Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision | ||
external:spr [2011/10/21 14:11] smejkalova |
external:spr [2012/01/19 10:47] (current) smejkalova |
||
---|---|---|---|
Line 5: | Line 5: | ||
===== People and contacts ===== | ===== People and contacts ===== | ||
- | * Silvie Cinková cinkova@ufal.mff.cuni.cz | + | * Silvie Cinková |
- | * Martin Holub holub@ufal.mff.cuni.cz | + | * Martin Holub <holub (at) ufal.mff.cuni.cz> |
- | * Lenka Smejkalová smejkalova@ufal.mff.cuni.cz | + | * Lenka Smejkalová |
- | * Vincent Kríž vincent.kriz@gmail.com | + | * Vincent Kríž |
Institute of Formal and Applied Linguistics | Institute of Formal and Applied Linguistics | ||
Line 16: | Line 16: | ||
Malostranské náměstí 25 | Malostranské náměstí 25 | ||
CZ-118 00 Praha | CZ-118 00 Praha | ||
+ | |||
+ | |||
+ | |||
Line 21: | Line 24: | ||
"CPA Verb Validation Sample 30 (En)" is a newly developed lexical resource. It contains descriptions of the following 30 English verbs: | "CPA Verb Validation Sample 30 (En)" is a newly developed lexical resource. It contains descriptions of the following 30 English verbs: | ||
+ | |||
+ | |//access // |// | ||
+ | |// | ||
+ | |// | ||
+ | |// | ||
+ | |// | ||
+ | |// | ||
+ | |||
+ | |||
+ | |||
+ | Here we present a small portion of the data just to illustrate its structure. For each verb we define a set of semantic patterns and provide a sample of manually annotated corpus concordances. Then we do a detailed interannotator disagreement analysis, error detection, and adjudication. | ||
- | |//access // |//ally// |// | ||
- | |// | ||
- | |// | ||
- | |hail |halt |part |plug |plough | | ||
- | |pour |smash |smell |steer |submit | | ||
- | |swell | throw |trouble |wake |yield | | ||
- | Here we present a small portion of the data just to show its structure. | ||
==== Annotation Scheme Description ==== | ==== Annotation Scheme Description ==== | ||
- | | + | |
* Annotation Manual - [[http:// | * Annotation Manual - [[http:// | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | ==== Pattern Definitions Preview ==== | ||
+ | A few examples of revised PDEV entries: | ||
+ | * [[http:// | ||
+ | * [[http:// | ||
+ | * [[http:// | ||
+ | |||
+ | An example of a pattern definition form in detail: | ||
+ | * [[http:// | ||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
+ | |||
==== Annotated Data Preview ==== | ==== Annotated Data Preview ==== | ||
+ | Each of the examples below contains a multiple annotated set of 50 corpus concordances with manual disagreement analysis and final adjudication. The adjudicated data are used as a gold standard sample. | ||
+ | * cool - [[http:// | ||
+ | * deny - [[http:// | ||
+ | * yield - [[http:// | ||
- | ==== Disagreement Analysis Preview ==== | ||
- | * Pattern definitions - we have revised the pattern definitions for 30 verbs. Here is the sample of three of them (after revision): | ||
- | * [[http:// | ||
- | * [[http:// | ||
- | * [[http:// | ||
- | * Annotation of 50 concordances per each of these verbs by three annotators. Here is a little sample of three verbs. There has been already finished the manual disagreements analysis and all instances were adjudicated and gold sample was created. | ||
- | * Manual disagreements analysis and adjudication: | ||
- | * cool - [[http:// | ||
- | * deny - [[http:// | ||
- | * yield - [[http:// | ||
- | * Automatic disagreements analysis - confusion matrix for each pair of the annotators in one file | ||
- | * cool - [[http:// | ||
- | * deny - [[http:// | ||
- | * yield - [[http:// | ||
- | * Inter-annotator agreement (Cohen' | ||
- | Results | + | |
- | ^verb | + | ==== Disagreement Analysis Preview ==== |
- | | | + | Confusion matrices for each pair of annotators and automatic disagreements analysis: |
- | |cool | 50| 16| 0.685| 0.743| 0.669| 0.646| | + | |
- | |deny | 50| 10| 0.524| 0.434| 0.571| 0.582| | + | |
- | |yield | + | |
+ | |||
- | TODO: ne odlisna barva pozadi a rozsirit chlivky a kurziva (v tabulce sloves) + vycentrovat na strance | ||
- | TODO: zkontrolovat obsah souboru (specifikace formulare) | ||