Granted Patent B2

US 12,724,964 · App. 18/073,519 · Granted Sep 1, 2026

Inventors: Jie Jiang (Shenzhen, CN); Guanglei Du (Shenzhen, CN); Zhilin Shi (Shenzhen, CN); Changwang Zhang (Shenzhen, CN); Jihong Zhang (Shenzhen, CN)

Assignee: TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED

G06F40/279G06F16/26G06F40/166G06F40/30

Abstract

Claims (103)

1 . A text mining method based on artificial intelligence performed by a computer device, the method comprising:

obtaining a plurality of training domain candidate terms;

constructing a positive sample pool and a negative sample pool by classifying the plurality of training domain candidate terms based on a general vocabulary database, including, for each respective training domain candidate term of the plurality of training domain candidate terms:

adding the respective training domain candidate term to the positive sample pool in response to a determination that the respective training domain candidate term belongs to the general vocabulary database; and

adding the respective training domain candidate term to the negative sample pool in response to a determination that the respective training domain candidate term does not belong to the general vocabulary database;

training a random forest model including a plurality of decision trees, including:

constructing a plurality of groups of domain candidate term samples, wherein each group of domain candidate term samples includes a positive domain candidate term sample selected from the positive sample pool and a negative domain candidate term sample selected from the negative sample pool; and

training each of the plurality of decision trees in the random forest model using a different group of the plurality of groups of domain candidate term samples;

obtaining domain candidate term features corresponding to domain candidate terms;

obtaining term quality scores corresponding to the domain candidate terms according to the domain candidate term features, including:

obtaining sub-scores corresponding to the domain candidate term features based on the domain candidate term features by using each of the plurality of decision trees in the random forest model; and

obtaining the term quality scores corresponding to the domain candidate terms according to the sub-scores corresponding to the domain candidate term features;

determining a new term from the domain candidate terms according to the term quality scores corresponding to the domain candidate terms;

obtaining an associated text according to the new term; and

determining a domain seed term as a domain new term in response to determining according to the associated text that the domain seed term satisfies a domain new term mining condition.

2 . The method according to claim 1 , further comprising:

obtaining sentences from a domain corpus;

performing term segmentation processing on each of the sentences to obtain a text sequence; and

obtaining the domain candidate terms according to the text sequence.

3 . The method according to claim 1 , wherein the determining a new term according to the term quality scores corresponding to the domain candidate terms comprises:

determining that a domain candidate term is a new term in response to the term quality score corresponding to the domain candidate term being greater than or equal to a quality score threshold; and

determining that a domain candidate term is not a new term in response to the term quality score corresponding to the domain candidate term being less than the quality score threshold.

4 . The method according to claim 1 , wherein the determining a new term according to the term quality scores corresponding to the domain candidate terms comprises:

obtaining term frequencies corresponding to the domain candidate terms;

determining that a domain candidate term is a new term in response to the term quality score corresponding to the domain candidate term being greater than or equal to a quality score threshold and a term frequency corresponding to the domain candidate term being greater than or equal to a first term frequency threshold; and

determining that a domain candidate term is not a new term in response to the term quality score corresponding to the domain candidate term being less than the quality score threshold and the term frequency corresponding to the domain candidate term being less than the first term frequency threshold.

5 . The method according to claim 1 , wherein the obtaining an associated text according to the new term comprises:

obtaining a search feedback result corresponding to the new term through a search engine, the search feedback result comprising at least one search result; and

determining, according to the search feedback result corresponding to the new term, first R search results with highest degrees of relevance from the at least one search result as the associated text corresponding to the new term, R being an integer greater than or equal to 1.

6 . The method according to claim 1 , further comprising:

obtaining the domain seed term;

determining an average term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the average term frequency being greater than or equal to a second term frequency threshold.

7 . The method according to claim 1 , further comprising:

obtaining the domain seed term;

determining a maximum term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the maximum term frequency being greater than or equal to a second term frequency threshold.

8 . A computer device, comprising: a memory, a processor, and a bus system, the bus system being configured to connect the memory and the processor to enable the memory and the processor to communicate with each other,

the memory being configured to store a program,

the processor being configured to execute the program in the memory and causing the computer device to perform a text mining method based on artificial intelligence including:

obtaining a plurality of training domain candidate terms;

constructing a positive sample pool and a negative sample pool by classifying the plurality of training domain candidate terms based on a general vocabulary database, including, for each respective training domain candidate term of the plurality of training domain candidate terms:

adding the respective training domain candidate term to the positive sample pool in response to a determination that the respective training domain candidate term belongs to the general vocabulary database; and

adding the respective training domain candidate term to the negative sample pool in response to a determination that the respective training domain candidate term does not belong to the general vocabulary database;

training a random forest model including a plurality of decision trees, including:

constructing a plurality of groups of domain candidate term samples, wherein each group of domain candidate term samples includes a positive domain candidate term sample selected from the positive sample pool and a negative domain candidate term sample selected from the negative sample pool; and

training each of the plurality of decision trees in the random forest model using a different group of the plurality of groups of domain candidate term samples;

obtaining domain candidate term features corresponding to domain candidate terms;

obtaining term quality scores corresponding to the domain candidate terms according to the domain candidate term features, including:

obtaining sub-scores corresponding to the domain candidate term features based on the domain candidate term features by using each of the plurality of decision trees in the random forest model; and

obtaining the term quality scores corresponding to the domain candidate terms according to the sub-scores corresponding to the domain candidate term features;

determining a new term from the domain candidate terms according to the term quality scores corresponding to the domain candidate terms;

obtaining an associated text according to the new term; and

determining a domain seed term as a domain new term in response to determining according to the associated text that the domain seed term satisfies a domain new term mining condition.

