CN103136221B - A kind of method for generating requirement templet, demand know method for distinguishing and its device - Google Patents
A kind of method for generating requirement templet, demand know method for distinguishing and its device Download PDFInfo
- Publication number
- CN103136221B CN103136221B CN201110379335.5A CN201110379335A CN103136221B CN 103136221 B CN103136221 B CN 103136221B CN 201110379335 A CN201110379335 A CN 201110379335A CN 103136221 B CN103136221 B CN 103136221B
- Authority
- CN
- China
- Prior art keywords
- type
- requirement
- query
- template
- queries
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 77
- 230000006870 function Effects 0.000 claims description 48
- 238000010586 diagram Methods 0.000 description 12
- 239000013598 vector Substances 0.000 description 8
- 230000007704 transition Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 210000004072 lung Anatomy 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a kind of method for generating requirement templet, demand knows method for distinguishing and its device, wherein the method for generation requirement templet includes:The kind subquery of demand type is obtained from search daily record;Seed Query generalization by the demand type is the candidate template of the demand type;The final template of the demand type is chosen from the candidate template of the demand type.Wherein demand is known method for distinguishing and is included:Obtain user's inquiry;The final template matched with user inquiry is determined in the final template that the method for generation requirement templet described previously is obtained, and the demand that the demand type corresponding to the final template that match has as user inquiry will be inquired about with the user.Through the above way, the scalability and maintainability of demand recognizer is made all to greatly improve.
Description
[ technical field ] A method for producing a semiconductor device
The present invention relates to natural language processing technology, and in particular, to a method for generating a requirement template, a method for identifying a requirement, and an apparatus thereof.
[ background of the invention ]
With the widespread use of search engines, search engine technology has developed greatly, and today's search engines have not only been satisfied with returning content that matches a user query, but rather attempt to return content that is relevant to the needs of the user query.
To return content related to the user query requirement, first of all, the requirement of the user during searching needs to be understood, that is, the user requirement needs to be identified. For example, through observation, a developer finds that a query ending with "mp 3" in a user's query is generally a music-like requirement, and the developer writes the rule that "the requirement ending with mp3 is a music requirement" into the requirement identification program. This method has the following two problems: firstly, the establishment of the rules depends on human observation, so that a large amount of manpower and material resources are consumed, and the establishment of the rules which comprehensively cover various requirements is difficult, so that the application program is difficult to identify various requirements of the user; secondly, the method is adopted to identify the user requirements, and as the rules are embedded into the online requirement identification program codes, the expandability and maintainability of the requirement identification program are greatly reduced.
[ summary of the invention ]
The technical problem to be solved by the present invention is to provide a method for generating a requirement template, a method for identifying a requirement, and a device thereof, so as to solve the defects that in the prior art, a program for identifying a user requirement is difficult to comprehensively identify various requirements of a user, and the expandability and the maintainability are poor.
The technical scheme adopted by the invention for solving the technical problem is to provide a method for generating a demand template, which comprises the following steps: acquiring a seed query of a demand type from a search log; generalizing the seed query of the demand type into a candidate template of the demand type; and selecting a final template of the requirement type from the candidate templates of the requirement type.
According to a preferred embodiment of the present invention, the step of obtaining a seed query of a requirement type includes: acquiring a preset initial seed query of the demand type; clustering all queries recorded in the search log according to a hierarchical clustering method; and determining a clustering hierarchy, so that the initial seed queries which are not less than a preset proportion in the hierarchy are clustered into the same class X, the total number of the queries contained in the class X in the hierarchy is minimum, and all the queries contained in the class X in the hierarchy are used as the seed queries of the demand type.
According to a preferred embodiment of the present invention, the step of obtaining the seed query of the requirement type includes: acquiring a preset initial seed query of the demand type; and learning a query with the similarity meeting preset requirements with the initial seed query from a search log by using an iterative learner, and taking the learned query and the initial seed query as the seed query of the demand type.
According to a preferred embodiment of the present invention, the step of obtaining a seed query of a requirement type includes: selecting N1 queries with the highest query times from the queries which cause the page of the demand type to be clicked in the search log as seed queries of the demand type, wherein N1 is a preset positive integer; or extracting the seed query of the demand type from a search log of the vertical search of the demand type.
According to a preferred embodiment of the present invention, the step of generalizing the seed query of the requirement type into the candidate template of the requirement type includes: replacing a part, matched with a preset entity word corresponding to the demand type, in the seed query of the demand type with a wildcard character of the category to which the preset entity word belongs; or replacing the part identified by the category identification function in the seed query of the requirement type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
According to a preferred embodiment of the present invention, the step of generalizing the seed query of the requirement type into the template of the requirement type further comprises: and replacing the terms with the length wildcard character for limiting the length of the terms, wherein the contribution degree of the terms to the requirement type in the seed query of the requirement type is lower than the preset contribution degree requirement.
According to a preferred embodiment of the present invention, when the final template of the requirement type is selected from the candidate templates of the requirement type, the final template of the requirement type is selected according to at least one of the following characteristics of the candidate templates of the requirement type: the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type; the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type; and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
According to a preferred embodiment of the present invention, the click feature of the candidate template W of the requirement type adopts the following methodCalculating the formula:where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
According to a preferred embodiment of the present invention, the similarity characteristic of the candidate template W of the requirement type is calculated in the following manner:wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
According to a preferred embodiment of the present invention, the matching capability feature of the candidate template W of the requirement type is calculated by the following method:wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the requirement type,representing the number of queries W matches in a search log of queries of various demand types.
