I want to create a walkability network for the city of Dortmund such that it includes all the line features that are walkable like footways, pedestrian streets etc. I downloaded OSM shape file from Geofarbrik website and now I see that there are a lot of inconsistencies in the roads data. Lots of streets are not classified as highway or footway or anything. Lots of footways are not present at all. Names of streets are missing and many more things like that. Now ofcourse I am aware that this is a consequence of using open source data and we have to deal with such inconsistencies that are a part of publicly sourced data almost always. But I just need to know if there is a better way to create a such a walkability network which could achieve some standard of accuracy rather than now editing all of the data by myself by adding or correcting all walkable polylines in this network using may be google maps or something to see if there is actually a footway in a particular street or not or if its walkable or not. That is gonna be one hell of a work. Phew!
I have found OSM to be the best available free dataset for pedestrian network analysis, but it does require cleaning first. Generally an OSM network requires at the very least planarized lines where there is not a "tunnel" or "bridge" tag (it does not go over or underground- you don't want your trail under a highway to have connectivity to an interchange!). It also usually requires speeds to be imputed off functional class. If your functional class field from OSM is not very dependable, I would suggest not using it. However, if OSM is the best available, maybe try to compare it to another series of datasets?
If you can't find a good dataset to compare against in centerline form, you might be able to use more aggregate geometries to prioritize which parts of the network you QAQC. This could be demographic data sets, or potentially amenity datasets from OSM so that you can prioritize connections around restaurants and other likely destinations that people might want to walk to.
Others might have better advice.
If you use the ArcGIS Editor for OpenStreetMap | Overview to download the data, the streets data will be much more viable for a network dataset. In fact, that toolbox even has a tool in it to create the network dataset for you. There's no tech support for the Editor for OpenStreetMap, but it should be fairly straightforward to use and I think it will get you a lot closer to what you want.
This is a very useful tool for working with OSM data, and it also have some decent handling of some of the odd network classifications you are seeing now in your network. I would give this a try, but keep in mind you still might have to do some QAQC still if the underlying data and its functional classifications are a concern.