CN113076395A - Semantic model training, search and display method, device, equipment and storage medium - Google Patents

Semantic model training, search and display method, device, equipment and storage medium Download PDF

Info

Publication number
CN113076395A
CN113076395A CN202110320886.8A CN202110320886A CN113076395A CN 113076395 A CN113076395 A CN 113076395A CN 202110320886 A CN202110320886 A CN 202110320886A CN 113076395 A CN113076395 A CN 113076395A
Authority
CN
China
Prior art keywords
search
data
semantic model
account
target
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.)
Granted
Application number
CN202110320886.8A
Other languages
Chinese (zh)
Other versions
CN113076395B (en
Inventor
黎晓东
刘旭刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Dajia Internet Information Technology Co Ltd
Original Assignee
Beijing Dajia Internet Information Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Dajia Internet Information Technology Co Ltd filed Critical Beijing Dajia Internet Information Technology Co Ltd
Priority to CN202110320886.8A priority Critical patent/CN113076395B/en
Publication of CN113076395A publication Critical patent/CN113076395A/en
Application granted granted Critical
Publication of CN113076395B publication Critical patent/CN113076395B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3346—Query execution using probabilistic model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The present disclosure relates to a semantic model training, search and display method, device, equipment and storage medium, and relates to the technical field of networks, wherein the method comprises the following steps: firstly, acquiring a plurality of search data; wherein the search data of the plurality of search data includes: a search term and a search result corresponding to the search term. Then, selecting sample data from the plurality of search data; the characteristic value of the search result in the sample data meets a preset condition; the feature values of the search results are determined based on the interaction performed by the account with the search results. Subsequently, training to obtain a semantic model based on sample data; wherein the semantic model is used for determining the relevance between the search terms and the search results. Because the characteristic value of the search result in the sample data meets the preset condition, the semantic model obtained by training based on the sample data is more accurate, and the technical problem that the semantic model obtained by training is inaccurate due to high uncertainty and irrelevance of the sample data is solved.

Description

Semantic model training, search and display method, device, equipment and storage medium
Technical Field
The present disclosure relates to the field of network technologies, and in particular, to a semantic model training method, a semantic model searching device, a semantic model searching apparatus, and a semantic model searching storage medium.
Background
The relevancy is used for representing the matching degree between the search word and the search target (title, content, image-text information and the like), and is one of important characteristics for determining the quality of the search result.
Currently, the degree of association is usually determined by using a well-trained semantic model. However, in practical applications, different users have different search tendencies, so that the uncertainty and irrelevancy of sample data are high, and the trained semantic model (trained from a large amount of sample data) is inaccurate.
Disclosure of Invention
The invention provides a semantic model training method, a semantic model searching display device, a semantic model searching display system and a semantic model searching display storage medium, and solves the technical problem that a trained semantic model is inaccurate due to high uncertainty and irrelevance of sample data.
The technical scheme of the embodiment of the disclosure is as follows:
according to a first aspect of the embodiments of the present disclosure, a semantic model training method is provided. The semantic model training method comprises the following steps: firstly, acquiring a plurality of search data; wherein the search data of the plurality of search data includes: a search term and a search result corresponding to the search term. Then, selecting sample data from the plurality of search data; the characteristic value of the search result in the sample data meets a preset condition; the feature values of the search results are determined based on the interaction performed by the account with the search results. Subsequently, training to obtain a semantic model based on sample data; wherein the semantic model is used for determining the relevance between the search terms and the search results.
Optionally, the specific method for selecting sample data from the plurality of search data includes: performing a preprocessing operation on the search data to obtain a plurality of processed search data; a preprocessing operation for removing noise data in the search data; and selecting sample data from the plurality of processed search data.
Optionally, in a case where the search word and the search result are texts, the specific method for performing the preprocessing operation on the search data to obtain the plurality of processed search data includes: deleting the text which is the same as the search word in the text of the search result to obtain the deleted search result; replacing the search result in the search data with the deleted search result, and determining the replaced search data as the processed search data; or deleting the search data meeting the deletion condition, and determining the deleted search data as the processed search data; the deletion conditions are as follows: at least one of the number of texts of the deleted search result is lower than the preset number, the number of texts of the search word is lower than the preset number, or the text meaning of the search word is different from the text meaning of the search result.
Optionally, the characteristic values include: a positive feedback characteristic value and a negative feedback characteristic value; the specific method for selecting sample data from the plurality of processed search data includes: determining a positive feedback characteristic value and a negative feedback characteristic value of the processed search data according to the interactive operation executed by the account on the search result; and acquiring search data of which the positive feedback characteristic value and the negative feedback characteristic value meet preset conditions from the processed search data, and determining the acquired search data as sample data.
Optionally, the determining a positive feedback characteristic value and a negative feedback characteristic value of the processed search data according to the interactive operation performed by the account on the search result includes: determining a positive feedback search result and a negative feedback search result in the processed search data according to the interactive operation executed by the account on the search result; the positive feedback search result is a search result of interactive operation executed by the account history; the negative feedback search result is a search result of the account history which has not performed interactive operation; determining a first ratio, and determining the first ratio as a forward feedback characteristic value of search data corresponding to a forward feedback search result; the first proportion is the proportion of the positive feedback search result in the search results of the processed search data; determining a second ratio, and determining the second ratio as a negative feedback characteristic value of the search data corresponding to the positive feedback search result; the second percentage is a percentage of negative feedback search results in search results of the plurality of processed search data.