The invention also provides a method for identifying the demand, which comprises the following steps: acquiring a user query; determining a final template matched with the user query from the final templates obtained by the method for generating the requirement template, and taking the requirement type corresponding to the final template matched with the user query as the requirement of the user query.
The invention also provides a device for generating the demand template, which comprises the following steps: the seed acquisition unit is used for acquiring a seed query of a demand type from the search log; a generalization unit, configured to generalize the seed query of the requirement type into a candidate template of the requirement type; and the selecting unit is used for selecting a final template of the requirement type from the candidate templates of the requirement type.
According to a preferred embodiment of the present invention, the seed obtaining unit includes: the first selection unit is used for acquiring a preset initial seed query of the demand type; the clustering unit is used for clustering all the queries recorded in the search log according to a hierarchical clustering method; and the determining unit is used for determining a clustering hierarchy, so that the initial seed queries not less than a predetermined proportion under the hierarchy are clustered into the same class X, the total number of the queries contained in the class X under the hierarchy is minimum, and all the queries contained in the class X under the hierarchy are used as the seed queries of the required types.
According to a preferred embodiment of the present invention, the seed obtaining unit includes: the second selection unit is used for acquiring the preset initial seed query of the demand type; and the learning unit is used for learning a query with the similarity meeting preset requirements with the initial seed query from a search log by using an iterative learner, and taking the learned query and the initial seed query as the seed query of the demand type.
According to a preferred embodiment of the present invention, when acquiring the seed query of the requirement type, the seed acquisition unit specifically selects N1 queries with the highest query frequency from queries that result in the page of the requirement type being clicked in the search log as the seed query of the requirement type, where N1 is a preset positive integer; or extracting the seed query of the demand type from a search log of the vertical search of the demand type.
According to a preferred embodiment of the present invention, when the generalization unit generalizes the seed query of the requirement type into the candidate template of the requirement type, specifically, a part of the seed query of the requirement type, which is matched with a preset entity word corresponding to the requirement type, is replaced with a wildcard character of a category to which the preset entity word belongs; or replacing the part identified by the category identification function in the seed query of the requirement type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
According to a preferred embodiment of the present invention, the generalization unit is further configured to replace a term in the seed query of the requirement type, which has a contribution degree lower than a preset contribution degree requirement on the requirement type, with a length wildcard for limiting a term length.
According to a preferred embodiment of the present invention, when the selecting unit selects the final template of the requirement type from the candidate templates of the requirement type, the selecting unit performs the selecting according to at least one of the following characteristics of the candidate template of the requirement type: the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type; the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type; and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
According to a preferred embodiment of the present invention, the selecting unit calculates the click characteristics of the candidate template W of the requirement type by using the following method:where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
According to a preferred embodiment of the present invention, the selecting unit calculates the similarity feature of the candidate template W of the requirement type by using the following method:wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
According to a preferred embodiment of the present invention, the selecting unit calculates the matching capability feature of the candidate template W of the requirement type by using the following method:wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the requirement type,representing the number of queries W matches in a search log of queries of various demand types.
The invention also provides a device for identifying the demand, which comprises: the query acquisition unit is used for acquiring user queries; a matching unit, configured to determine, from the final templates obtained by the device for generating a requirement template, a final template matched with the user query, and use a requirement type corresponding to the final template matched with the user query as a requirement of the user query.
According to the technical scheme, through the mode, developers do not need to write the identification rules into the requirement identification program manually, and the machine automatically generates the requirement template by using the query recorded in the search log to identify the requirements of the users. In the online program, after the user query is obtained, the requirement type of the user can be well judged by using the requirement template generated offline. In the mode of identifying the user requirements, the requirement template is automatically generated, so that manpower and material resources are saved, and meanwhile, the requirement template and the requirement identification are generated to realize the separation of offline and online, so that the expandability and maintainability of the requirement identification program are greatly improved.
[ description of the drawings ]
FIG. 1 is a flow chart illustrating an embodiment of a method for generating a requirement template and a method for identifying a requirement in the present invention;
FIG. 2 is a schematic diagram of a structure for clustering queries in a search log according to the present invention;
FIG. 3 is a schematic diagram of a matching tree in accordance with the present invention;
FIG. 4 is a block diagram schematically illustrating the structure of an embodiment of an apparatus for generating a requirement template and an apparatus for identifying a requirement in the present invention;
FIG. 5 is a block diagram illustrating the structure of an embodiment of a seed obtaining unit according to the present invention;
fig. 6 is a block diagram schematically illustrating the structure of another embodiment of the seed obtaining unit in the present invention.
[ detailed description ] embodiments
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1, fig. 1 is a flowchart illustrating a method for generating a requirement template and a method for identifying a requirement according to an embodiment of the present invention. As shown in fig. 1, the present embodiment is divided into an offline part and an online part, wherein the offline part is a flowchart of an embodiment of a method for generating a requirement template, and the online part is a flowchart of an embodiment of a method for identifying a requirement. The method for generating the demand template by the offline part comprises the following steps:
step S101: a seed query of a requirement type is obtained from a search log (querylog).
Step S102: and generalizing the seed query of the corresponding demand type into a candidate template of the corresponding demand type.
Step S103: and selecting a final template of the corresponding demand type from the candidate templates of the corresponding demand type.
The above steps are specifically described below.
In step S101, a seed query for a requirement type includes queries that can reflect the requirement type expressed in various ways. For example, the seed query "step surprise theme song" and "three-inch paradise" are different in expression, but the requirements for expression are the same, and the same song is queried.
There are various ways to obtain the seed query of the requirement type, and the following describes the way to obtain the seed query by using a specific embodiment.
