Comments and suggestions on the WM ontology v2.1

[lgalescu]

terminology on the way I understand how this ontology is structured (by levels):

1. root: wm_compositional
   • suggest changing to wm
2. category: fixed set of top-level types [concept, entity, process, property, time]
   • the /entity/ category is a hodge-podge; i see no objective reason for them not to be moved wholesale under /concept/
   • ontologists use "concept" and "entity" generically, to apply to all things; i would recommend renaming concept/ to /object/, which is more widely accepted usage
3. domain: only used (somewhat) systematically for the /concept/ sub-hierarchy
   • there are a few cases of use in the /process/ hierarchy
4. the lower levels form an is-a hierarchy for the respective domain/process
   • at least this seems to be the general idea; exceptions are probably incidental

general observations:

• examples from metadata are a huge mess!
• we probably need a lot more properties -- often it is the modifier that conveys the critical info (eg, "high temperatures" might reduce yields, not just "temperature"; or "hybrid seeds" increase yields, not simply "seeds").

notes on specific nodes:

wm_compositional/concept/agriculture/disease/crop_disease wm_compositional/concept/agriculture/disease/livestock_disease

• remove (compositional) -- /health/disease should apply to crops, livestock, people, children, whatever (i am not entirely sure how

wm_compositional/concept/agriculture/pest/locust/solitary/solitary_hopper wm_compositional/concept/agriculture/pest/locust/solitary/solitary_locust wm_compositional/concept/agriculture/pest/locust/gregarious/hopper_band wm_compositional/concept/agriculture/pest/locust/gregarious/locust_swarm

• for these four, it seems the intention was to get combos of behavior (solitary vs gregarious) with phase (immature vs mature). but the naming is misleading.
• might want to look at grounding data to see if all these are useful or mere hair-splitting. if the latter, nodes should be combined.
• if all the distinctions are desirable, i suggest renaming for clarity, using separate upper nodes for individuals and swarms/bands: locust/solitary/solitary_hopper => locust/immature_locust locust/solitary/solitary_locust => locust/mature_locust locust/gregarious/hopper_band => locust_swarm/immature_locust_swarm (or _band) locust/gregarious/locust_swarm => locust_swarm/mature_locust_swarm

wm_compositional/concept/crisis_or_disaster/conflict/crime wm_compositional/concept/crisis_or_disaster/conflict/tension

• crime is not conflict; tension is typically not a kind of conflict
• neither of them rises to the level of crisis/disaster
• maybe create a "social" domain and put there things like crime, tension, corruption?

wm_compositional/concept/crisis_or_disaster/disease

• rename to epidemic

wm_compositional/concept/crisis_or_disaster/depression_recession

• rename to recession -- a depression is a long recession

wm_compositional/concept/crisis_or_disaster/evacuation

• move under /process/

wm_compositional/concept/crisis_or_disaster/taskforce

• doesn't belong in the crisis/disaster domain
• maybe move under /entity/person_or_group?

wm_compositional/concept/economy/domestic_market

• rename to market or commerce (n.b.: there is no international_market!)

wm_compositional/concept/economy/livelihoods/rural_livelihoods wm_compositional/concept/economy/livelihoods/urban_livelihoods

• rename livelihoods => livelihood
• is the rural/urban distinction reflected in the data?

wm_compositional/concept/environment/meteorology/precipitation wm_compositional/concept/environment/meteorology/temperature

• seems to me that temperature is a property of weather -- thoughts?

wm_compositional/concept/environment/meteorology/weather

• i would replace meteorology with climate in all the above; or just delete

wm_compositional/concept/health/nutrition/malnutrition

• maybe move under /disease/ ?

wm_compositional/concept/health/welfare

• isn't welfare more general than just health?

wm_compositional/concept/infrastructure/construction_materials

• i think this fits better under /goods/ (they're consumables, not infrastructure)

wm_compositional/concept/infrastructure/transportation_infrastructure/road_infrastructure

• too much infrastructure!!!
• rename transportation_infrastructure => transportation
• rename road_infrastructure => roads

wm_compositional/concept/infrastructure/transportation_infrastructure/rail_infrastructure

• rename rail_infrastructure => railroads

wm_compositional/concept/infrastructure/WASH/sanitation_infrastructure wm_compositional/concept/infrastructure/WASH/water_facilities

• remove WASH node

wm_compositional/concept/migration/human_migration

• rename to refugees

wm_compositional/concept/migration/IDP wm_compositional/concept/migration/returnees

• should these three all go under person_or_group?

wm_compositional/concept/physical_market

• yuck! whoever calls a market physical?!? maybe rename to marketplace?
• also, it's funny that this appears all alone; maybe move under /economy/ ?

