Testing taxonomic relations
1 / 1

Testing taxonomic relations Verena Rube (Project B6, Project Leader: Peter Koch) * - PowerPoint PPT Presentation

  • Uploaded on

Testing taxonomic relations Verena Rube (Project B6, Project Leader: Peter Koch) *. Research questions

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
Download Presentation

PowerPoint Slideshow about 'Testing taxonomic relations Verena Rube (Project B6, Project Leader: Peter Koch) *' - tessa

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Testing taxonomic relations verena rube project b6 project leader peter koch

Testing taxonomic relationsVerena Rube (Project B6, Project Leader: Peter Koch)*

  • Research questions

  • 1) How to access folk’s belief concering taxonomic relations which are said to hold on the semantic side of motivational relations? You can present subjects with test sentences like the following ones:

    • (1) Un’autostradaè un tipo distrada (A ‘motor way where you can drive very fast’ is a kind/type of ‘road, highway’)

    • (2)Un maxiprocesso è un tipo di processo (A‘process with a long duration and a lot of testimonies’ is a kind/type of ‘process (law)’)

  • 2) Is there a difference between examples like the one in (1) and the one in (2)?

  • Croft/Cruse (2004) : (1) = taxonymy; (2) = hyponymy.

  • Explanation: the core of the meaning of maxiprocesso is not a specification of the core of the meaning of processo. Processo is taxonomized on the basis of what kind of lawsuit is involved: e.g. processo civile.

  • →The diagnostic sentence “X is a kind/type of Y” would be adequate only in example (1).

  • Selection of motivated stimuli

  • stimulus candidate pairs (prepared by Yannick Versley (SFB 441, A1)):

    • internet corpus search with two shallow patterns: X_pl come Y_def (Xs such as Y), Y e altr[ei] X_pl (X and other Ys) for Italian hyponym/hyperonym noun pairs

  • - search restricted to pairs with a certain degree of surface similarity

  • - candidate pairs ranked according to an averaged smoothed pointwisemutualinformation statistic over two patterns (figure of merit).

  • 56 pairs from the list were chosen together with appropriate meanings: taxonomic pairs, hyponymic pairs and pairs which couldn’t be related

  • taxonomically or hyponymically in all possible readings, like possibilità - impossibilità.

  • A preliminary motivation test with informants was executed for the morphologically more complex word of the stimulus candidate pair and

  • more than half of the tested stimuli were motivated in the way it was predicted.

Rating task

37 subjects rated on a 7 point scale the appropriateness of the sentence “X è un tipo di Y” (X is a type of Y) filled with the stimulus pairs.

The meaning of each stimulus word was specified by an example sentence.

The 3 x 2 factorial design included the factor relation type: taxonomic pairs (8), hyponymic pairs (8) and pairs related in other ways (8 (16))

and the figure of merit of the pairs.


(1) Taxonomic pairs should be rated higher than hyponymic pairs and the other pairs should be rated very low.

(2) The figure of merit factor should influence only the taxonomic and hyponymic pairs.



F1 (2, 72)= 393 p< ,001 F1 (2, 74)= 11,279 p< ,001

F2 (2, 14)=121,1p< ,001 F2 (2, 6)= ,448 p< ,207

→ hypothesis (1) confirmed → hybrid interaction

Taxonomic pairs vs. hyponymic pairs: F1 (1, 36)= 36,95 p< ,001; F2 (1, 7)= 11,169 p= ,012

Items with high figure of merit were rated only slightly better than those with a low figure of merit: F1 (1,36)= 13,745 p= ,001; F2 (1,3)= ,465

p= ,544. The hybrid interaction which is significant for subjects would also suggest not considering the effect of that factor.

  • Conclusions

    • Hypothesis (1) was confirmed: the diagnostic sentence “X is a type of Y” is less applicable for hyponymic stimulus pairs and its not applicable for other kinds of relations.

    • But: Croft/Cruse’s hypothesis was that the sentence is not applicable to hyponymic stimulus pairs, the ratings though were still better than 4 which marks the zero point of the scale.

    • Further research

    • The experiment should be repeated with a Yes/No task that requires participants to make taxonomic/nontaxonomic judgments.

    • Prediction:

    • → taxonomic and hyponymic pairs should be judged as taxonomic because both were rated better than 0.

*I would like to thank Yannick Versley, Sam Featherston, Janina Radò, Oliver Bott & Mingya Liu for their help.