Granted Patent B2

US 12,634,336 · App. 18/736,735 · Granted May 19, 2026

Inventor: Chih-Ming Chen (New Taipei City, TW)

Assignee: WISTRON CORPORATION

H04L63/1441H04L51/02

Abstract

Claims (61)

1 . An artificial intelligence frontend system, executed on an electronic device, comprising an input frontend and an output frontend, the input frontend comprising an input filter element, the output frontend comprises an output filter element, wherein

the input frontend is configured to execute the following steps:

(a) execute an external clarification procedure in response to receiving a query to obtain an external clarification query, and direct the external clarification query to the input filter element;

(b) execute, via the input filter element, a filtering procedure on the external clarification query to filter the external clarification query, and send, via the input filter element, the external clarification query which is filtered to an external chatbot;

(c) in response to the chatbot raising an internal clarification question, replying to the internal clarification question; and

(d) in response to the chatbot raising a final question, redirecting an internal clarification query to the chatbot; and

the output frontend is configured to execute the following steps after step (d);

(e) receiving, via the output filter element of the output frontend ( 301 ), at least one chat answer replied by the chatbot to the internal clarification query; and

(f) finding, by the output frontend, a best chat answer from the at least one chat answer; wherein step (f) comprises the following steps:

(f1) based on a coherence and clarity algorithm, evaluating coherence and clarity of each of the at least one chat answer to obtain a coherence and clarity score for each of the at least one chat answer;

(f2) evaluating accuracy of each of the at least one chat answer to obtain an accuracy score for each of the at least one chat answer;

(f3) based on a relevance algorithm, evaluating relevance between each of the at least one chat answer and the internal clarification query to obtain a relevance score for each of the at least one chat answer; and

(f4) based on the coherence and clarity score, the accuracy score, and the relevance score for each of the at least one chat answer, selecting the best chat answer from the at least one chat answer.

2 . The artificial intelligence frontend system according to claim 1 , wherein the external clarification procedure comprises the following steps:

(a1) setting the query as a current query;

(a2) judging whether the current query meets an explicit criterion or not;

(a3) in response to judging that the current query meets the explicit criterion, outputting the current query to serve as the external clarification query, and exiting the external clarification procedure;

(a4) in response to judging that the current query does not meet the explicit criterion, issuing an asked question based on the current query; and

(a5) in response to receiving a response corresponding to the asked question, setting the response as the current query, and then executing steps (a2) to (a5).

3 . The artificial intelligence frontend system according to claim 1 , wherein the output frontend is configured to execute the following step after step (f):

(g) executing, via the output filter element of the output frontend, a modification procedure on the best chat answer to modify the best chat answer.

4 . An artificial intelligence frontend operation method, suitable for an artificial intelligence frontend system, wherein the artificial intelligence frontend system comprises an input frontend and an output frontend, the input frontend comprises an input filter element, the output frontend comprises an output filter element, and the artificial intelligence frontend operation method comprises:

(a) executing, by the input frontend, an external clarification procedure in response to receiving a query to obtain an external clarification query, and directing the external clarification query to the input filter element;

(b) executing, by the input frontend via the input filter element, a filtering procedure on the external clarification query to filter the external clarification query, and sending, via the input filter element, the external clarification query which is filtered to an external chatbot;

(c) in response to the chatbot raising an internal clarification question, replying, by the input frontend, to the internal clarification question; and

(d) in response to the chatbot raising a final question, redirecting, by the input frontend an internal clarification query to the chatbot, by the input frontend;

(e) receiving, via the output filter element of the output frontend, at least one chat answer replied by the chatbot to the internal clarification query; and

(f) finding, by the output frontend, a best chat answer from the at least one chat answer; wherein step (f) comprises the following steps:

(f1) based on a coherence and clarity algorithm, evaluating coherence and clarity of each of the at least one chat answer to obtain a coherence and clarity score for each of the at least one chat answer;

(f2) evaluating accuracy of each of the at least one chat answer to obtain an accuracy score for each of the at least one chat answer;

(f3) based on a relevance algorithm, evaluating relevance between each of the at least one chat answer and the internal clarification query to obtain a relevance score for each of the at least one chat answer; and

(f4) based on the coherence and clarity score, the accuracy score, and the relevance score for each of the at least one chat answer, selecting the best chat answer from the at least one chat answer.