wm_compositional/concept/population/demographics/age wm_compositional/concept/population/demographics/gender wm_compositional/concept/population/demographics/race_or_ethnicity

• remove demographics node

wm_compositional/concept/service/information/surveillance_system

• service only appears in this path -- move concept under /security/?

wm_compositional/concept/storage/agricultural_storage

• move under /infrastructure/ (maybe under farm_infrastructure?)

wm_compositional/concept/storage/general_storage_facilities

• move under /infrastructure/

wm_compositional/concept/storage/household_food_storage

• very specific! any evidence from grounding data that it is useful?

wm_compositional/concept/storage/stockpile_goods

• odd node! any evidence from grounding data that it is useful?

wm_compositional/concept/trend

• i'm sure this gets grounded, but is it useful? a trend is a direction of change, and those changes ought to be captured by increase/decrease statements

wm_compositional/concept/water

• move under /goods/ ( with food, fuel, etc.)

wm_compositional/concept/workforce

• move to the /economy/ domain

wm_compositional/process/depreciate

• there is no opposite process for appreciation in value

wm_compositional/process/intake

• is this for food intake? if so, rename to food_intake to make it explicit

wm_compositional/process/population/emigrate wm_compositional/process/population/immigrate

• these two should be moved under /migrate/

wm_compositional/process/train_or_educate/

• rename train_or_educate => e.g., teach (or just train)

wm_compositional/process/train_or_educate/humanitarian_training/emergency_preparedness_training

• do we expect other subtypes of /humanitarian_training/? otherwise remove

wm_compositional/property/insecurity

• remove ; should represent as (security :polarity -)

wm_compositional/property/instability

• remove ; should represent as (stability :polarity -)

wm_compositional/property/nonutilization

• remove ; should represent as (utilization :polarity -)

wm_compositional/property/unavailability

• remove ; should represent as (availability :polarity -)

wm_compositional/entity/animal

• what is this for? /concept/agriculture/livestock ought to capture some animals -- is this for everything else? it would be strange to have two subsets of a broad category in two different parts of the hierarchy

wm_compositional/entity/artifact

• what is this for? /concept/goods ought to capture lots of artifacts

wm_compositional/entity/geo-location

• how about just location?

wm_compositional/entity/health_worker

• a health worker is a person, so it probably should go under person_or_group

wm_compositional/entity/person_or_group

• wouldn't people be more economical?

wm_compositional/time/season

• is this really the only temporal concept we need??

[MihaiSurdeanu]

Wow! These are great @lgalescu ! Thank you for the effort!

@zupon, @BeckySharp: can you please review and incorporate these changes?

[MihaiSurdeanu]

@zupon, @BeckySharp: we can discuss the more difficult ones at the WM project meeting.

[zupon]

I can get started on the easier ones now, and we can definitely discuss the hard cases at the WM project meeting.

[zupon]

Many of the simpler changes are covered by #99, but there are some specific "hard" cases we wanted to discuss before changing so we don't inadvertently mess something up for someone.

1. Moving entity/ and/or time/ inside the concept/ category. I'm told that BBN might treat these differently, so there may be reason to keep them separate. @chanys does that sound correct?
2. Eliminating redundancy in e.g. infrastructure/transportation_infrastructure/road_infrastructure/. Given the whole ontology path, there's a lot of extra infrastructure present. If we only see the intermediate node or leaf node (e.g. just transportation or just roads, per the suggestions), do we lose important information?
3. Things like crime and tension don't really fit in the crisis_or_disaster domain. Should we create a new domain like social and move them there, or rename crisis_and_disaster to be more inclusive of things like crime and tension?

Some smaller cases to decide on:

• Should we replace meteorology domain node with climate, delete that intermediate node, or leave it as-is?
• Move malnutrition from nutrition domain to disease domain?
• Keep entity/animal node? Is it different enough from concept/agricultural/livestock?
• Keep entity/artifact, or delete and rely on existing goods node?
• In storage domain, are nodes like stockpile_goods and household_food_storage useful?

[chanys]

On Q1: It is fine to move entity and time under concept. We don't need them under their own distinct branch. We just need them to be represented somewhere in the ontology, so that we can represent the Event-participants and Event-time.

On Q2:

• When doing renaming, ought to ensure that we don't introduce duplicate node names.
• BBN (at least) does not rely on the names of each node to do grounding. But we do rely on the meta-keywords under each node to some extent. So making a branch shorter, doesn't really affect us, so long as the meta-keywords under that node moves along as well.

[zupon]

Closed by #99