SE-MSLC: Semantic Entropy-Driven Keyword Analysis and Multi-Stage Logical Combination Recall for Search Engine

. 2025 Sep 16;27(9):961. doi: 10.3390/e27090961

Algorithm 2 SE-KIA (online component)

1:
Input: User query $q = {k_{1}, k_{2}, \dots, k_{n}}$ , keyword co-occurrence word graph $G = (V, E)$ , weight factor $▵ w e i g h t = 2$ , POS weight table $W$
2:
Output: Keyword weights ${w_{1}, w_{2}, \dots, w_{n}}$
3:
Step 1: Static Weight Assignment
4:
for each keyword $k_{i} \in q$ do
5:
Retrieve part-of-speech tag $POS (k_{i})$
6:
Assign static weight: $w_{i} \leftarrow W [POS (k_{i})]$
7:
end for
8:
Step 2: Dynamic Weight Adjustment
9:
Generate candidate pairs: $P \leftarrow {(k_{i}, k_{j}) ∣ 1 \leq i < j \leq n}$
10:
for each pair $(k_{i}, k_{j}) \in P$ do
11:
if $(k_{i}, k_{j})$ exists in G then
12:
Retrieve precomputed relative importance $w_{i j} \leftarrow G . get_importance (k_{i}, k_{j})$
13:
if $w_{i j} = 1$ and $w_{i} < w_{j}$ then
14:
$w_{i} = w_{j} + ▵ w e i g h t$
15:
else if $w_{i j} = - 1$ and $w_{i} > w_{j}$ then
16:
$w_{j} = w_{i} + ▵ w e i g h t$
17:
end if
18:
end if
19:
end for
20:
Step 3: New Queries Logging
21:
Insert q into log database $L$
22:
Periodically (e.g., daily): extract new pairs $\tilde{P}$ from $L$ with frequency $f_{i j} \geq τ$
23:
Update G with $\tilde{P}$
24:
return ${w_{1}, w_{2}, \dots, w_{n}}$