polish-English pairs of derivations

November 28th 2000
2
Outline

• Defining terms
• Derivational group
• Derivational base
• Affixes
• Similarity of and within derivational groups
• Procedure of comparison
• Conclusions

3
Derivational group

• A well-ordered system constructed around an
  underived entry word concentrating all the
  derivatives connected with it by means of direct
  or indirect process of derivation
• a hierarchical structure in which each element
  functions as a link between other derivatives and
  the BASE

4
Derivational base

• The item to which an affix is added to derive a
  new word-form
• the word-forms consisting of the derivational
  base and an affix are called DERIVATIVES
• e.g. STYLE - STYLIZE - STYLIZER
• e.g. CENTRE - CENTRIC - CENTRICALLY

5
Affix

• a morpheme that is added to a word, and which
  changes the meaning or function of the word
• affixes are bound-forms that can be added
• to the beginning of a word a prefix, e.g.
  unkind
• to the end of a word suffix, e.g. kindness

6
Similarity within derivational groups

• Four kinds of similarities within derivational
  groups are considered. Three types of
  translational similarity
• translational similarity between morphemes
• translational similarity between derivatives
• translational similarity between derivational
  groups
• and one type of grapho-etymological similarity
• graphemic and etymological similarity between
  bases

7
degrees of translational similarity between
morphemes (incl. bases)

• def. translational similarity between L1 and L2
  morphemes is a degree to which L1 morpheme can
  correctly be rendered as a corresponding L2
  morpheme (i.e. morphemes occupying the same
  position with respect to the base).
• no similarity, e.g.
• ponad- vs. -less in P. ponad-czasowy, E.
  time-less)
• 1st degree of similarity, e.g.
• bez- vs. -less in P. bez-glosny, E. voice-less
• 2nd degree of similarity, e.g.
• -ik vs. -er in P. glosn-ik, E. loudspeak-er
• -czas- vs. time- in P. ponad-czas-owy, E.
  time-less)

8
degrees of translational similarity between
derivatives

• def. a joint translational similarity between
  all the corresponding morphemes of the Polish and
  English derivatives
• e.g. Pol. Eng.
• za- a-
• les- forest
• ac ? ?
• whereby two morphemes are corresponding iff they
  occupy the same position with respect to the base.

9
degrees of translational similarity between
derivative groups

• similarity between derivational groups is a
  function of
• the grapho-etymological similarity of their
  bases,
• and the translational similarity of all their
  derivatives.

10
Degrees of graphemic-etymological similarity
between derivational bases

• def. Similarity established between two bases
  with respect to their etymological and graphemic
  features with the assumption of their
  translational equivalence
• no similarity, e.g. dom vs. house
• remote similarity, e.g. brat vs. brother
• close similarity, e.g. styl vs. style

irrespective of the translational equivalence of
their derivatives
11
Scale of translational similarity between
derivatives

• This scale used here consists of 12 levels of
  similarity counted from 11 to 0, where 0 stands
  for the lowest level of similarity and 11 denotes
  the highest level of similarity.

0 1 2 3 4 5
6 7 8 9 10
11
12
Treatment of compound derivatives

• If a single compound derivative of the form
  A-B or AB (but not A B) has an equivalent
  in the other language in the form of 2 separate
  words C D then it is included into our
  classification as long as
• C is a direct translation of A and D is a direct
  translation of B
• or C is a direct translation of B and D is a
  direct translation of A.
• This convention has been adopted because
• Jadackas derivational groups contain only
  derivatives of the form AB or A-B, but no A
  B derivatives
• Jadackas work constituted the main and most
  reliable source of derivatives and derivational
  groups considered in the study.

13
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity

• 11. P. BASE1 BASE2 SUFFIX
• E. BASE1 BASE2 SUFFIX
• e.g. slowo - word
• slowo-twór-stwo ? word form-ation
• 10. E. BASE1 (BASE2 SUFFIX)
• P. (BASE2 SUFFIX) BASE1
• e.g. krew - blood
• blood-stain-ed ? poplamio-ny krwia
• 9. E. BASE1 BASE2
• P. BASE2 (BASE1 SUFFIX)
• e.g. glos - voice
• voice-mail ? poczta glos-owa

Compound derivatives 1
14
Compound derivatives 2
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity

• 8. P. BASE1 BASE2
• E. BASE1 BASE2
• e.g. slowo - word
• pól-slowo ? half-word
• 7. E. BASE1 BASE2
• P. BASE2 BASE1
• e.g. styl - style
• free-style ? styl wolny

15
0 1 2 3 4 5
6 7 8 9 10
11

• 6. P. BASE SUFFIX
• E. BASE SUFFIX
• e.g. las - forest
• les-nik forest-er
• P. BASE SUFFIX SUFFIX
• E. BASE SUFFIX SUFFIX
• e.g. styl - style
• styl-ist-yczny styl-ist-ic
• P. PREFIX BASE SUFFIX
• E. PREFIX BASE SUFFIX
• e.g. las - forest
• wy-les-anie de-forest-ation
• P. PREFIX BASE SUFFIX SUFFIX
• E. PREFIX BASE SUFFIX SUFFIX
• e.g. centrum - centre
• de-centr-al-izowac de-centr-al-ize

Scale of similarity
Single derivatives 1
16
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity

• 5. P. PREFIX BASE SUFFIX
• E. BASE SUFFIX SUFFIX
• e.g. dziecko - child
• bez-dziet-nosc ? child-less-ness
• 4. P. PREFIX BASE SUFFIX
• E. BASE SUFFIX
• e.g. pan - lord
• wielko-pan-ski ? lord-ly
• 3. P. PREFIX BASE SUFFIX
• E. PREFIX BASE
• e.g. las - forest
• za-les-ac ? a-forest