Optionally, the obtaining search data in which the positive feedback characteristic value and the negative feedback characteristic value satisfy the preset condition from the plurality of processed search data, and determining the obtained search data as sample data includes: acquiring forward feedback data of which the forward feedback characteristic value meets a first preset condition from the plurality of processed search data; the first preset condition is used for indicating that the probability of the processed search data being selected as the forward feedback data corresponds to the forward feedback characteristic value; acquiring negative feedback data with a negative feedback characteristic value meeting a second preset condition from the plurality of processed search data; the second preset condition is used for indicating that the probability that the processed search data is selected as negative feedback data corresponds to the negative feedback characteristic value; and determining that the sample data comprises positive feedback data and negative feedback data.
According to a second aspect of the embodiments of the present disclosure, there is provided a search display method. The search display method comprises the following steps: responding to a target search word input by an account, and acquiring a plurality of target search results corresponding to the target search word; determining the association degree between the target search word and a plurality of target search results according to a semantic model trained in advance; the semantic model is obtained by training according to any one of the semantic model training methods in the first aspect; and displaying the target search results according to the magnitude sequence of the association degree between the target search word and the target search results.
According to a third aspect of the embodiments of the present disclosure, a semantic model training apparatus is provided. The semantic model training device comprises: the device comprises an acquisition unit, a determination unit and a training unit; an acquisition unit configured to acquire a plurality of search data; the search data of the plurality of search data includes: search terms and search results corresponding to the search terms; the determining unit is used for selecting sample data from the plurality of search data acquired by the acquiring unit; the characteristic value of the search result in the sample data meets a preset condition; the characteristic value of the search result is determined according to the interactive operation performed on the search result by the account; the training unit is used for training to obtain a semantic model based on the sample data determined by the determining unit; the semantic model is used to determine the degree of association between the search terms and the search results.
Optionally, the determining unit is specifically configured to: performing a preprocessing operation on the search data to obtain a plurality of processed search data; a preprocessing operation for removing noise data in the search data; and selecting sample data from the plurality of processed search data.
Optionally, in a case that the search word and the search result are texts, the determining unit is specifically configured to: deleting the text which is the same as the search word in the text of the search result to obtain the deleted search result; replacing the search result in the search data with the deleted search result, and determining the replaced search data as the processed search data; or deleting the search data meeting the deletion condition, and determining the deleted search data as the processed search data; the deletion conditions are as follows: at least one of the number of texts of the deleted search result is lower than the preset number, the number of texts of the search word is lower than the preset number, or the text meaning of the search word is different from the text meaning of the search result.
Optionally, the characteristic values include: a positive feedback characteristic value and a negative feedback characteristic value; a determination unit, specifically configured to: determining a positive feedback characteristic value and a negative feedback characteristic value of the processed search data according to the interactive operation executed by the account on the search result; and acquiring search data of which the positive feedback characteristic value and the negative feedback characteristic value meet preset conditions from the processed search data, and determining the acquired search data as sample data.
Optionally, the determining unit is specifically configured to: determining a positive feedback search result and a negative feedback search result in the processed search data according to the interactive operation executed by the account on the search result; the positive feedback search result is a search result of interactive operation executed by the account history; the negative feedback search result is a search result of the account history which has not performed interactive operation; determining a first ratio, and determining the first ratio as a forward feedback characteristic value of search data corresponding to a forward feedback search result; the first proportion is the proportion of the positive feedback search result in the search results of the processed search data; determining a second ratio, and determining the second ratio as a negative feedback characteristic value of the search data corresponding to the positive feedback search result; the second percentage is a percentage of negative feedback search results in search results of the plurality of processed search data.
Optionally, the determining unit is specifically configured to: acquiring forward feedback data of which the forward feedback characteristic value meets a first preset condition from the plurality of processed search data; the first preset condition is used for indicating that the probability of the processed search data being selected as the forward feedback data corresponds to the forward feedback characteristic value; acquiring negative feedback data with a negative feedback characteristic value meeting a second preset condition from the plurality of processed search data; the second preset condition is used for indicating that the probability that the processed search data is selected as negative feedback data corresponds to the negative feedback characteristic value; and determining that the sample data comprises positive feedback data and negative feedback data.
According to a fourth aspect of the embodiments of the present disclosure, there is provided a search display apparatus. The search display device includes: the device comprises an acquisition unit, a determination unit and a display unit; the system comprises an acquisition unit, a search unit and a search unit, wherein the acquisition unit is used for responding to a target search word input by an account and acquiring a plurality of target search results corresponding to the target search word; the determining unit is used for determining the association degree between the target search word acquired by the acquiring unit and a plurality of target search results according to the semantic model trained in advance; the semantic model is obtained by training according to any one of the semantic model training methods in the first aspect; and the display unit is used for displaying the plurality of target search results according to the magnitude sequence of the association degrees between the target search words and the plurality of target search results determined by the determination unit.