Embodiment one of obtaining a seed query for a requirement type:
step S1011: and acquiring a preset initial seed query of the type of the demand, and clustering all queries recorded in the search log according to a hierarchical clustering method.
Step S1012: and determining a clustering hierarchy, so that not less than a predetermined proportion of initial seed queries in the hierarchy are clustered into the same class X, the total number of queries contained in the class X in the hierarchy is minimum, and all the queries contained in the class X in the hierarchy are used as seed queries of the type of the demand. It is to be understood that, the predetermined ratio is used as a ratio value, the value range should be greater than or equal to zero and less than or equal to 1, and when the predetermined ratio is 1 (i.e. 100%), no less than the predetermined ratio of the initial seed queries is actually all the initial seed queries.
The initial seed query may be obtained by manually selecting a query having that type of requirement. The embodiment can perform hierarchical clustering in a bottom-up or top-down manner. The bottom-up hierarchical clustering method comprises the following steps: taking each query as a class, and then combining the two most similar classes continuously through iteration until all the queries are combined into one class; the top-down hierarchical clustering method comprises the following steps: all queries are first treated as one class, and then through iteration, the most dissimilar query is found to be split out into two classes until each query is self-classified.
In clustering, the similarity between different queries is obtained by calculating the cosine similarity between feature word vectors corresponding to different queries, wherein the feature word corresponding to each query can be extracted from the search result corresponding to the query. In addition, the user click data corresponding to each query may also be used as a feature to calculate the similarity between different queries. The method for judging the similarity between different queries in the clustering is not limited, and can be carried out in any way known by the technical personnel in the field.
The above process of obtaining a seed query for a requirement type is described below by way of a specific example. Suppose the initial seed query of music demand type is "pace surprise theme song", "three cun paradise mp 3" and "love me chinese song for free trial listening". The queries in the search logs are: 1. step surprise piece tail song, 2 step surprise theme song, 3, three-inch paradise, 4, Yanzhan three-inch paradise, 5, three-inch paradise mp3, 6, good-hearing song, 7 love my Chinese song for free trial listening, 8, step surprise online viewing, 9, step surprise 30, 10, step surprise no deletion version, 11, crossing the story of step surprise, and then referring to fig. 2 for hierarchical clustering results of the above queries.
FIG. 2 is a diagram illustrating the results of clustering queries in a search log according to the present invention. As shown in FIG. 2, at the first level of the clustering results, queries 1, 2 are one type, queries 3, 4 are one type, queries 8, 9 are one type, and the remaining queries are each one type. On the second level of clustering results, queries 1, 2 are one type, queries 3, 4, 5 are one type, queries 8, 9, 10 are one type, and the remaining queries are each one type. At the third level of the query, queries 1, 2, 3, 4, 5 are one class, queries 8, 9, 10 are one class, and the remaining queries are each one class. At the fourth level of queries, queries 1, 2, 3, 4, 5, 6 are one class, queries 8, 9, 10 are one class, and the remaining queries are each one class. On the fifth level of queries, queries 1, 2, 3, 4, 5, 6, 7 are one class, queries 8, 9, 10 are one class, and query 11 is itself one class. At the sixth level of queries, queries 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 are one class, and query 11 is itself one class. At the seventh level of queries, all queries form a class. It can be seen that the initial seed queries are queries 2, 5, 7, respectively, and if the aforementioned predetermined ratio is 1, it is required to determine a level under which all initial seed queries (i.e. queries 2, 5, 7) are in the same class and the number of queries included in this class should be the smallest, obviously, at the fifth level, there is a class satisfying this condition, and all queries included in this class are queries 1, 2, 3, 4, 5, 6, 7, respectively, so that queries 1, 2, 3, 4, 5, 6, 7 are the final seed queries of the music demand type.
Embodiment two of obtaining a seed query of a requirement type:
step S101 a: and acquiring a preset initial seed query of the type of the demand.
Step S102 b: and learning a query with the similarity meeting the requirement with the initial seed query from the search log by using an iterative learner, and taking the learned query and the initial seed query as the seed query of the requirement type.
Similar to the first embodiment, the initial seed query may be obtained by manually selecting a query with the requirement type, and in this embodiment, the similarity between different queries also needs to be obtained, and the calculation method of the similarity between different queries is similar to that described in the first embodiment, and is not described herein again. The iterative learner can be obtained by any machine learning method with or without supervision, and the invention is not described in detail.
In addition to the first and second embodiments described above, when a seed query of a requirement type is obtained, the determination may be made according to the query that caused the page of the requirement type to be clicked. For example: and selecting N1 queries with the highest query times from the queries which cause the clicked pages of the requirement type as seed queries of the requirement type, wherein N1 is a preset positive integer. For example, if a page for a pop music download is heavily clicked by queries such as "Zhou Jie Lung album download", "Zhou Jie Lung flower stand", "still Van Te xi", etc., these queries can be used as seed queries for the type of pop music demand. In addition, a seed query for a demand type can be extracted from a search log of a vertical search for the demand type. The vertical search is a search for an industry or a field, and when a user searches a travel website, the user expresses a requirement related to travel, but not a requirement related to catering. Therefore, when the seed query of the requirement related to the travel needs to be obtained, the seed query can be directly extracted from the search logs in the travel search field.
Please continue to refer to fig. 1. In step S102, the query is generalized into a candidate template, that is, the query is limited by wildcards to generate the candidate template.
Specifically, the implementation of step S102 includes the following, see examples three, four, and five.
