Back to main page
What are Aito's core concepts?
Aito is an synthesis of a database and AI technology, and it introduces concepts and idea
from both of these realms. Some of the database concepts and AI concepts have different names,
yet are conceptually very similar or the same.
In this chapter, we introduce the core concepts of the Aito intelligence layer.
The basic concepts relating to the data are the following:
Table. Aito is a database-like entity, that stores its data in relational tables consisting of rows and
columns, a bit like an SQL database.
Schema. Each table has a strict schema, as in an SQL database. The schema contains the column names and
value types, plus additional meta information about linking, and the analysis
of column values.
Values. The table - of course - contains values at the intersection of each row and column. Strings, Integers, Long integers and double precision floating points are supported, as well as their optional versions.
Analyzers and features. Aito does not do the statistical reasoning at the level of values, but at reasons at the level of 'binary features'.
- Specialized analyzers are used to decompose potentially large values (e.g.
long text values) into individual binary features.
- For example, if field 'age' has a value 54, this can be turned into feature
age:54. If a field 'text' value 'horses are running'
is analyzed with English language analyzer, it can be turned into features
text:'hors' and text:'run'.
The analyzers do not only split the text into words, but it can also drop
overly common words, and reduce the words into their normal forms.
- Analyzers may be familiar from search engines like Lucene or its derivatives
like ElasticSearch and Solr.
- While the main data is stored in a table, the features are stored in a separate
feature data frame
The concepts related to the use of Aito are the following:
Context. All Aito's operations are done in a context. Context defines a scenario, where there are knowns and unknowns.
- Context has always some examined 'root table'. For example: this table may
be the impression table.
- Context has some knowns. For example, in the impression context, we may know
and we may know the page the user is int.
- The context often has unknowns, which are not specified. For example, it may
not be known, whether
the displayed content was clicked.
Relation. In Aito, relation refers to the statistical examination of a set of features.
- For example, the relation (name:'playstation', productCategory:electronics')
is used to
examine playstation's statistical relationship with the 'electronics' product
- Also, the relation (user:'bob', product.category:'laptop', click:true) can
be used to
examine Bob's preference for laptops.
- Aito often examines hundreds or thousands of relationships per query to find
statistical relationships that may help to provide results.