Labeling product identifiers and navigating products

• US 9,323,838 B2
• Filed: 09/03/2013
• Issued: 04/26/2016
• Est. Priority Date: 09/05/2012
• Status: Active Grant

• Current Assignee: Alibaba Group Holding Ltd.
• Original Assignee: Alibaba Group Holding Ltd.
• Inventors: Sun, Changlong, Zeng, Anxiang
• Primary Examiner(s): Rostami, Mohammad S
• Application Number:
  US14/017,161
• Publication Number:
  US 20140067815A1
• Time in Patent Office: 966 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/35 Clustering; Classification

Abstract

The present disclosure provides example methods and apparatuses of labeling product identifiers and methods of navigating products. Description information of one or more products is extracted. The description information of the products is clustered into a text. A subject analysis is applied to the text by using a text analysis method based on subject models to obtain one or more subjects and definition names for the subjects. A subject that is correlated to the description information of the product is used as an identifier of the product to label the product. The present techniques label the products with identifiers that have one or more user dimension attributes so that users may easily and intuitively find their desired products.

Claims

1. A method comprising:

extracting description information of multiple products;

clustering the description information of the multiple products belonging to a particular model into a first text;

processing the first text by segmenting the first text to one of

remove from the first text one or more terms whose term frequencies are higher than a first set threshold, and

remove from the first text one or more terms whose term frequencies are lower than a second set threshold;

clustering, after processing the first text, first texts of products belonging to different models into a second text;

applying a subject analysis to the second text to obtain one or more subjects;

defining one or more names for the one or more subjects respectively;

assigning a respective name of a respective subject correlated to description information of a respective product as an identifier of the respective product; and

labeling the respective product by using the identifier,

wherein the applying the subject analysis to the second text to obtain one or more subjects comprises;

setting a number of subjects in one or more subject models;

applying the subject analysis to the second text by using a text analysis method based on the one or more subject models;

obtaining a number of subsets corresponding to the number of subjects from a set of terms in the second text, the number of subsets being equal to the number of subjects, a respective subset corresponding to a respective subject; and

according to the respective subset that one or more terms in the description information of the products locate, correlating the description information of the products to the respective subject corresponding to the respective subset.

2. The method as recited in claim 1, further comprising:

prior to the extracting the description information of the multiple products, categorizing the multiple products,

wherein the extracting the description information of the multiple products comprises extracting description information of multiple products under a particular category, and

the second text includes the description information of the multiple products under the particular category.

3. The method as recited in claim 1, wherein the one or more subject models include a probabilistic latent semantic analysis (PLSA) or a latent Dirichlet allocation (LDA).

4. The method as recited in claim 1, wherein the description information of the multiple products comprises title information, attribute information, or a combination of the title information and the attribute information.

5. The method as recited in claim 1, wherein the description information includes title information, the method further comprising:

determining whether the title information of the respective product includes an additional identifier in an additional identifier database that is pre-established for a particular category that the respective product belongs; and

in response to determining that the title information of the respective product includes the additional identifier, obtaining the additional identifier, and using the additional identifier to label the respective product.

6. The method as recited in claim 5, further comprising establishing the additional identifier database, the establishing including:

according to a log of products under the particular category, extracting one or more inquiry terms whose inquiry frequencies are higher than a preset threshold, segmenting the extracted one or more inquiry terms;

removing one or more category words and stop words to obtain one or more keywords; and

assigning the one or more keywords as additional identifiers to establish the additional identifier database.

7. A method comprising:

extracting description information of multiple products;

clustering the description information of the multiple products belonging to a particular model into a first text;

processing the first text by segmenting the first text to one of;

remove from the first text one or more terms whose term frequencies are higher than a first set threshold, and

remove from the first text one or more terms whose term frequencies are lower than a second set threshold;

clustering, after processing the first text, first texts of products belonging to different models into a second text;

applying a subject analysis to the second text to obtain one or more subjects;

correlating the multiple products to the one or more subjects; and

navigating the multiple products according to a respective subject correlated to a respective product,

wherein the applying the subject analysis to the second text to obtain one or more subjects comprises;

setting a number of subjects in one or more subject models;

applying the subject analysis to the second text by using a text analysis method based on the one or more subject models;

obtaining a number of subsets corresponding to the number of subjects from a set of terms included in the second text, the number of subsets being equal to the number of subjects, a respective subset corresponding to a respective subject; and

according to the respective subset that one or more terms in the description information of the products locate, correlating the description information of the products to the respective subject corresponding to the respective subset.

8. The method as recited in claim 7, further comprising:

prior to the extracting the description information of the multiple products, categorizing the multiple products,

wherein the extracting the description information of the multiple products comprises extracting description information of multiple products under a particular category and the second text includes the description information of the multiple products under the particular category.

9. The method as recited in claim 7, wherein the description information of the multiple products comprises title information, attribute information, or a combination of the title information and the attribute information.

10. The method as recited in claim 7, wherein the navigating the multiple products according to the respective subject correlated to the respective product comprises:

determining a display sequence of the one or more subjects according to numbers of clicks of the products corresponding to the subjects respectively; and

displaying the one or more subjects in a navigation guide at a user interface according to the display sequence.

14. The method as recited in claim 7, wherein the one or more subject models include a probabilistic latent semantic analysis (PLSA) or a latent Dirichlet allocation (LDA).

11. An apparatus comprising a memory and a processor that executes computer executable instructions stored in the memory to cause the processor to:

extract description information of multiple products;

cluster the description information of the multiple products belonging to a particular model into a first text;

process the first text by segmenting the first text to one of

remove from the first text one or more terms whose term frequencies are higher than a first set threshold, and

remove from the first text one or more terms whose term frequencies are lower than a second set threshold;

cluster, after the first text is processed, first texts of products belonging to different models into a second text;

apply a subject analysis to the second text to obtain one or more subjects and defines one or more names for the one or more subjects, respectively; and

assign a respective name of a respective subject correlated to description information of a respective product as an identifier of the respective product and label the respective product by using the identifier,

wherein the computer executable instructions stored in the memory cause the processor to;

set a number of subjects in one or more subject models;

apply the subject analysis to the second text by using a text analysis method based on the one or more subject models;

obtain a number of subsets corresponding to the number of subjects from a set of terms included in the second text, the number of subsets being equal to the number of subjects, a respective subset corresponding to a respective subject; and

according to the respective subset that one or more terms in the description information of the products locate, correlate the description information of the products to the respective subject corresponding to the respective subset.

12. The apparatus as recited in claim 11, wherein the description information of the multiple products comprises title information, and the computer executable instructions stored in the memory cause the processor to:

determine whether the title information of the respective product includes an additional identifier in an additional identifier database pre-established for a particular category that the respective product belongs; and

in response to a result of determining that is positive, obtain the additional identifier and uses the additional identifier in addition to the identifier to label the respective product.

13. The apparatus as recited in claim 12, wherein the computer executable instructions stored in the memory cause the processor to:

according to a log of products under the particular category, extract one or more inquiry terms whose inquiry times are higher than a preset threshold;

segment the extracted one or more inquiry terms;

remove one or more category words and stop words to obtain one or more keywords; and

use the one or more keywords as additional identifiers to establish the additional identifier database.