Grass&Green: with more entities to enhance the classification

Posted by grass_and_green on 9 February 2016 in English (English)

Hi mappers,

OSM, as a VGI mapping project, inherits various challenges: naive contributors, flexible contribution mechanism, and uncertainty of spatial data. The facts that rise subjective classification problems in the resulting data. Whether a piece of land covered by grass is classified as “park”, “garden”, “meadow”, or “forest”; whether a water body is classified as “pond”, “lake”, or “reservoir”. In OSM project, such of these classification answers are likely dependent on contributors’ perceptions. However, the appropriate classification of entity is strongly related to some qualitative observations and quantitative measures.

Thus, Grass&Green is a tool that has been developed to help contributors to assign the most appropriate classification of some entities of grass-related features. The role of contributors is needed not only in adding new data, but also to revise and confirm the existence one.

• The tool has an access and contribution to users accounts

Alt text

• It has very supportive and easy to use interface

Alt text

On the right hand, it presents the entity under investigation with some qualitative descriptions of its context. On the left hand, it show the OSM entity combined with some recommendations and suggestion, with full flexibility to update/confirm/reject these given recommendations.

• The tool provides the user with textual and visual descriptions of the target features and similar identical features.
Alt text

Example of classification improvements:

o The next was only grass with fence, the contributors with our tool makes it “Garden” which is more expressive information than grass with fence.

Alt text

o The next was labeled as “Park”, however its characteristics is far from being park for amusements and entertainment. Out tool correct it to the reasonable classification as “Garden”

Alt text

The status of contributions: We have around 250 participants from more over that 30 counties. They checked around 3000 entities in about 3 months. However, we are seeking for more confirmation and rejections. The data need to be check and we adapt the methodology of crowd-sourcing revision.

Alt text

Alt text

They are mostly agree or partial agree with our recommendations and working to improve the classification quality of these features.

So, please feel free to contribute and follow us on Twitter and Facebook. Send us your feedbacks. Every small contribution even for one entity would be appreciated.

Comment from Warin61 on 9 February 2016 at 22:19

‘Forest’ has different meaning for different people!!!

To me it means ‘trees’ that are intended to produce products e.g. wooden planks, beams (‘lumber’) wood pulp (for paper) etc.

Another existing tag is ‘natural=wood’ … that meaning has migrated away from my understanding of natural to include ‘artificial’. The key ‘natural’ is poorly though out .. has never been formally discussed on the tagging group and should have the status ‘inuse’ at best. Hopefully this will eventually be replaced with something like ‘landcover=trees’

Comment from BushmanK on 9 February 2016 at 23:45

Idea of this tool is just great.

But unfortunately, it is incapable to solve the problem of ambiguous definitions.

You are suggesting landuse=forest while there is natural=wood, as it was mentioned above. To be able to use this tool correctly, there should be clear consensus on exact meaning (and, therefore, on distinctive features) of landuse=forest, natural=wood, landuse=grass, landuse=meadow, natural=grassland. So, it seems like it’s just someone’s willful choice.

And what the hell is landuse_1=*?

Also, these tags have “compound” meaning, combining description of floral community and its usage.

I have a diary entry on this topic:

Comment from BushmanK on 10 February 2016 at 02:41

Also, referring to Wikipedia and definitions from the dictionaries is completely wrong since OSM does not use natural language to describe objects - these words, used for keys and values are borrowed from the English language, however, in OSM they should be used as special terms, which sometimes have a bit different (more strict or more broad) definition. So, semantically, this work could be making similar mess as there was before.

Login to leave a comment