According to a fifth aspect of the embodiments of the present disclosure, there is provided a semantic model training apparatus, which may include: a processor and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement any one of the above-described optional semantic model training methods of the first aspect.
According to a sixth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium having instructions stored thereon, which, when executed by a processor of an electronic device, enable the electronic device to perform any one of the above-mentioned optional semantic model training methods of the first aspect.
According to a seventh aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program for executing the optional semantic model training method according to any one of the first aspect by a processor.
According to an eighth aspect of the embodiments of the present disclosure, there is provided a search display apparatus, which may include: a processor and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the search display method in the second aspect described above.
According to a ninth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium having instructions stored thereon, which, when executed by a processor of an electronic device, enable the electronic device to perform the search display method of the second aspect described above.
According to a tenth aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program executed by a processor to perform the search display method as in the second aspect.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:
based on the first aspect, in the disclosure, the semantic model training device may select, from the plurality of search data, sample data in which a feature value of a search result satisfies a preset condition after acquiring the plurality of search data, and train to obtain the semantic model based on the sample data. Because the characteristic value of the search result in the sample data meets the preset condition, the semantic model obtained by training based on the sample data is more accurate, and the technical problem that the semantic model obtained by training is inaccurate due to high uncertainty and irrelevance of the sample data is solved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.
Fig. 1 is a schematic view of an application scenario provided in the embodiment of the present disclosure;
FIG. 2 is a flow chart of a semantic model training method provided by an embodiment of the present disclosure;
FIG. 3 is a flow chart illustrating a semantic model training method according to an embodiment of the present disclosure;
FIG. 4 is a flowchart illustrating a semantic model training method according to an embodiment of the disclosure;
FIG. 5 is a flowchart illustrating a semantic model training method according to an embodiment of the disclosure;
FIG. 6 is a flowchart illustrating a semantic model training method according to an embodiment of the disclosure;
FIG. 7 is a flowchart illustrating a semantic model training method according to an embodiment of the disclosure;
FIG. 8 is a flowchart illustrating a semantic model training method according to an embodiment of the disclosure;
fig. 9 is a flowchart illustrating a search display method provided by an embodiment of the present disclosure;
FIG. 10 is a schematic structural diagram illustrating a semantic model apparatus provided by an embodiment of the present disclosure;
fig. 11 is a schematic structural diagram illustrating a search display apparatus according to an embodiment of the present disclosure;
fig. 12 is a schematic structural diagram of a semantic model device or a search display device according to an embodiment of the present disclosure.
Detailed Description
In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, and/or components.
The data to which the present disclosure relates may be data that is authorized by a user or sufficiently authorized by parties.
As described in the background art, since different users have different search tendencies, the uncertainty and irrelevancy of sample data are high, which results in inaccuracy of the trained semantic model.
Based on this, the embodiment of the disclosure provides a semantic model training method, and a semantic model training device may select sample data from a plurality of search data after obtaining the plurality of search data, where a feature value of a search result meets a preset condition, and train to obtain a semantic model based on the sample data. Because the characteristic value of the search result in the sample data meets the preset condition, the semantic model obtained by training based on the sample data is more accurate, and the technical problem that the semantic model obtained by training is inaccurate due to high uncertainty and irrelevance of the sample data is solved.
In addition, after the semantic model is obtained through training, the search display device may obtain a plurality of target search results corresponding to the target search word in response to the target search word input by the account, determine the association degree between the target search word and the plurality of target search results according to the semantic model obtained through training by any one of the optional semantic model training methods in the first aspect, and display the plurality of target search results according to the magnitude sequence of the association degree between the target search word and the plurality of target search results, so that the user may quickly and accurately find the target search result corresponding to the target search word, and user experience is improved.
The semantic model training and searching display method provided by the embodiment of the disclosure is exemplarily described below with reference to the accompanying drawings:
fig. 1 is a schematic view of an application scenario provided by an embodiment of the present disclosure, as shown in fig. 1, the application scenario may include: a semantic model training device 110 and a search display device 120. Wherein, the semantic model training device 110 is in communication connection with the search display device 120.
In some embodiments, the semantic model training device 110 may be a server, a terminal, or other electronic devices for performing semantic model training, which is not limited in this disclosure.
The server may be a data server of some multimedia resource service platforms, and may be used to store and process multimedia resources. For example, the multimedia resource service platform may be a short video application service platform, a news service platform, a live broadcast service platform, a shopping service platform, a take-away service platform, a shared service platform, a functional website, and the like. The multimedia resources provided by the short video application service platform can be some short video works, the multimedia resources provided by the news service platform can be some news information, the multimedia resources provided by the live broadcast service platform can be live broadcast works and the like, and the rest are not repeated one by one. The present disclosure is not limited to a particular type of multimedia asset service platform.
In some embodiments, the server may be a single server, or may be a server cluster composed of a plurality of servers. In some embodiments, the server cluster may also be a distributed cluster. The present disclosure is also not limited to a specific implementation of the server.
In some embodiments, the search display apparatus 120 may be a display device, a terminal, or other electronic devices for performing search display, which is not limited in this disclosure.
