Asterisk

Tools for computing asset risk with respect to goals.

Today advanced decision-making and artificial intelligence techniques are used primarily by financial and security industries. This is an attempt to create a tool that would be useful to apply the advanced decision-making techniques for people's personal lives, with respect to any goals, and any assets. We learned math at school, so let's apply it not only at work, but also at home.

pip install asterisk                                         # only Python2.7 for now

Asterisk implements a wrapper on top of the WikiData API, as a source for schemas of objects, so that we could automatically have multilingual, well-defined concepts, with respect to which to define assets. Then, it will use the functionality of packages stepio and plandf, and more, to provide the means to create computable, multilingual, hierarchial plans with respect to goals defined as concepts, and in terms of concepts. Lookup and create multilingual concepts on www.wikidata.org, use them here. For example, a concept of a loan is Q189539. Edit the P1552 property ("has quality") to define the schema attributes. Ideally, we will define the goals in terms of quality ranges, however Asterisk provides a means to define the extra attributes freely by passing dicts, because often concepts on WikiData do not yet have standardized set of qualities, and your personal plans or personal concepts may be non-standard.

Examples

When instantiating an item from schema, by default, the .fact=False, and the instances are prepended with asterisk *, meaning that it is something hypothesized, not really existing. If the thing that you are instantiating actually exists in real life (e.g., a specific human, a specific loan), pass the .fact=True (and you can also add attribute such as Lithuania's republic passport ID, or Bondora loan ID, or the ID in ANY OTHER DATABASE to provide actual reference to identify with real thing.) The real things are prepended with dot . rather than asterisk.

You can define a goal as a set of assets within an output of a step, part of plan. At the moment. the .Plan does not yet work, but you can already instantiate lists like:

import asterisk as rx

# Concepts
Dog = rx.Concept('Q144')
Cat = rx.Concept('Q146')
Human = rx.Concept({'en': 'human', 'lt': u'žmogus', 'zh': u'人'})
Family = rx.Concept({'en': 'family', 'de': 'Familie', 'lt': 'šeima'})
Dollar = rx.Concept({'en': 'US Dollar'})

# Instances (Assets)
alice = Human({'name': 'Alice', 'legs': 2, 'person_id': 'LT123148188101041041'}, fact=True)
bob = Human({'name': 'Bob', 'legs': 2, 'person_id': 'EE18416109410238120'}, fact=True)
dog = Dog({'name': 'Tako', 'legs': 4, 'owner': bob})
cat = Cat({'legs': 4, 'owner': alice})
family = Family({'name': "Alice and Bob's family", 'mission': "To enjoy each other's minds, activities together, and exploration of the Universe."})
dollar = Dollar()

# List of Steps
[
    {'input': {alice: 1, bob: 1},
    'output': {family: 1}},
    
    {'input': {dollar: 200},
    'output': {cat: 1}},

    {'input': {dollar: 250},
    'output': {cat: 1}}
]

Classes defined with "Q" retrieve and map schemas from WikiData (e.g., Q144 is a dog, and Q146 is a cat), while the ones defined without "Q" with multi-lingual dictionaries instead, are not mapped to actual concepts, and are provided as a "last resort", when you only need a multilingual word, so that you could display the plans for different audiences in their own language. Using "Q" automatically loads the language aliases in all available languages on WikiData, so you don't have to care about writing out all aliases. Instantiating from WikiData also provides much more information, like the definitions of the concept in many languages, relations, etc.

Future:

It would not be hard to enable this tool to optimize for everything. For example:

• Need to optimize your social network feed for messags of your interest? Each message is an asset., so, can be optimized.

• Need optimize for food price / quality, and you have a link to a supermarket website with products? Each food item is an asset.

• Need to find friends of common interests? Each contact in your social networks is a potential friend. Each can be treated as an asset*

• Need to find an interesting job? Each job offer in freelancing platforms is an asset*

Things of higher hierarchical complexity should also be possible. For example,

• What events to attend, based on topic, location, schedule.

And so on, including, of course, what securities to buy (any kind of portfolio optimization).

• Choose Any Action.

Hierarchical Multilingual Plans

# Future: Multilingual plans with respect to defined concepts, display-able and computable...
plan = rx.Plan(
    [
        {'input': {alice: 1, bob: 1},
        'output': {family: 1}},
        
        {'input': [
            {'input': {dollar: 200},
            'output': {cat: 1}},

            {'input': {dollar: 250},
            'output': {cat: 1}}
            ],
         'output': '*'
         }
    ]
)

plan.risk()

Read the intial thoughts on Asterisk, and the follow-up thoughts.