Example three:
and replacing the part matched with the preset entity word corresponding to the corresponding demand type in the seed query of the corresponding demand type with the wildcard character of the category to which the preset entity word belongs.
For example, the seeds of the music demand type are inquired by a step frightening tail song and a Yanzhan three-inch paradise, and the preset entity words corresponding to the music demand type are as follows: when the user walks by a certain distance (belonging to the category of the name of a movie television show), the artist (belonging to the category of the name of a singer) and the three-inch auditorium (belonging to the category of the name of a song), the user can respectively generalize the step by a certain distance to the step by a certain distance and respectively generalize the step by a certain distance to the step by a certain distance and respectively generalize the step by a.
Example four:
and replacing the part of the seed query of the corresponding demand type, which is identified by the category identification function, with the wildcard of the category corresponding to the category identification function, wherein the category identification function is a function for identifying the category, which is defined according to the attribute of the category.
The category identification function includes a name identification function, a symbol identification function, an english identification function, a number identification function, a date identification function, a commodity model identification function, and the like, wherein the name, the symbol, the english, and the like represent attributes of the corresponding category. It should be understood that the category identification function is not limited to the above categories, and all the category identification functions that can be realized by those skilled in the art are included in the scope of the present invention.
The category identification function in the fourth embodiment may also be used in combination with the preset entity word in the third embodiment, so as to enhance the correctness of the wildcards in the candidate template.
Seed query of the following 5 video demand types:
1. "latest movies by article performance"
2. "May \29709;
3. "movie of grand lecture together with grand couple" of Sun Li Deng "
4. 'model ice self-bending-down luggage video pickup'
5. "all high definition video of michael jackson"
If the name recognition function is used in conjunction with a pre-defined vocabulary of entity words containing a star category, the following candidate templates can be obtained: the latest movie played by [ Star ], the inspirational movie played by [ Star ], the movie played by [ Star ] [ Star ] wife together, [ Star ] the self-stoop for collecting luggage videos, and all high-definition videos of [ F: name ], wherein [ Star ] is a wildcard meeting the name of a Star of a preset Star category entity word, and [ F: name ] is a wildcard of a person name that can be recognized by a person name recognition function. However, in another example where the query is "where there is a movie review article with rising sun as usual", although the "article" matches a preset star category entity word, since the name recognition function does not recognize the "article" in this query as a name, the "article" in this query is not generalized, thereby improving the accuracy of the candidate template.
The name recognition function may be defined according to a co-occurrence probability of the term, for example, statistics of large-scale corpus resources, a probability of co-occurrence of a term with a context term is determined according to the co-occurrence probability of the term, and the term is considered as the name of the person when the probability is greater than a set threshold. Other types of class identification functions can also be defined according to the characteristics of the identification type, and are not described herein again.
Example five:
on the basis of the third embodiment and the fourth embodiment, the terms with the contribution degree lower than the preset contribution degree requirement on the corresponding demand type in the seed query of the corresponding demand type are replaced by length wildcards for limiting the length of the terms.
For example, 5 seed queries, i.e., "last performed" in query 1, "inspired performed" in query 2, "mastered with couples" in query 3, "stoop to pick up baggage by oneself" in query 4, and "all high definition" in query 5, in the example of embodiment four, have a low degree of contribution to determining whether a query belongs to the video demand type, and thus, these terms may be replaced with length wildcards. For example, replace "last in evolution" in query 1 with [ W:1-4], where [ W:1-4] is a wildcard representing 1 to 4 words in length. In the fourth embodiment, based on the processing in the third and fourth embodiments, the 5 seed queries further replace the terms with the contribution degree lower than the preset contribution degree requirement of the video demand type to obtain the following candidate templates:
1. [ Star ] [ W:1-4] motion Picture
2. [ Star ] [ W:1-5] motion Picture
3. [ Star ] [ Star ] [ W:1-7] motion Picture
4. [ Star ] [ W:1-8] video
5. [ F: name ] [ W:1-5] video
The candidate template 1 and the candidate template 2 may also be merged according to a certain merging strategy, for example, by using the maximum distance of the matching length interval of the length wildcard in the candidate templates to be merged, and the candidate template 1 and the candidate template 2 are merged as follows: [ Star ] [ W:1-5] motion Picture.
The contribution of the terms in the query to the corresponding demand type can be obtained by calculating the cosine distance between the vector formed by the terms with the n-gram granularity in the query and the vector formed by the terms with the corresponding demand type, wherein the n-gram refers to a segment formed by n terms with the minimum granularity capable of being independently expressed, for example, "last word" is a 2-gram formed by the terms with the minimum granularity "last word" and "last word". The concept of n-grams can be referred to various existing word segmentation techniques, and will not be described in detail here. Determining the words of the corresponding requirement type may also be performed by using various existing technologies, such as manual selection or mining in the corpus of the corresponding requirement type, and the like, which are not the key points of the present invention, and thus are not described herein again.
As can be seen from the above description, through step S102, candidate templates of corresponding requirement types can be obtained, but since there may be excessive generalization in the generalization process, such as a template "[ singer name ] [ W:1-4 ]", not only a query of "Seiyitan three-inch Tang" but also a query of "Seiyitan height" can be matched, assuming that this template is a template of a music requirement type, it is obvious that the requirement expressed by the matching latter query is not of a music type, which indicates that a sufficiently accurate requirement template is to be obtained, and needs to be selected from the candidate templates obtained in step S102. Therefore, in step S103, the candidate template generated in step S102 is selected to obtain the final template.