The terminal may be a mobile phone, a tablet computer, a desktop, a laptop, a handheld computer, a notebook, an ultra-mobile personal computer (UMPC), a netbook, a cellular phone, a Personal Digital Assistant (PDA), an Augmented Reality (AR) Virtual Reality (VR) device, and other devices that can be installed and use a content community application (e.g., a fast hand), and the specific form of the electronic device is not particularly limited by the present disclosure. The system can be used for man-machine interaction with a user through one or more modes of a keyboard, a touch pad, a touch screen, a remote controller, voice interaction or handwriting equipment and the like.
In some embodiments, when the semantic model training device 110 and the search display device 120 are both terminals, the semantic model training device 110 and the search display device 120 may be independent devices or may be integrated in the same terminal, which is not specifically limited in this disclosure.
It is easy to understand that when the semantic model training device 110 and the search display device 120 are integrated in the same terminal, the communication mode between the semantic model training device 110 and the search display device 120 is the communication between the internal modules of the terminal. In this case, the communication flow between the two is the same as the "communication flow between the semantic model training device 110 and the search display device 120 when they are independent of each other".
For the convenience of understanding, the present disclosure will be mainly described by taking an example in which the semantic model training device 110 and the search display device 120 are independently provided.
The semantic model training method provided by the embodiment of the disclosure can be applied to the semantic model training device in the application scenario shown in fig. 1.
As shown in fig. 2, when the semantic model training method is applied to the semantic model training apparatus, the semantic model training method may include:
s201, the semantic model training device acquires a plurality of search data.
Wherein the search data of the plurality of search data includes: a search term and a search result corresponding to the search term.
The search result may be a multimedia resource, an account for issuing the multimedia resource, or a title of the multimedia resource, which is not limited in this disclosure.
Illustratively, taking the search result as an account (searched account) for issuing the multimedia resource as an example, the semantic model training device obtains a historical search log of each account from a server or an electronic device, and determines search data according to data in the historical search log.
Optionally, each time the account performs a search action in a historical period, the server or the electronic device records the search action in a session control (session) manner and forms a search log. The search log includes: the account name of the search account, the search term, and the account name of the searched account corresponding to the search term.
Optionally, the search log further includes: the search behavior identification (session ID), whether the searched account is exposed, whether the searched account clicks the searched account, whether the searched account focuses on the searched account, and whether the behavior of the searched account clicking the searched account is at least one of high quality clicks.
Illustratively, the search logs are shown in Table 1.
TABLE 1
Item Content providing method and apparatus
Searching account names of accounts a
Search term b
Account name of searched account c
Search behavior identification Search action 10
Whether the searched account is exposed Is that
Whether the searching account clicks the searched account Is that
Whether the searching account is concerned with the searched account Is that
Whether the behavior of clicking searched account by searching account is high-quality click Is that
Wherein, whether the searched account is exposed or not means that: after the search account inputs the search terms in the search box, the electronic device pages a plurality of searched accounts corresponding to the search terms due to the size of a display interface of the electronic device, and the searched accounts are displayed on a first page of display part. In this case, after the search account inputs the search terms in the search box, the searched account directly displayed by the electronic device is the exposed account; the electronic device needs to respond to the operation that the account clicks the second page or the next page, and the searched account which can be displayed is unexposed.
The meaning that the behavior of the search account clicking on the searched account is not a high quality click is as follows: when the searching account clicks the searched account, the searching account may click the searched account due to misoperation, and may also be attracted by the head portrait and account name of the searched account, and the searching account enters the multimedia resource publishing page of the searched account and finds out of the searched account without interest. In this case, the behavior of the search account clicking on the searched account is not a high quality click.
Accordingly, the behavior of a search account clicking on a searched account is high quality click with the implication that: the searched account clicked by the search account is the searched account which has performed the interactive operation for the history of the search account or the searched account related to the search account, and the behavior of the search account clicking the searched account is high-quality clicking. The searched account or the searched account related to the searched account, on which the interactive operation is performed in the history of the searched account, comprises: the searched account that the search account has focused on, the searched account that the search account clicks to focus on at the current time, the searched account that has an association with the search account (e.g., a friend relationship, etc.), the searched account that has a duration of browsing the multimedia resource issued by the searched account exceeds a preset duration, the searched account that has performed an interactive operation in other histories, or the searched account associated with the search account, which is not limited in this disclosure.
It should be noted that, a specific determination manner may refer to a determination manner in the prior art, and this disclosure is not repeated herein.
S202, the semantic model training device selects sample data from the plurality of search data.
The characteristic value of the search result in the sample data meets a preset condition; the feature values of the search results are determined based on the interaction performed by the account with the search results.
Specifically, after the plurality of search data are acquired, since different accounts have different search tendencies, the semantic model training device can determine the feature value of the search result in each search data, and select, from the plurality of search data, the search data whose feature value of the search result meets the preset condition as sample data. In this way, the characteristic value of the search result is determined according to the interactive operation executed by the account on the search result, and when the characteristic value of the search result in the sample data meets the preset condition, the semantic model training device trains to obtain the semantic model according to the sample data with higher relevance, so that the accuracy of the semantic model is improved.
