taarifa / TaarifaWaterpoints

Waterpoint management system for Tanzania
http://taarifa.org
Other
19 stars 21 forks source link

Water Point Identification #15

Open willowbl00 opened 10 years ago

willowbl00 commented 10 years ago

Each water point has an official unique identifier, used by the Ministry of Water, and expected to be known by the water engineers. This is not reasonable to ask people to input via text or any other method - the number is long, often missing visually, and could be mistyped easily.

The same dataset has a few complimentary ways of indicating a water point's location - basin (not useful), region, LGA (local gov authority) (could be district or city), ward, sub village, water point name.

By using USSD (thanks, Rav!), we could nest menus leading people from "Where do you live" to narrow down until they have one choice between water point names. Or start with the water point names, and deconflate based on responding to further whereabouts. This becomes more complex (and costly) using SMS. Jeremy and Mark were working on 3-word indicators for any water point. Much of this can be handled via geolocation on smart phones where available.

Both of these paths need at least SOME level of fuzzy matching. There are many spellings of each way of referencing, so if we're asking for free-form text, the matching has to occur on our end, and should assume imprecision.

kynan commented 10 years ago

Could anyone elaborate on those 3-word indicators? How do they work? How are they derived? Is this an automated process or does someone need to assign an indicator to each waterpoint individually?

Bericender commented 10 years ago

We can`t use USSD until we have a shortcode number in Tanzania. Most people will most likely not have smart phones either, so geotagged SMS are not ideal either.

For the moment I will implement it using the water point name, while giving them a confirmation text including LGA, ward and sub village

jmorley commented 10 years ago

"Jeremy and Mark were working on 3-word indicators for any water point." - I'm not sure we ever were?

We're still discussing this. An option suggested by JP would be to use the closest mobile-money point id (supposedly 5-digits but possibly needing network ID too; widely used) as an alternative identifier.

My issue is that we don't want to invent and proliferate unique IDs. I'd suggest we stick with the long identifier at least for the demos. The SMS/input methods should have a graceful fallback for "unknown" ID (not least because many points don't have an identified written on them). In that case some informal identifier (such as the "Where do you live?" response), essentially free text, would be needed. We then review on the basis of the trip to Iringa.

willowbl00 commented 10 years ago

I think we're going to lose our audience immediately if we try to use the long ID. It is a technologist's solution. NAME + LOCATION is a reasonable stopgap in Iringa, until we have to worry about scaling, via USSD and shortcodes.

After re-connecting and uploading the information, if that NAME + LOCATION doesn't exist in the database (so it can be linked to the UID), add it. The water engineers who know the area will be able to deconflict it.

jangojarango commented 10 years ago

Regarding the 3-word location indicators, it was Gary and Mark having a conversation about http://what3words.com/ , which does sound quite intriguing – though I only glanced at it and don’t know how much it would cost…

dgorissen commented 10 years ago

@Bericender any update on this?