In step S103, the candidate templates are selected by using a classifier, that is, the candidate templates are divided into correct templates and incorrect templates by using the classifier, where the correct templates are the final templates to be selected as the corresponding types of requirements.
For the classifier, the most important factor influencing the classification result is the feature extracted from the candidate template. In the present invention, features that can be extracted include, but are not limited to: click characteristics, similarity characteristics and matching capability characteristics. The click feature is used for representing the probability that a page of the corresponding demand type can be clicked due to the query covered by the candidate template of the corresponding demand type, the similarity feature is used for representing the degree of commonality between one candidate template of the corresponding demand type and all candidate templates of the corresponding demand type, and the matching capability feature is used for representing the capability of the candidate template of the corresponding demand type for matching the query of the corresponding demand type.
Specifically, the click feature of the candidate template W can be expressed by formula (1):
wherein click (W) represents the click characteristic of W,representing the number of times W caused the corresponding demand type page (URL) to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
For example, all the queries covered by the candidate template "[ title of movie ] theme song download" in the search log include "step-by-step surprise theme song download" and "water enterprising theme song download", among the clicks caused by these two queries, 100 clicks point to music websites, and 5 clicks point to other websites, and then the click feature value of this candidate template "[ title of movie ] theme song download" with respect to the music requirement is 100/105. And the queries covered by the candidate template of the ' movie drama name ' online reading ' include ' step surprise online reading ' and ' Shuihu online reading ', among the click points caused by the two queries, only 3 click points to the music website, and 100 click points to other websites (mainly reading websites), so that it is obvious that the click characteristic value of the candidate template of the ' movie drama name ' online reading ' relative to the music requirement is 3/103, and the click characteristic value of the candidate template of the ' movie drama name ' online reading ' relative to the reading requirement is easily judged by the method, so that the reading requirement is more likely.
The similarity characteristic of the candidate template W can be calculated by the formula (2):
wherein similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the corresponding requirement type.
For example, there are A, B, C candidate templates under the music requirement type, the similarity feature of candidate template a (similarity), (a), (B), (C), (a, a), (C, etc.) is S (a, B) + S (a, C).
The similarity between one candidate template X and another candidate template Y can be obtained by calculating the cosine distance between the word vector obtained by X and the word vector obtained by Y. The word vectors of the candidate template X or Y may be obtained in various manners, for example, extracting the keyword from the query matched with X or Y to form a word vector, or extracting the keyword from the search result caused by the query matched with X or Y to form a word vector, where the extraction manner may be any manner that may occur to those skilled in the art, and the present invention is not limited thereto.
The matching capability feature of the candidate template W can be calculated by equation (3):
wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the corresponding requirement type,representing the number of queries W matches in a search log of queries of various demand types.
The search log formed by the queries of the corresponding requirement type refers to a log file in which only the queries of the requirement type are recorded. For example, a search log taken from a music website would obviously record all queries that should be of the music demand type.
With continued reference to fig. 1, the method for demand identification of the on-line portion in fig. 1 includes:
step S201: a user query is obtained.
Step S202: determining a final template matched with the user query from the final templates obtained by the method for generating the requirement template, and taking the requirement type corresponding to the final template matched with the user query as the requirement of the user query.
In step S202, a final template matching the user query is determined, and matching may be performed by using a tree-structured algorithm. In the tree structure, each node represents a state, wherein the root node represents an initial state, the leaf nodes represent the states of successful template matching, and the intermediate nodes represent intermediate states in the matching process. Connecting an edge between two nodes is referred to as a state transition condition.
Referring to fig. 3, fig. 3 is a schematic diagram of a matching tree according to the present invention. For the query "three inch paradise 2011", the initial state is 1, and since "three inch paradise" meets the state transition condition "[ song title ]", the state transitions from state 1 to state 3, and since "2011" meets the state transition condition "[ F: time ]" (indicating that it can be recognized by the time recognition function), the state transitions from state 3 to state 9 (i.e., leaf nodes indicating that the matching is successful). The combination of state transition conditions that make up the state transition route forms a template that matches the query.
Since the requirement type corresponding to the template matching the query is known, after the matching process is finished, it can be determined that the user query has a requirement consistent with the matched template.
It should be understood that the above examples are provided by way of example using a query tree algorithm, and are not intended to limit the matching process of the present invention, and in fact, any known matching algorithm may be used herein, for example, using a regular expression to perform matching, and the present invention will not be described in detail herein since the matching algorithm is known to those skilled in the art.
Please refer to fig. 4. Fig. 4 is a schematic block diagram of the structure of an embodiment of the device for generating a requirement template and the device for identifying a requirement in the present invention, wherein the lower part of the line is a schematic diagram of the device for generating a requirement template, and the upper part of the line is a schematic diagram of the device for identifying a requirement. As shown in fig. 4, the apparatus for generating a requirement template includes: a seed obtaining unit 301, a generalization unit 302 and a selection unit 303.
The seed obtaining unit 301 is configured to obtain a seed query of a requirement type. A generalization unit 302, configured to generalize the seed query of the corresponding requirement type into a candidate template of the corresponding requirement type. The selecting unit 303 is configured to select a final template of the corresponding requirement type from the candidate templates of the corresponding requirement type.
Referring to fig. 5, fig. 5 is a schematic block diagram illustrating a structure of a seed obtaining unit according to an embodiment of the present invention. As shown in fig. 5, the seed acquisition unit 301 includes: a first selecting unit 3011, a clustering unit 3012, and a determining unit 3013.