S203, training the semantic model training device to obtain the semantic model based on the sample data.
Wherein the semantic model is used for determining the relevance between the search terms and the search results.
Specifically, after the sample data is determined, the semantic model training device trains based on the sample data to obtain the semantic model.
The semantic model is a new data model which is added with a brand-new data constructor and data processing primitives on the basis of a relational model and is used for expressing complex structures and rich semantics. In the disclosure, the semantic model training device may determine the degree of association between the search term and the search result according to the semantic model obtained by training.
Optionally, the semantic model training device may use a Bidirectional Encoding (BERT) model based on a transformation model as a model for word vector extraction, and on the basis, train to obtain the semantic model based on the determined sample data by using a network structure of a deep structure semantic based semantic (DSSM) model.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S201 to S203, after the semantic model training device acquires a plurality of search data, sample data with a feature value of a search result satisfying a preset condition may be selected from the plurality of search data, and based on the sample data, the semantic model is trained to obtain the semantic model. Because the characteristic value of the search result in the sample data meets the preset condition, the semantic model obtained by training based on the sample data is more accurate, and the technical problem that the semantic model obtained by training is inaccurate due to high uncertainty and irrelevance of the sample data is solved.
In an embodiment, referring to fig. 2 and fig. 3, in the above S202, the method for selecting sample data from a plurality of search data by the semantic model training device specifically includes S301 to S302.
S301, the semantic model training device executes preprocessing operation on the search data to obtain a plurality of processed search data.
Wherein the preprocessing operation is used to remove noisy data in the search data.
Specifically, after a plurality of search data are acquired, because a large amount of noise data is included in the search data, the semantic model training device can perform preprocessing operation on the search data to remove the noise data in the search data so as to obtain a plurality of processed search data, and the training quality of the processed search data is ensured.
S302, the semantic model training device selects sample data from the processed search data.
Specifically, after preprocessing is performed on the search data to obtain a plurality of processed search data, the semantic model training device selects sample data from the plurality of processed search data.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S301 to S302, after a plurality of search data are acquired, because the search data include a large amount of noise data, the semantic model training device can perform a preprocessing operation of removing the noise data on the search data to obtain a plurality of processed search data, thereby ensuring the training quality of the processed search data and further improving the accuracy of the semantic model.
In an embodiment, referring to fig. 3, as shown in fig. 4, in the case that the search term and the search result are texts, in S301, the method for the semantic model training device to perform a preprocessing operation on the search data to obtain a plurality of processed search data specifically includes S401-S402.
S401, deleting the text which is the same as the search word in the text of the search result by the semantic model training device to obtain the deleted search result.
For example, when the search word is "standalone game" and the search result is "standalone game one time of flight and sky", the "one time of flight and sky" has a correlation with the "standalone game" because the "one time of flight and sky" is the standalone game. In this case, the semantic model training device deletes the "one-machine game" in the search result "one-machine game one-time-of-flight" and retains the "one-time-of-flight" as the search result after deletion.
S402, replacing the search result in the search data with the deleted search result by the semantic model training device, and determining the replaced search data as the processed search data.
And after the text which is the same as the search word in the text of the search result is deleted to obtain the deleted search result, replacing the search result in the search data with the deleted search result by the semantic model training device, and determining the replaced search data as the processed search data.
For example, when the deleted search result corresponding to the search word "stand-alone game" is "one-time-of-flight", the search word in the replaced search data is "stand-alone game" and the search result is "one-time-of-flight".
The technical scheme provided by the embodiment at least has the following beneficial effects: from S401 to S402, the semantic model training device may delete the text that is the same as the search word in the text of the search result to obtain the deleted search result. Because the deleted search result has the relevance with the search word, the search result in the search data is replaced by the deleted search result, and the replaced search data is determined as the processed search data, so that the training quality of the search data is improved, and the accuracy of the semantic model can be further improved.
In an embodiment, referring to fig. 3, as shown in fig. 5, in the case that the search term and the search result are texts, in S302, the method for the semantic model training apparatus to perform a preprocessing operation on the search data to obtain a plurality of processed search data specifically includes S501.
S501, deleting the search data meeting the deletion condition by the semantic model training device, and determining the deleted search data as the processed search data.
Wherein, the deleting conditions are as follows: at least one of the number of texts of the deleted search result is lower than the preset number, the number of texts of the search word is lower than the preset number, or the text meaning of the search word is different from the text meaning of the search result.
Illustratively, when the search word is "stand-alone game" and the search result is "stand-alone game a", the semantic meaning is not rich enough because the text quantity of "a" is too short, and the relevance between "stand-alone game a" and "stand-alone game" is very high. In this case, the search data corresponding to the "stand-alone game a" is meaningless as positive feedback data, and affects the training quality of the search data as negative feedback data, and therefore, the search data is deleted.
The technical scheme provided by the embodiment at least has the following beneficial effects: as can be seen from S501, the semantic model training device may delete the search data satisfying the deletion condition, and determine the deleted search data as the processed search data. The deletion condition is at least one of the condition that the number of the deleted texts of the search result is lower than the preset number, the condition that the text meaning of the search word is lower than the preset number or the condition that the text meaning of the search word is different from the text meaning of the search result, so that the semantic model training device improves the training quality of the deleted search data and can further improve the accuracy of the semantic model.
