FoR-UNL

From UNL Wiki
(Difference between revisions)
Jump to: navigation, search
(→Descriptors)
 
(33 intermediate revisions by one user not shown)
Line 1: Line 1:
−
FRAU (FRAmework of reference for UNL) is a guideline used to describe achievements of natural languages in relation to UNL. It was inspired by the [http://www.coe.int/t/dg4/linguistic/CADRE_EN.asp Common European Framework of Reference for Languages (CEFR)], and its main goal is to provide a method for assessing the availability and quality of natural language resources inside the UNL framework.
+
FoR-UNL ('''F'''ramework '''o'''f '''R'''eference for '''UNL''') is a guideline used to describe achievements of natural languages in relation to UNL. It was inspired by the [http://www.coe.int/t/dg4/linguistic/Cadre1_en.asp Common European Framework of Reference for Languages (CEFR)], and its main goal is to provide a method for assessing the availability and quality of natural language resources inside the [[UNL System]].
  
 
== Reference Levels ==
 
== Reference Levels ==
−
FRAU divides languages into three broad divisions which can be divided into six levels:
+
The FoR-UNL classifies languages in three broad divisions which can be divided into six levels, according to the recall and precision of the corresponding resources:
 
*'''A''' - Basic Level
 
*'''A''' - Basic Level
 
**'''A1''' - Breakthrough or beginner
 
**'''A1''' - Breakthrough or beginner
Line 18: Line 18:
 
{|border="1" align="center" cellpadding="5"
 
{|border="1" align="center" cellpadding="5"
 
!Level
 
!Level
−
!UNL-NL Dictionary<br />(entries)
+
!Dictionary<br />(base forms)
−
!NL-UNL Dictionary<br />(entries)
+
!Grammar
−
!UNL-NL Grammar<br />(sentences)
+
−
!NL-UNL Grammar<br />(sentences)
+
 
|-
 
|-
 
|align="center"|A1
 
|align="center"|A1
−
|align="center"|2,000
+
|align="center"|5,000
−
|align="center"|2,000
+
|align="center"|Morphology: NP
−
|align="center"|RC-A1
+
−
|align="center"|500
+
 
|-
 
|-
 
|align="center"|A2
 
|align="center"|A2
−
|align="center"|5,000
+
|align="center"|10,000
−
|align="center"|5,000
+
|align="center"|Morphology: others
−
|align="center"|RC-A2
+
−
|align="center"|1,000
+
 
|-
 
|-
 
|align="center"|B1
 
|align="center"|B1
−
|align="center"|10,000
+
|align="center"|20,000
−
|align="center"|10,000
+
|align="center"|Syntax: NP
−
|align="center"|RC-B1
+
−
|align="center"|2,000
+
 
|-
 
|-
 
|align="center"|B2
 
|align="center"|B2
−
|align="center"|20,000
+
|align="center"|40,000
−
|align="center"|20,000
+
|align="center"|Syntax: VP
−
|align="center"|RC-B2
+
−
|align="center"|3,000
+
 
|-
 
|-
 
|align="center"|C1
 
|align="center"|C1
−
|align="center"|35,000
+
|align="center"|70,000
−
|align="center"|35,000
+
|align="center"|Syntax: IP
−
|align="center"|RC-C1
+
−
|align="center"|5,000
+
 
|-
 
|-
 
|align="center"|C2
 
|align="center"|C2
−
|align="center"|50,000
+
|align="center"|100,000
−
|align="center"|50,000
+
|align="center"|Syntax: CP
−
|align="center"|RC-C2
+
−
|align="center"|8,000
+
 