The first selecting unit 3011 is configured to obtain an initial seed query of a preset corresponding requirement type. The clustering unit 3012 is configured to cluster all queries recorded in the search log according to a hierarchical clustering method. The determining unit 3013 is configured to determine a clustering hierarchy, so that at least a predetermined proportion of the initial seed queries at the hierarchy are grouped into the same class X, and the total number of queries included in the class X at the hierarchy is the minimum, and all queries included in the class X at the hierarchy are used as seed queries of corresponding requirement types.
Referring to fig. 6, fig. 6 is a schematic block diagram illustrating a structure of a seed obtaining unit according to another embodiment of the present invention. As shown in fig. 6, the seed acquisition unit 301 includes: a second extraction unit 301a and a learning unit 301 b.
The second selecting unit 301a is configured to obtain an initial seed query of a preset corresponding requirement type. The learning unit 301b is configured to learn, from a search log, a query whose similarity to the initial seed query meets a preset requirement using an iterative learner, and use the learned query and the initial seed query as a seed query of a corresponding requirement type.
In addition to the manners shown in fig. 5 and fig. 6, the seed obtaining unit 301 may further select, from the queries that result in the pages of the corresponding requirement types being clicked in the search logs, N1 queries with the highest query frequency as the seed queries of the corresponding requirement types, or extract the seed queries of the corresponding requirement types from the search logs that reflect the vertical search of the corresponding requirement types.
Please continue to refer to fig. 4. When the generalization unit 302 generalizes the seed query of the corresponding requirement type into the candidate template of the corresponding requirement type, the following concrete steps are performed:
the first method is as follows: replacing the part matched with the preset entity word corresponding to the corresponding demand type in the seed query of the corresponding demand with a wildcard character of the category to which the preset entity word belongs, or
The second method comprises the following steps: and replacing the part of the seed query of the corresponding demand type, which is identified by the category identification function, with the wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
When the generalization unit 302 generalizes the seed query of the corresponding requirement type into the candidate template of the corresponding requirement type, on the basis of the above two manners, the method further includes replacing the terms, whose contribution degree to the corresponding requirement type is lower than the preset contribution degree requirement, in the seed query of the corresponding requirement type with the length wildcard character for limiting the term length.
When the selecting unit 303 selects the final template of the corresponding requirement type from the candidate templates of the corresponding requirement type, it performs the process according to at least one of the following features of the candidate template of the corresponding requirement type:
the method comprises the steps of firstly, clicking characteristics, wherein the clicking characteristics are used for representing the probability that the page of the corresponding demand type can be clicked due to the query covered by the candidate template of the corresponding demand type.
And the similarity characteristic is used for representing the common degree of one candidate template of the corresponding demand type in all the candidate templates of the corresponding demand type.
And thirdly, matching the capability characteristics, wherein the capability is used for representing the capability of the candidate template of the corresponding demand type to match the query of the corresponding demand type.
Specifically, the selecting unit 303 calculates the click characteristics of the candidate template W of the corresponding requirement type in the following manner:
where click (W) represents the click characteristic of W,representing the number of times W caused the corresponding demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
Specifically, the selecting unit 303 calculates the similarity characteristic of the candidate template W of the corresponding requirement type by using the following method:
wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the corresponding requirement type.
Specifically, the selecting unit 303 calculates the matching capability characteristics of the candidate template W of the corresponding requirement type by using the following method:
wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the corresponding requirement type,representing the number of queries W matches in a search log of queries of various demand types.
As shown in fig. 4, the demand recognition apparatus for an on-line portion includes: query acquisition unit 401 and matching unit 402. The query obtaining unit 401 is configured to obtain a user query, and the matching unit 402 is configured to determine a final template matching the user query from the final templates obtained by the apparatus for generating a requirement template, and use a requirement type corresponding to the final template matching the user query as a requirement of the user query.
The matching unit 402 may use any known matching algorithm to determine the final template matching the user query, which is not limited by the invention.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (30)
1. A method of generating a requirements template, the method comprising:
acquiring a seed query of a demand type from a search log;
generalizing the seed query of the demand type into a candidate template of the demand type;
selecting a final template of the demand type from the candidate templates of the demand type;
the step of obtaining a seed query for a demand type includes:
acquiring a preset initial seed query of the demand type;
clustering all queries recorded in the search log according to a hierarchical clustering method;
and determining a clustering hierarchy, so that the initial seed queries which are not less than a preset proportion in the hierarchy are clustered into the same class X, the total number of the queries contained in the class X in the hierarchy is minimum, and all the queries contained in the class X in the hierarchy are used as the seed queries of the demand type.
2. The method of claim 1, wherein generalizing the seed query for the demand type into the candidate template for the demand type comprises:
replacing a part, matched with a preset entity word corresponding to the demand type, in the seed query of the demand type with a wildcard character of the category to which the preset entity word belongs; or,
and replacing the part identified by the category identification function in the seed query of the demand type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
3. The method of claim 2, wherein the step of generalizing the seed query for the demand type into a template for the demand type further comprises:
and replacing the terms with the length wildcard character for limiting the length of the terms, wherein the contribution degree of the terms to the requirement type in the seed query of the requirement type is lower than the preset contribution degree requirement.
4. The method according to claim 1, wherein when selecting the final template of the requirement type from the candidate templates of the requirement type, the method is performed according to at least one of the following characteristics of the candidate template of the requirement type:
the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type;
the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type;
and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
5. The method according to claim 4, wherein the click feature of the candidate template W of the requirement type is calculated in the following way:
where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
6. The method according to claim 4, wherein the similarity characteristic of the candidate template W of the requirement type is calculated by the following method:
wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
7. The method according to claim 4, wherein the matching capability characteristic of the candidate template W of the requirement type is calculated by the following method:
wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the requirement type,representing the number of queries W matches in a search log of queries of various demand types.