In one embodiment, as shown in fig. 6 in conjunction with fig. 3, the characteristic values include: a positive feedback characteristic value and a negative feedback characteristic value. In the above S302, the method for selecting sample data from the plurality of processed search data specifically includes: S601-S602.
S601, the semantic model training device determines a positive feedback characteristic value and a negative feedback characteristic value of the processed search data according to interactive operation executed by the account on the search result.
Specifically, when the semantic model training device trains the semantic model based on sample data, the sample data needs to include positive feedback data and negative feedback data, so as to improve the accuracy of the semantic model. Therefore, when selecting sample data from the plurality of processed search data, the semantic model training device firstly determines the positive feedback characteristic value and the negative feedback characteristic value of the processed search data according to the interactive operation executed by the account on the search result.
Wherein the forward feedback characteristic value is used for representing the probability of the processed search data as the forward feedback data. Correspondingly, the negative feedback characteristic value is used for representing the probability that the processed search data is used as negative feedback data.
S602, the semantic model training device acquires search data of which the positive feedback characteristic value and the negative feedback characteristic value meet preset conditions from the processed search data, and determines the acquired search data as sample data.
Specifically, after determining the positive feedback characteristic value and the negative feedback characteristic value of the processed search data according to the interactive operation performed by the account on the search result, the semantic model training device acquires the search data of which the positive feedback characteristic value and the negative feedback characteristic value meet the preset condition from the plurality of processed search data, and determines the acquired search data as sample data.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S601 to S602, since the positive feedback feature value is used to indicate the probability that the processed search data is used as the positive feedback data, and the negative feedback feature value is used to indicate the probability that the processed search data is used as the negative feedback data, the semantic model training device acquires, from the processed search data, the search data whose positive feedback feature value and negative feedback feature value satisfy the preset condition as the sample data training semantic model, so that the quality of the sample data is improved, and the accuracy of the semantic model can be further improved.
In an embodiment, referring to fig. 6 and as shown in fig. 7, in the above S601, the method for determining the positive feedback feature value and the negative feedback feature value of the processed search data according to the interactive operation performed by the account on the search result by the semantic model training apparatus specifically includes: S701-S703.
S701, determining a positive feedback search result and a negative feedback search result in the processed search data by the semantic model training device according to interactive operation executed by the account on the search result.
And the positive feedback search result is a search result of interactive operation executed by the account history. And the negative feedback search result is a search result of the account history which has not performed the interactive operation.
For example, after the account enters a search term in the search box, the electronic device may display a plurality of search results. And after the account clicks the search result, the electronic equipment responds to the clicking operation of the account and judges whether the clicking operation of the account is high-quality clicking. And if the clicking operation of the account is high-quality clicking, determining the searching result clicked by the account as a forward feedback searching result. The forward feedback search result may be a search result that the account has paid attention to, or a search result that the account has clicked and paid attention to at the current time, or a search result that has a relationship with the account (e.g., a friend relationship, etc.), or a search result that a duration of browsing the search result exceeds a preset duration, or a result that an interactive operation has been performed in other histories, or a result associated with the account, which is not limited in this disclosure.
Correspondingly, the semantic model training device determines the search results of the account history which have not performed the interactive operation as negative feedback search results, namely the search results of the processed search data except the positive feedback search results.
S702, the semantic model training device determines a first ratio, and determines the first ratio as a forward feedback characteristic value of search data corresponding to a forward feedback search result.
After positive feedback search results and negative feedback search results in the processed search data are determined, the sample data determining device determines a first proportion, and determines the first proportion as a positive feedback characteristic value of the search data corresponding to the positive feedback search results. The first proportion is the proportion of the forward feedback search result in the search results of the processed search data.
Exemplary, the processed search data is shown in table 2.
TABLE 2
Account Search term Target account Whether high quality clicks
1 a x Is that
1 a y Whether or not
1 b x Whether or not
2 a x Whether or not
2 b m Whether or not
3 a x Whether or not
3 b m Whether or not
In this case, the sample data specifying means specifies the forward feedback characteristic value of the target account X under the search term a as 1/3 with the search term as the dimension. Under search term b, the forward feedback feature value for target account X is 0 (0/1).
S703, the semantic model training device determines a second ratio, and determines the second ratio as a negative feedback characteristic value of the search data corresponding to the positive feedback search result.
After determining the positive feedback search result and the negative feedback search result in the plurality of processed search data, the sample data determining device determines a second ratio, and determines the second ratio as a negative feedback characteristic value of the search data corresponding to the positive feedback search result. And the second proportion is the proportion of the negative feedback search result in the search results of the processed search data.
With reference to table 2, the sample data determining apparatus determines that the negative feedback characteristic value of the target account X under the search term a is 2/3 with the search term as the dimension. The negative feedback feature value for target account X under search term b is 1 (1/1).
