Publishing product information
The present disclosure provides a method and an apparatus for publishing product information. The present disclosure provides a method for publishing product information. Based on a stored search click log of buyers, correlation information between inquiry words and categories in the search click log is calculated. A keyword input by the seller is matched to the inquiry words. The keyword may be a word or a phrase that includes one or more words. If the keyword is matched to at least one inquiry word, at least one category corresponding to the matched inquiry word is obtained based on the correlation information. The product information is stored under one or more categories of the obtained categories. The present techniques improve the accuracy rate of recommended categories to the seller and the return rate of the published product information.
1. A method comprising:
performing a matching of an input keyword to one or more inquiry words in a search click log;
rewriting the input keyword in response to determining that the input keyword does not match any inquiry word in the search click log;
obtaining one or more categories corresponding to at least one inquiry word in the search click log based at least in part on correlation information in response to determining that the rewritten keyword matches the at least one inquiry word in the search click log, wherein the correlation information is calculated between a plurality of inquiry words and a plurality of categories in the search click log by calculating confidence degrees of each category corresponding to a respective inquiry word, and wherein a confidence degree of a respective category corresponding to the respective inquiry word comprises a weighted combination of a first conditional probability between the respective inquiry word and the respective category when the respective category is clicked after the respective inquiry word is received, and a second conditional probability between the respective inquiry word and the respective category when one or more products under the respective category are clicked after the respective inquiry word is received; and
publishing product information under the one or more obtained categories.
2. A method as recited in claim 1 , further comprising:
ranking the one or more obtained categories based on respective relevant degrees with the rewritten keyword; and
selecting at least one category from the one or more obtained categories based on a result of the ranking.
3. A method as recited in claim 1 , wherein rewriting the input keyword comprises deleting one or more words from the input keyword.
4. A method as recited in claim 1 , rewriting the input keyword comprises:
labeling a respective importance value for each word included in the input keyword; and
deleting one or more words whose respective importance values are lower than a preset importance value threshold from the input keyword.
5. A method as recited in claim 4 , wherein the respective importance value for each word included in the input keyword is labeled based on one or more of a respective syntax, grammar, semantics, or statistical characteristic of a respective word.
6. A method as recited in claim 1 , further comprising:
in response to determining that the rewritten keyword does not match any inquiry word in the search click log,
classifying the input keyword into one or more characteristics;
calculating a posterior probability of each of the one or more characteristics under each category in the search click log; and
determining at least one category which posterior probability is higher than a preset threshold as a category that matches the input keyword.
7. A method as recited in claim 6 , wherein the one or more characteristics include one or more of:
a product that determines whether the input keyword is a product name;
a brand that determines whether the input keyword is a brand name;
a model that determines whether the input keyword is a model name;
a centric word in the input keyword;
one or more nouns in the input keyword;
the centric word and a noun left to the centric word; or
the centric word and a noun right to the centric word.
8. A method as recited in claim 6 , wherein calculating the posterior probability of each of the one or more characteristics under each category comprises using a formula of
p
(
y
|
x
)
=
1
Z
(
x
)
exp
(
∑
j
λ
j
f
j
(
x
,
y
)
)
,
wherein:
y represents a respective category in the search click log;
x represents the input keyword;
f j (x, y) represents a j th characteristic of x under the respective category y;
λ j represents a weight of the j th characteristic; and
Z(x) represents a normalization factor.
9. A method as recited in claim 1 , wherein the first conditional probability comprises a ratio between a number of times that the respective category is clicked within a period of time when the respective inquiry word is received, and a number of times that the respective inquiry word is received within the period of time.
10. A method as recited in claim 1 , wherein the second conditional probability comprises a ratio between a number of times that the one or more products under the respective category are clicked within a period of time when the respective inquiry word is received, and a number of times that the respective inquiry word is received within the period of time.
11. A method as recited in claim 1 , further comprising ranking the one or more obtained categories based on respective confidence degrees.
12. A method as recited in claim 1 , wherein the search click log is established from clicking behaviors of buyers.
13. A method as recited in claim 1 , wherein the input keyword is received from a seller.
14. A method comprising:
performing a matching of an input keyword to one or more inquiry words in a search click log; in response to determining that the input keyword does not match any inquiry word in the search click log, rewriting the input keyword;
performing a matching of the rewritten keyword to a plurality of inquiry words in the search click log; and
in response to determining that the rewritten keyword does not match any inquiry word in the search click log,
classifying the input keyword into one or more characteristics,
calculating a probability of each characteristic of the one or more characteristics under each category in the search click log, and
determining at least one category which probability is higher than a preset threshold as a category that matches the input keyword.
15. A method as recited in claim 14 , wherein rewriting the input keyword comprises:
labeling a respective importance value for each word included in the input keyword; and
deleting a word which respective importance value is lower than a preset importance value threshold.
16. One or more computer storage media storing executable instructions that, when executed by one or more processors, cause the one or more processors to perform acts comprising:
performing a matching of an input keyword to one or more inquiry words in a search click log; in response to determining that the input keyword does not match any inquiry word in the search click log, rewriting the input keyword;
performing a matching of the rewritten keyword to a plurality of inquiry words in the search click log; and
in response to determining that the rewritten keyword does not match any inquiry word in the search click log,
classifying the input keyword into one or more characteristics,
calculating a probability of each characteristic of the one or more characteristics under each category in the search click log, and
determining at least one category having a probability being higher than a preset threshold as a category that match the input keyword.
17. The one or more computer storage media as recited in claim 16 , wherein rewriting the input keyword comprises:
labeling a respective importance value for each word included in the input keyword; and
deleting a word which respective importance value is lower than a preset importance value threshold from the input keyword.
18. The one or more computer storage media as recited in claim 17 , wherein the respective importance value for each word included in the input keyword is labeled based on one or more of a respective syntax, grammar, semantics, or statistical characteristic of a respective word.
19. The one or more computer storage media as recited in claim 16 , wherein the one or more characteristics include one or more of:
a product that determines whether the input keyword is a product name;
a brand that determines whether the input keyword is a brand name;
a model that determines whether the input keyword is a model name;
a centric word in the input keyword;
one or more nouns in the input keyword;
the centric word and a noun left to the centric word; or
the centric word and a noun right to the centric word.