|}
 
|}
  
−
Where:<br />
+
== Methodology ==
−
*The numbers are cumulative. The number of entries and sentences required for the level C2 include the entries and sentences of the lower levels (i.e., in order to go from A1 to A2, a language must provide 3,000 new dictionary entries for each dictionary and address new 500 sentences in each direction).
+
In order to achieve the results above, each language must observe the following procedures<ref>This flow is not observed in some special cases, such as Latin, for instance</ref>. Each level is a prerequisite for the next (A2 is not open before A1 is finished, and so on):
−
*UNL-NL Dictionary is the number of UW's addressed in the UNL-NL dictionary according to the frequency of use. For instance: in order to achieve the level A1, languages must have addressed the 2,000 most frequent UW's of the UNL Dictionary (i.e., they should have completed [[MIR|MIR A1]])
+
*Dictionary flow (dictionary projects):
−
*NL-UNL Dictionary is the number of natural language lemmas addressed in the NL-UNL dictionary according to the frequency of use. For instance: in order to achieve the level A1, languages must have addressed their 2,000 most frequent lemmas (i.e., they should have completed [[BRUNO|BRUNO A1]])
+
**A1: [[MIR|MIR-A1]] > [[NADIA|NADIA-A1]] > [[BRUNO|BRUNO-A1]]
−
*UNL-NL Grammar is the number of UNL sentences of the [[Corpus|UNL Reference Corpus]] that the languages are able to generate. For instance: in order to achieve the level A1, languages must have succeeded in generating the sentences from the corpus [[RC-A1]] with [[EUGENE]].
+
**A2: [[MIR|MIR-A2]] > [[NADIA|NADIA-A2]] > [[BRUNO|BRUNO-A2]]
−
*NL-UNL Grammar is the number of natural language sentences of the [[NL Reference Corpus]] that the languages are able to analyze. For instance: in order to achieve the level A1, languages must have succeeded in analyzing the sentences from the NL Corpus A1 with [[IAN]].
+
**B1: [[MIR|MIR-B1]] > [[NADIA|NADIA-B1]] > [[BRUNO|BRUNO-B1]]
−
 
+
**B2: [[MIR|MIR-B2]] > [[NADIA|NADIA-B2]] > [[BRUNO|BRUNO-B2]]
−
== Assessment ==
+
**C1: [[MIR|MIR-C1]] > [[NADIA|NADIA-C1]] > [[BRUNO|BRUNO-C1]]
−
A '''dictionary entry''' is considered valid if verified by at least one editor.<br />
+
**C2: [[MIR|MIR-C2]] > [[NADIA|NADIA-C2]] > [[BRUNO|BRUNO-C2]]
−
A '''grammar''' is considered valid if its [[F-measure]] for the reference corpus is equal or higher than 0.8.
+
*Grammar flow (corpus projects):
 +
**A1: [[UGO|UGO-A1]] > [[CORNELIA|CORNELIA-A1]]
 +
**A2: [[UGO|UGO-A2]] > [[CORNELIA|CORNELIA-A2]]
 +
**B1: [[UGO|UGO-B1]] > [[CORNELIA|CORNELIA-B1]]
 +
**B2: [[UGO|UGO-B2]] > [[CORNELIA|CORNELIA-B2]]
 +
**C1: [[UGO|UGO-C1]] > [[CORNELIA|CORNELIA-C1]]
 +
**C2: [[UGO|UGO-C2]] > [[CORNELIA|CORNELIA-C2]]

Latest revision as of 15:49, 20 February 2014

FoR-UNL (Framework of Reference for UNL) is a guideline used to describe achievements of natural languages in relation to UNL. It was inspired by the Common European Framework of Reference for Languages (CEFR), and its main goal is to provide a method for assessing the availability and quality of natural language resources inside the UNL System.

Reference Levels

The FoR-UNL classifies languages in three broad divisions which can be divided into six levels, according to the recall and precision of the corresponding resources:

  • A - Basic Level
    • A1 - Breakthrough or beginner
    • A2 - Waystage or elementary
  • B - Intermediate Level
    • B1 - Threshold or intermediate
    • B2 - Vantage or upper intermediate
  • C - Advanced Level
    • C1 - Effective Operational
    • C2 - Mastery

Descriptors

The descriptors below inform what is required for a language to be classified in each level:

Level Dictionary
(base forms) 		Grammar
A1 5,000 Morphology: NP
A2 10,000 Morphology: others
B1 20,000 Syntax: NP
B2 40,000 Syntax: VP
C1 70,000 Syntax: IP
C2 100,000 Syntax: CP

Methodology

In order to achieve the results above, each language must observe the following procedures[1]. Each level is a prerequisite for the next (A2 is not open before A1 is finished, and so on):

Retrieved from "http://unlweb.net/wiki/index.php?title=FoR-UNL&oldid=7550"
Software