9 . The computer device according to claim 8 , wherein the method further comprises:

obtaining sentences from a domain corpus;

performing term segmentation processing on each of the sentences to obtain a text sequence; and

obtaining the domain candidate terms according to the text sequence.

10 . The computer device according to claim 8 , wherein the determining a new term according to the term quality scores corresponding to the domain candidate terms comprises:

determining that a domain candidate term is a new term in response to the term quality score corresponding to the domain candidate term being greater than or equal to a quality score threshold; and

determining that a domain candidate term is not a new term in response to the term quality score corresponding to the domain candidate term being less than the quality score threshold.

11 . The computer device according to claim 8 , wherein the determining a new term according to the term quality scores corresponding to the domain candidate terms comprises:

obtaining term frequencies corresponding to the domain candidate terms;

determining that a domain candidate term is a new term in response to the term quality score corresponding to the domain candidate term being greater than or equal to a quality score threshold and a term frequency corresponding to the domain candidate term being greater than or equal to a first term frequency threshold; and

determining that a domain candidate term is not a new term in response to the term quality score corresponding to the domain candidate term being less than the quality score threshold and the term frequency corresponding to the domain candidate term being less than the first term frequency threshold.

12 . The computer device according to claim 8 , wherein the obtaining an associated text according to the new term comprises:

obtaining a search feedback result corresponding to the new term through a search engine, the search feedback result comprising at least one search result; and

determining, according to the search feedback result corresponding to the new term, first R search results with highest degrees of relevance from the at least one search result as the associated text corresponding to the new term, R being an integer greater than or equal to 1.

13 . The computer device according to claim 8 , wherein the method further comprises:

obtaining the domain seed term;

determining an average term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the average term frequency being greater than or equal to a second term frequency threshold.

14 . The computer device according to claim 8 , wherein the method further comprises:

obtaining the domain seed term;

determining a maximum term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the maximum term frequency being greater than or equal to a second term frequency threshold.

15 . A non-transitory computer-readable storage medium, storing a computer program that, when executed by a processor of a computer device, causes the computer device to perform a text mining method based on artificial intelligence including:

obtaining a plurality of training domain candidate terms;

constructing a positive sample pool and a negative sample pool by classifying the plurality of training domain candidate terms based on a general vocabulary database, including, for each respective training domain candidate term of the plurality of training domain candidate terms:

adding the respective training domain candidate term to the positive sample pool in response to a determination that the respective training domain candidate term belongs to the general vocabulary database; and

adding the respective training domain candidate term to the negative sample pool in response to a determination that the respective training domain candidate term does not belong to the general vocabulary database;

training a random forest model including a plurality of decision trees, including:

constructing a plurality of groups of domain candidate term samples, wherein each group of domain candidate term samples includes a positive domain candidate term sample selected from the positive sample pool and a negative domain candidate term sample selected from the negative sample pool; and

training each of the plurality of decision trees in the random forest model using a different group of the plurality of groups of domain candidate term samples;

obtaining domain candidate term features corresponding to domain candidate terms;

obtaining term quality scores corresponding to the domain candidate terms according to the domain candidate term features, including:

obtaining sub-scores corresponding to the domain candidate term features based on the domain candidate term features by using each of the plurality of decision trees in the random forest model; and

obtaining the term quality scores corresponding to the domain candidate terms according to the sub-scores corresponding to the domain candidate term features;

determining a new term from the domain candidate terms according to the term quality scores corresponding to the domain candidate terms;

obtaining an associated text according to the new term; and

determining a domain seed term as a domain new term in response to determining according to the associated text that the domain seed term satisfies a domain new term mining condition.

16 . The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

obtaining sentences from a domain corpus;

performing term segmentation processing on each of the sentences to obtain a text sequence; and

obtaining the domain candidate terms according to the text sequence.

17 . The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

obtaining the domain seed term;

determining an average term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the average term frequency being greater than or equal to a second term frequency threshold.

18 . The non-transitory computer-readable storage medium according to claim 15 , wherein the method further comprises:

obtaining the domain seed term;

determining a maximum term frequency of the domain seed term according to the associated text; and

determining that the domain seed term satisfies the domain new term mining condition in response to the maximum term frequency being greater than or equal to a second term frequency threshold.

Assignments (1)

Priority Claims (1)

Continuity (2)

References Cited (20)