It should be understood that the semantic model training device may also determine the positive feedback characteristic value and the negative feedback characteristic value of the search data corresponding to the negative feedback search result according to the method for determining the positive feedback characteristic value and the negative feedback characteristic value of the search data corresponding to the positive feedback search result, and the specific steps may refer to the above S702 and S703, which is not described in detail in this disclosure.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S701 to S703, when determining the positive feedback feature value and the negative feedback feature value of the processed search data according to the interactive operation performed by the account on the search result, the semantic model training device may determine the corresponding feature values by taking the positive feedback search result and the negative feedback search result as the percentage of the search results of the plurality of processed search data, respectively, so as to improve the quality of the sample data, and further improve the accuracy of the semantic model.
In an embodiment, with reference to fig. 7 and as shown in fig. 8, in the above S602, the method for acquiring, by the semantic model training device, search data in which a positive feedback feature value and a negative feedback feature value satisfy a preset condition from a plurality of processed search data, and determining the acquired search data as sample data specifically includes: S801-S803.
S801, the semantic model training device obtains forward feedback data with forward feedback characteristic values meeting first preset conditions from the processed search data.
Specifically, after the forward feedback characteristic value of each search data is determined, the semantic model training device acquires the forward feedback data of which the forward feedback characteristic value meets a first preset condition from the plurality of search data. And the first preset condition is used for indicating that the probability of the processed search data being selected as the forward feedback data corresponds to the forward feedback characteristic value.
Alternatively, the forward feedback feature value may be a probability that the search data is selected as the forward feedback data. In this case, the first preset condition is a probability that the search data is selected as the forward feedback data.
Illustratively, the plurality of processed search data and the forward feedback characteristic values thereof are: the forward feedback feature value of search data a is 1/5, the forward feedback feature value of search data B is 2/5, the forward feedback feature value of search data C is 3/5, the forward feedback feature value of search data D is 4/5, and the forward feedback feature value of search data E is 1 (5/5). In this case, if the semantic model needs to acquire 3 pieces of forward feedback data, the probability that search data a is selected as the forward feedback data is 20%, the probability that search data B is selected as the forward feedback data is 40%, the probability that search data C is selected as the forward feedback data is 60%, the probability that search data D is selected as the forward feedback data is 80%, and the probability that search data E is selected as the forward feedback data is 100%. And the semantic model training device selects three forward feedback data from the five search data according to the probability.
S802, the semantic model training device obtains negative feedback data with a negative feedback characteristic value meeting a second preset condition from the processed search data.
Specifically, after the negative feedback characteristic value of each search data is determined, the semantic model training device acquires negative feedback data, of which the negative feedback characteristic value satisfies a second preset condition, from the plurality of search data. And the second preset condition is used for indicating that the probability that the processed search data is selected as the negative feedback data corresponds to the negative feedback characteristic value.
Alternatively, the negative feedback feature value may be a probability that the search data is selected as the negative feedback data. In this case, the second preset condition is the probability that the search data is selected as negative feedback data.
Illustratively, the plurality of processed search data and the negative feedback characteristic values thereof are: the negative feedback feature value of the search data a is 1(5/5), the negative feedback feature value of the search data B is 4/5, the negative feedback feature value of the search data C is 3/5, the negative feedback feature value of the search data D is 2/5, and the negative feedback feature value of the search data E is 1/5. In this case, if the semantic model needs to acquire 3 pieces of negative feedback data, the probability that the search data a is selected as the negative feedback data is 100%, the probability that the search data B is selected as the negative feedback data is 80%, the probability that the search data C is selected as the negative feedback data is 60%, the probability that the search data D is selected as the negative feedback data is 40%, and the probability that the search data E is selected as the negative feedback data is 20%. And the semantic model training device selects three negative feedback data from the five search data according to the probability.
And S803, the semantic model training device determines that the sample data comprises positive feedback data and negative feedback data.
After the positive feedback data and the negative feedback data are obtained, the semantic model training device determines that the sample data comprises the positive feedback data and the negative feedback data.
Optionally, because the number of the search data is huge, the semantic model training device may summarize and select sample data including positive feedback data and negative feedback data from part of the search data.
For example, if there are 100 pieces of search data, and all the search terms in the 100 pieces of search data are different. In this case, the semantic model training apparatus may use 20 search data as a training sample set, and select 1 positive feedback data and 3 negative feedback data from the training sample set. In this case, the merged semantic model training apparatus determines that the 1 positive feedback data and the 3 negative feedback data are included in the sample data.
For example, in combination with table 2, the semantic model training apparatus may use the search word a as a dimension, and determine that one of the forward feedback data corresponding to 4 search words a is: account 1, search term a, target account x. Correspondingly, the data determining device determines that one of the data corresponding to the 4 search terms a is the forward feedback data: account 1, search term a, target account y.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S801 to S803, it can be known that the semantic model training device may determine that the sample data includes positive feedback data and negative feedback data according to the positive feedback feature values and the negative feedback feature values in the plurality of processed search data, so that the quality of the sample data is improved, and the accuracy of the semantic model may be further improved.
The search display method provided by the embodiment of the present disclosure may be applied to the search display device in the application scenario shown in fig. 1.
As shown in fig. 9, when the search display method is applied to the search display apparatus, the search display method may include: and S901-S903.