Single derivatives 2
17

• 2. P. BASE SUFFIX
• E. BASE ____
• e.g. slowo - word
• slow-nik ? word-book
• P. BASE SUFFIX
• E. BASE
• e.g. dziecko - child
• diec-inka ? child
• 1. P. BASE SUFFIX SUFFIX
• E. _____ _______ SUFFIX
• e.g. slowo - word
• slow-nik-arz ? lexico-graph-er
• P. BASE SUFFIX
• E. _____ SUFFIX
• e.g. znak - sign
• znacz-nik ? mark-er

0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity
Single derivatives 3
18
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity

• 0. E. BASE BASE
• P. _____
• e.g. time - czas
• time-piece ? zegarek
• P. BASE SUFFIX
• E. _____
• e.g. kosc - bone
• kos-tka ? ankle
• E. PREFIX BASE
• P. _______
• e.g. child - dziecko
• grand-child ? wnuk

Single derivatives 4
19
Experiment

• 540 Polish-English pairs of derivatives were
  judged as to their similarity according to the
  12-point scale presented above
• the translational similarity points for each pair
  of derivatives obtained for each of the Polish
  and English bases together with the
  grapho-etymological similarity between these
  bases were analysed statistically

20
Statistical tests applied in the study

• in spite of nonnormality of the data the
  following parametric tests were applied
• MANOVA for
• for translational similarity between derivatives
  by
• grapho-etymological similarity between the basis
  these derivatives were obtained from, and
• direction of translation
• (Polish-English based on Jadacka 95 and Collins
  Polish-English Electronic Dictionary,
• English-Polish based on Harper-Collins
  Electronic Dictionary and Collins English-Polish
  Electronic Dictionary)
• Multiple Range Tests for
• translational similarity of the derivatives,
  irrespective of whether they were obtained
  through Polish-English or English-Polish
  translation
• by grapho-etymological similarity between the
  Polish and English bases they were derived from
• Multiple Range Tests for
• translational similarity of the derivatives
  obtained through Polish-English translation
• by grapho-etymological similarity between the
  Polish and English bases they were derived from
• additionally some non-parametric tests were
  applied
• Mann-Whitney W test to compare
• medians of the similarity points obtained for the
  derivatives in Polish-English translation
• with the medians of the similarity points
  obtained for the derivatives in English-Polish
  translation

21
Some results MANOVA

• Type III Sums of Squares was used
• All F-ratios were based on the residual mean
  square error.
• Source
  Sum of Squares Df Mean Square F-Ratio
  P-Value
• Agraph_ethym_sim_betw_bases
  590,704 2 295,352 53,53
  0,0000
• Bdirection_of_translation
  195,227 1 195,227 35,38
  0,0000
• RESIDUAL 2957,27
  536 5,5173
• TOTAL (CORRECTED) 3903,44
  539
• The P-values test the statistical significance of
  each of the sources. Since P-values are less than
  0,05,
• these grapho-etymological similarity between
  bases and the direction of translation have a
  statistically
• significant effect on the translational
  similarity between the derivatives obtained from
  these bases
• at the 95,0 confidence level.

22
Some results Multiple Range Tests

• Contrast
  Difference /- Limits
• 0 - 1 0,197742
  1,25397
• 0 - 2 -2,60124
  0,488299
• 1 - 2 -2,79898
  1,30672
• denotes a statistically significant difference.
• which means that the derivational groups
• of the Polish-English bases that were judged
  to bear no similarity with respect to their
  grapho-etymological features, and the
  derivational groups
• of the bases that were judged to be remotely
  similar with respect to their grapho-etymological
  features
• (i.e. 0-1) do not differ significantly with
  respect to the similarity of the derivatives that
  constitute derivational groups of each of these
  basis.
• on the other hand, groups derived from bases
  that differed in their etymology and graphemic
  representation (contrasts 0-2 and 1-2) have
  significantly different derivatives as far as the
  translational similarity of these derivatives is
  concerned.

23
1 2 5 7
8
Frequency Cumulative
540 observations 100
24
Applications of the study

• The results of the study provide insights into
  the possibility of automatic translation of
  UNKNOWN L1 derivatives on the basis of
• the L2 equivalents of the component morphemes of
  L1 derivative
• the degree of grapho-etymological similarity
  between the bases of these derivatives

25

• For example
• assume
• we do not know the equivalent of a derivative
  lesnik
• we can interpret bases even if they are modified
  by other morphemes (las ? les-)
• we know the equivalents of the component
  morphemes
• les- ( las) ? forest
• -nik ? -er
• we know the grapho-etymological similarity
  between the bases ( 0)
• Hence, we guess with a relatively small certainty
  that
• English equivalent of lesnik is forester

26
Pessimistic scenario for automatic translation of
derivatives
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity
27
Optimistic scenario for automatic translation of
derivatives
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity
28
Very optimistic scenario for automatic
translation of derivatives
0 1 2 3 4 5
6 7 8 9 10
11
Scale of similarity
29
Conclusions

• COMPOSITIONALITY The meaning of the derivative
  is a direct function of the meaning of its
  morphemes in app. 38-56 of cases
• Assuming we know the equivalents of all the
  morphemes of an L1 derivative we have app. 38-56
  chance of producing a comprehensible L2
  derivative
• The grapho-etymological similarity of L1 and L2
  bases influences the translational similarity of
  their derivational groups