8. A method of generating a requirements template, the method comprising:
acquiring a seed query of a demand type from a search log;
generalizing the seed query of the demand type into a candidate template of the demand type;
selecting a final template of the demand type from the candidate templates of the demand type;
the step of obtaining the seed query of the demand type includes:
acquiring a preset initial seed query of the demand type;
and learning a query with the similarity meeting preset requirements with the initial seed query from a search log by using an iterative learner, and taking the learned query and the initial seed query as the seed query of the demand type.
9. The method of claim 8, wherein generalizing the seed query for the demand type into the candidate template for the demand type comprises:
replacing a part, matched with a preset entity word corresponding to the demand type, in the seed query of the demand type with a wildcard character of the category to which the preset entity word belongs; or,
and replacing the part identified by the category identification function in the seed query of the demand type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
10. The method of claim 9, wherein the step of generalizing the seed query of the demand type into a template of the demand type further comprises:
and replacing the terms with the length wildcard character for limiting the length of the terms, wherein the contribution degree of the terms to the requirement type in the seed query of the requirement type is lower than the preset contribution degree requirement.
11. The method according to claim 8, wherein when selecting the final template of the requirement type from the candidate templates of the requirement type, the method is performed according to at least one of the following characteristics of the candidate template of the requirement type:
the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type;
the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type;
and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
12. The method according to claim 11, wherein the click feature of the candidate template W of the requirement type is calculated as follows:
where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
13. The method according to claim 11, wherein the similarity characteristic of the candidate template W of the requirement type is calculated by:
wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
14. The method according to claim 11, wherein the matching capability characteristic of the candidate template W of the requirement type is calculated by:
wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the requirement type,representing the number of queries W matches in a search log of queries of various demand types.
15. A method for demand identification, the method comprising:
acquiring a user query;
determining a final template matched with the user query from final templates obtained by the method for generating a requirement template as claimed in any one of claims 1 to 7 or any one of claims 8 to 14, and taking the requirement type corresponding to the final template matched with the user query as the requirement of the user query.
16. An apparatus for generating a requirements template, the apparatus comprising:
the seed acquisition unit is used for acquiring a seed query of a demand type from the search log;
a generalization unit, configured to generalize the seed query of the requirement type into a candidate template of the requirement type;
the selecting unit is used for selecting a final template of the requirement type from the candidate templates of the requirement type;
the seed acquisition unit includes:
the first selection unit is used for acquiring a preset initial seed query of the demand type;
the clustering unit is used for clustering all the queries recorded in the search log according to a hierarchical clustering method;
and the determining unit is used for determining a clustering hierarchy, so that the initial seed queries not less than a predetermined proportion under the hierarchy are clustered into the same class X, the total number of the queries contained in the class X under the hierarchy is minimum, and all the queries contained in the class X under the hierarchy are used as the seed queries of the required types.
17. The apparatus according to claim 16, wherein the generalization unit, when generalizing the seed query of the requirement type into the candidate template of the requirement type, specifically replaces a part of the seed query of the requirement type that matches a preset entity word corresponding to the requirement type with a wildcard of a category to which the preset entity word belongs; or,
and replacing the part identified by the category identification function in the seed query of the demand type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
18. The apparatus of claim 17, wherein the generalization unit is further configured to replace a term in the seed query for the requirement type whose contribution to the requirement type is lower than a preset contribution requirement with a length wildcard for limiting term length.
19. The apparatus according to claim 16, wherein the selecting unit selects the final template of the requirement type from the candidate templates of the requirement type according to at least one of the following characteristics of the candidate template of the requirement type:
the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type;
the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type;
and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
20. The apparatus according to claim 19, wherein the selecting unit calculates the click feature of the candidate template W of the requirement type by:
where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,indicating the number of times that all queries covered by W in the search log caused all pages to be clicked.
21. The apparatus according to claim 19, wherein the selecting unit calculates the similarity feature of the candidate template W of the requirement type by:
wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
22. The apparatus according to claim 19, wherein the selecting unit calculates the matching capability characteristic of the candidate template W of the requirement type by:
wherein match (W) represents the matching capability characteristic of W,representing W in said type of demandThe number of queries matched in the search log of queries,representing the number of queries W matches in a search log of queries of various demand types.
23. An apparatus for generating a requirements template, the apparatus comprising:
the seed acquisition unit is used for acquiring a seed query of a demand type from the search log;
a generalization unit, configured to generalize the seed query of the requirement type into a candidate template of the requirement type;
the selecting unit is used for selecting a final template of the requirement type from the candidate templates of the requirement type;
the seed acquisition unit includes:
the second selection unit is used for acquiring the preset initial seed query of the demand type;
and the learning unit is used for learning a query with the similarity meeting preset requirements with the initial seed query from a search log by using an iterative learner, and taking the learned query and the initial seed query as the seed query of the demand type.
24. The apparatus according to claim 23, wherein the generalization unit, when generalizing the seed query of the requirement type into the candidate template of the requirement type, specifically replaces a part of the seed query of the requirement type that matches a preset entity word corresponding to the requirement type with a wildcard of a category to which the preset entity word belongs; or,
and replacing the part identified by the category identification function in the seed query of the demand type with a wildcard of the category corresponding to the category identification function, wherein the category identification function is a function defined according to the attribute of one category and used for identifying the category.
25. The apparatus of claim 24, wherein the generalization unit is further configured to replace a term in the seed query for the requirement type whose contribution to the requirement type is lower than a preset contribution requirement with a length wildcard for limiting term length.
