Solid State Technology
Volume: 63 Issue: 6
Publication Year: 2020
19669
Archives Available @ www.solidstatetechnology.us
weight is added to each topic defines the relation between them as the average of the averages of the shared
weights according to the following equation where w is the weight to calculate, t
1
is the first topic value, t
2
is
the second topic value, t
1
w
i
is the weight of a shared keyword in the first topic, t
2
w
i
is the corresponding
weight in the other topic, and n is the number of the shared topics:
If the Analyzer could not find any shared keywords, the weight is simply set to 0. This means the two
topics contextually are unrelated. Hence, the Analyzer calculates its initial value as the average of the weights
multiplied by their values. Where n is the count of the weights, the value v is:
If the value is 0, then all the topics involved in the case are unrelated, and the case is rejected.
The next step is to define the significance threshold. Statistically speaking, 0.05 and 0.01 the common
significance threshold used. However, the case is different here. It is natural that new cases would have low
values. The significance threshold should be determined by that value and the value of the topics concerned
as the minimum outliers' limit. This significance value will determine which topic is related to the case or not.
The significance of each topic is calculated as its value multiplied by its weight to its parent multiplied by its
parent's significance. Where the topic is t, the value is v, its weight to its parent is w, and its parent is p, the
significance S is:
The significance of the case itself is always 1. The analyzer would start by assigning the significance of
the topics of the case. If any of the topics is insignificant, this means the case itself does not make sense as
one of the topics involved does not belong to the context of the case. Hence, the Analyzer will proceed with
neglecting to analyze the insignificant topic. Otherwise, the Analyzer calculates the significance of the
keywords of the topic and consider them as related topics and proceeds to their keywords or unrelated topics
and neglects them accordingly.After establishing the tree, the Analyzer will iterate updating the weights of
the topics until the difference between the iteration is insignificant.
The next step is to determine the shared keywords and combine them. If a topic is significant and is a
keyword for more than one topic, it is a shared keyword topic. It will have more than one significance value
corresponding to each topic it is a keyword for it. The significance of the topic is the sum of the calculated
significances. Then, the Analyzer replaces the significance values by their probability. Then, the Analyzer
recalculates the case value as a sum of products of the topic values multiplied to their significance values.
Where the significance probability is s, the topic value is t, and the number of topics in the tree is n, the case
value v is:
The value of the case represents the state of this case. Then, the significance threshold is updated again. If
the difference between the new value and the old value is higher than the significance threshold, the Analyzer
processes the model again based on the new base value and the new significance threshold.
The Analyzer recognizes four types of topics in the tree. The first type is the case topics mentioned in the
case. The second type is the shared topics which are keywords of more than one topic. The third type is the
Direct Influence Topics DIT which are revealed in the keywords of all the case topics. The fourth type is
Indirect Influence Topics IIT which are revealed only once.
Do'stlaringiz bilan baham: |