5 . The artificial intelligence frontend operation method according to claim 4 , wherein the external clarification procedure comprises the following steps:

(a1) setting the query as a current query;

(a2) judging whether the current query meets an explicit criterion or not;

(a3) in response to judging that the current query meets the explicit criterion, outputting the current query to serve as the external clarification query, and exiting the external clarification procedure;

(a4) in response to judging that the current query does not meet the explicit criterion, issuing an asked question based on the current query; and

(a5) in response to receiving a response corresponding to the asked question, setting the response as the current query, and then executing steps (a2) to (a5).

6 . The artificial intelligence frontend operation method according to claim 4 , wherein the artificial intelligence frontend operation method comprises executing the following step after step (f):

(g) executing, via the output filter element of the output frontend, a modification procedure on the best chat answer to modify the best chat answer.

7 . The artificial intelligence frontend operation method according to claim 4 , wherein the coherence and clarity algorithm comprises a neural local coherence analysis model.

8 . The artificial intelligence frontend operation method according to claim 4 , wherein the coherence and clarity algorithm comprises an automatic evaluation of text coherence algorithm.

9 . The artificial intelligence frontend operation method according to claim 4 , wherein step (f2) comprises the following steps:

(f21) finding, by a structural semantic similarity model, multiple pieces of data most relevant to the query from multiple trusted sources; and

(f22) based on these pieces of data, providing the accuracy score for each of the at least one chat answer.

10 . The artificial intelligence frontend operation method according to claim 4 , wherein the relevance algorithm comprises a structural semantic similarity model.

11 . The artificial intelligence frontend operation method according to claim 4 , wherein the filtering procedure comprises:

(b1) executing an input transformation on the external clarification query as a defense mechanism to eliminate adversarial perturbations.

12 . The artificial intelligence frontend operation method according to claim 11 , wherein the filtering procedure comprises based on the context of a dialogue, adding a more detailed query requirement into the external clarification query undergoing the input transformation.

13 . The artificial intelligence frontend operation method according to claim 11 , wherein the filtering procedure comprises: transforming the external clarification query undergoing the input transformation into an open question form.

14 . A non-transitory computer-readable recording medium with a stored program, wherein an artificial intelligence frontend system is formed at a logic level when a processing unit loads and executes the stored program, the artificial intelligence frontend system comprises an input frontend and an output frontend, and the input frontend comprises an input filter element, the output frontend comprises an output filter element, and executes the following operations:

(a) executing, by the input frontend, an external clarification procedure in response to receiving a query to obtain an external clarification query, and directing the external clarification query to the input filter element;

(b) executing, by the input frontend via the input filter element, a filtering procedure on the external clarification query to filter the external clarification query, and sending, via the input filter element, the external clarification query which is filtered to an external chatbot;

(c) in response to the chatbot raising an internal clarification question, replying, by the input frontend, to the internal clarification question; and

(d) in response to the chatbot raising a final question, redirecting, by the input frontend an internal clarification query to the chatbot, by the input frontend;

(e) receiving, via the output filter element of the output frontend, at least one chat answer replied by the chatbot to the internal clarification query; and

(f) finding, by the output frontend, a best chat answer from the at least one chat answer; wherein step (f) comprises the following steps:

(f1) based on a coherence and clarity algorithm, evaluating coherence and clarity of each of the at least one chat answer to obtain a coherence and clarity score for each of the at least one chat answer;

(f2) evaluating accuracy of each of the at least one chat answer to obtain an accuracy score for each of the at least one chat answer;

(f3) based on a relevance algorithm, evaluating relevance between each of the at least one chat answer and the internal clarification query to obtain a relevance score for each of the at least one chat answer; and

(f4) based on the coherence and clarity score, the accuracy score, and the relevance score for each of the at least one chat answer, selecting the best chat answer from the at least one chat answer.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (22)