26. The apparatus according to claim 23, wherein the selecting unit selects the final template of the requirement type from the candidate templates of the requirement type according to at least one of the following characteristics of the candidate template of the requirement type:
the click feature is used for representing the probability that the page of the demand type can be clicked due to the query covered by the candidate template of the demand type;
the similarity characteristic is used for representing the common degree of one candidate template of the requirement type and all candidate templates of the requirement type;
and the matching capability feature is used for characterizing the capability of the candidate template of the demand type for matching the query of the demand type.
27. The apparatus according to claim 26, wherein the selecting unit calculates the click feature of the candidate template W of the requirement type by:
where click (W) represents the click characteristic of W,representing the number of times W caused the demand type page to be clicked on for all queries covered in the search log,all queries that represent coverage of W in the search log resultNumber of times all pages are clicked.
28. The apparatus according to claim 26, wherein the selecting unit calculates the similarity feature of the candidate template W of the requirement type by:
wherein, similarity (W) represents the similarity characteristic of W,representing the sum of the similarity between W and all other candidate templates of the requirement type.
29. The apparatus according to claim 26, wherein the selecting unit calculates the matching capability characteristic of the candidate template W of the requirement type by:
wherein match (W) represents the matching capability characteristic of W,representing the number of queries W matches in the search log formed by queries of the requirement type,representing the number of queries W matches in a search log of queries of various demand types.
30. An apparatus for demand identification, the apparatus comprising:
the query acquisition unit is used for acquiring user queries;
a matching unit, configured to determine a final template matching the user query from the final templates obtained by the apparatus for generating a requirement template according to any one of claims 16 to 22 or any one of claims 23 to 29, and use a requirement type corresponding to the final template matching the user query as a requirement of the user query.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110379335.5A CN103136221B (en) | 2011-11-24 | 2011-11-24 | A kind of method for generating requirement templet, demand know method for distinguishing and its device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110379335.5A CN103136221B (en) | 2011-11-24 | 2011-11-24 | A kind of method for generating requirement templet, demand know method for distinguishing and its device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103136221A CN103136221A (en) | 2013-06-05 |
CN103136221B true CN103136221B (en) | 2017-06-06 |
Family
ID=48496057
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201110379335.5A Active CN103136221B (en) | 2011-11-24 | 2011-11-24 | A kind of method for generating requirement templet, demand know method for distinguishing and its device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103136221B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104331456B (en) * | 2014-10-31 | 2018-02-09 | 百度在线网络技术(北京)有限公司 | Classification proper name method for digging and device |
CN106294476B (en) * | 2015-06-05 | 2020-10-16 | 北京搜狗科技发展有限公司 | Feature word relation obtaining method and device |
CN104991943A (en) * | 2015-07-10 | 2015-10-21 | 百度在线网络技术(北京)有限公司 | Music searching method and apparatus |
CN106919542B (en) | 2015-12-24 | 2020-04-21 | 北京国双科技有限公司 | Rule matching method and device |
CN110209780B (en) * | 2018-08-07 | 2023-03-10 | 腾讯科技(深圳)有限公司 | Question template generation method and device, server and storage medium |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101984423B (en) * | 2010-10-21 | 2012-07-04 | 百度在线网络技术(北京)有限公司 | Hot-search word generation method and system |
CN102096717B (en) * | 2011-02-15 | 2013-01-16 | 百度在线网络技术(北京)有限公司 | Search method and search engine |
-
2011
- 2011-11-24 CN CN201110379335.5A patent/CN103136221B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN103136221A (en) | 2013-06-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107862027B (en) | Retrieve intension recognizing method, device, electronic equipment and readable storage medium storing program for executing | |
CN108280114B (en) | Deep learning-based user literature reading interest analysis method | |
CN105956053B (en) | A kind of searching method and device based on the network information | |
CN108268600B (en) | AI-based unstructured data management method and device | |
CN109829104A (en) | Pseudo-linear filter model information search method and system based on semantic similarity | |
US20130060769A1 (en) | System and method for identifying social media interactions | |
WO2017070656A1 (en) | Video content retrieval system | |
CN106202294B (en) | Related news computing method and device based on keyword and topic model fusion | |
CN108509521B (en) | Image retrieval method for automatically generating text index | |
AU2011210535A1 (en) | Joint embedding for item association | |
JP2010170529A (en) | Method and system for object classification | |
WO2010014082A1 (en) | Method and apparatus for relating datasets by using semantic vectors and keyword analyses | |
CN110888991A (en) | Sectional semantic annotation method in weak annotation environment | |
CN103136221B (en) | A kind of method for generating requirement templet, demand know method for distinguishing and its device | |
CN103778206A (en) | Method for providing network service resources | |
CN106294358A (en) | The search method of a kind of information and system | |
CN113641707B (en) | Knowledge graph disambiguation method, device, equipment and storage medium | |
CN112000929A (en) | Cross-platform data analysis method, system, equipment and readable storage medium | |
WO2023057988A1 (en) | Generation and use of content briefs for network content authoring | |
CN103761286A (en) | Method for retrieving service resources on basis of user interest | |
CN106372123B (en) | Tag-based related content recommendation method and system | |
Patwardhan et al. | ViTag: Automatic video tagging using segmentation and conceptual inference | |
Manjula et al. | Visual and tag-based social image search based on hypergraph ranking method | |
Luberg et al. | Information retrieval and deduplication for tourism recommender sightsplanner | |
US12050612B2 (en) | Generation and use of topic graph for content authoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |