Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020
19667
Archives Available @ www.solidstatetechnology.us
Data-driven
UABI
Business-driven
Profile
Product
Market
Organiztion
Environment
Storage
Retrieval
Analysis
Representation
Figure 2 UABI orientations
UABI model acts as a parallel sustainable system(multi-agent system for example) that implements the
data-driven aspect of the model to handle operations on the business-driven aspect. This proposes the four
basic abstract concepts of the model are (Storage, Retriever, Analyzer, and Exhibitor). These concepts are to
be implemented by the software objects like declaring a storage object for DBMS, OLAP, or Data Warehouse
used in the application. If OLAP is used, it would also implement an analyzer. As our model is a
programming model, we will talk at the programming level and consider the four basic concepts as four
abstract objects.
In addition to the four basic abstract objects declared above, UABI model contains an Information object
represents the details about the business-driven aspect to be handled by the data-driven objects. This
information object has the following attributes: topic, content,keywords, value, and co-occurrence weights.
The topic is the identifier semantic noun that uniquely and fully refers to the content. For other information
objects, the topic might be used as a keyword, but never as a topic but once it is declared. The content is the
long string collected from the resources. The keywords list is a list of topics figured out from the content.
Each keyword has a weight between 1 and -1. Where the Keywords group is K and the weights are W, the
value of each topic is v(k) and the value V of a topic is calculated as:
As we notice, calculating the value attribute involves the values of other topics. As the model needs to use
the calculated value attribute iteratively and aggregately, it is very efficient to keep the value result saved
after calculation and recalculated upon updates. This will dramatically increase the performance efficiency
and reduce retrieving the value cost from
to
.
A language processing algorithm figures the keywords throughout the content. Theoretically, every
syntactical noun is a keyword. However, a relevance algorithm navigates the context to define the relation
between the topic and the keywords. The software might use Natural Language Processing (NLP) algorithms
for this task. A reinforcement technique should be used to analyze the context and figure how much a
keyword is mentioned and define its responserelation as negative or positive. Each occurrence adds 1 to the
probability of the relation. The weight is a max-min score between the highest occurrence and the mean. All
occurrences bellow the mean are ignored.So, where keyword occurrence is c, the absolute weight W is:
The sign is determinedthrow the context if it is negative or positive. Moreover, the weighting process is a
two-way relation. This means if a keyword is related to a topic throw some weight, the topic also is a
Do'stlaringiz bilan baham: |