S901, the search display device responds to the target search words input by the account and obtains a plurality of target search results corresponding to the target search words.
Specifically, the account may enter the target search term in the search box when searching for the desired search result. In this case, the search display means acquires a plurality of target search results corresponding to the target search term in response to the target search term input from the account.
Illustratively, the account wants to view a stand-alone game, and thus inputs a target search word of "stand-alone game" in the search box. In this case, the search display means acquires a plurality of target search results corresponding to the "stand-alone game" in response to the account inputting the target search word of the "stand-alone game": the game is a single game with one flying sky, a single game with three games and a single game with four games.
S902, the search display device determines the association degree between the target search word and a plurality of target search results according to the semantic model trained in advance.
Wherein, the semantic model is obtained by training according to the semantic model method of any one of the above-mentioned fig. 2-fig. 8.
In connection with the above example, after acquiring a plurality of target search results corresponding to the "standalone game" (the "standalone game one time sky", "the" standalone game three "," the "standalone game four", etc.) in response to the account inputting the target search word of the "standalone game", the search display apparatus inputs the acquired target search word and the plurality of target search results corresponding to the target search word into the semantic model trained in advance. The semantic model may directly output a result, which is a degree of association between the target search term and the plurality of target search results.
And S903, the search display device displays the target search results according to the magnitude sequence of the relevance between the target search words and the target search results.
After determining the degree of association between the target search term and the plurality of target search results, the search display device displays the plurality of target search results in the order of magnitude of the degree of association between the target search term and the plurality of target search results.
The technical scheme provided by the embodiment at least has the following beneficial effects: from S901-S903, the search display device may obtain a plurality of target search results corresponding to the target search word in response to the target search word input by the account, determine the association degree between the target search word and the plurality of target search results according to a pre-trained semantic model, and display the plurality of target search results according to the magnitude order of the association degree between the target search word and the plurality of target search results, so that the user may quickly and accurately find the target search result corresponding to the target search word, and user experience is improved.
It is understood that, in practical implementation, the electronic device/server according to the embodiments of the present disclosure may include one or more hardware structures and/or software modules for implementing the corresponding semantic model training and search display methods, and these hardware structures and/or software modules may constitute an electronic device. Those of skill in the art will readily appreciate that the present disclosure can be implemented in hardware or a combination of hardware and computer software for implementing the exemplary algorithm steps described in connection with the embodiments disclosed herein. Whether a function is performed as hardware or computer software drives hardware depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.
Based on such understanding, the embodiment of the present disclosure further provides a semantic model training device, and fig. 10 shows a schematic structural diagram of the semantic model training device provided in the embodiment of the present disclosure. As shown in fig. 10, the semantic model training apparatus may include: an acquisition unit 1001, a determination unit 1002, and a training unit 1003.
An acquisition unit 1001 configured to acquire a plurality of search data; the search data of the plurality of search data includes: a search term and a search result corresponding to the search term.
A determining unit 1002, configured to select sample data from the plurality of search data acquired by the acquiring unit 1001; the characteristic value of the search result in the sample data meets a preset condition; the feature values of the search results are determined based on the interaction performed by the account with the search results.
A training unit 1003, configured to train to obtain a semantic model based on the sample data determined by the determining unit 1002; the semantic model is used to determine the degree of association between the search terms and the search results.
Optionally, the determining unit 1002 is specifically configured to: performing a preprocessing operation on the search data to obtain a plurality of processed search data; a preprocessing operation for removing noise data in the search data; and selecting sample data from the plurality of processed search data.
Optionally, in a case that the search word and the search result are texts, the determining unit 1002 is specifically configured to: deleting the text which is the same as the search word in the text of the search result to obtain the deleted search result; replacing the search result in the search data with the deleted search result, and determining the replaced search data as the processed search data; or deleting the search data meeting the deletion condition, and determining the deleted search data as the processed search data; the deletion conditions are as follows: at least one of the number of texts of the deleted search result is lower than the preset number, the number of texts of the search word is lower than the preset number, or the text meaning of the search word is different from the text meaning of the search result.
Optionally, the characteristic values include: a positive feedback characteristic value and a negative feedback characteristic value; the determining unit 1002 is specifically configured to: determining a positive feedback characteristic value and a negative feedback characteristic value of the processed search data according to the interactive operation executed by the account on the search result; and acquiring search data of which the positive feedback characteristic value and the negative feedback characteristic value meet preset conditions from the processed search data, and determining the acquired search data as sample data.
Optionally, the determining unit 1002 is specifically configured to: determining a positive feedback search result and a negative feedback search result in the processed search data according to the interactive operation executed by the account on the search result; the positive feedback search result is a search result of interactive operation executed by the account history; the negative feedback search result is a search result of the account history which has not performed interactive operation; determining a first ratio, and determining the first ratio as a forward feedback characteristic value of search data corresponding to a forward feedback search result; the first proportion is the proportion of the positive feedback search result in the search results of the processed search data; determining a second ratio, and determining the second ratio as a negative feedback characteristic value of the search data corresponding to the positive feedback search result; the second percentage is a percentage of negative feedback search results in search results of the plurality of processed search data.
Optionally, the determining unit 1002 is specifically configured to: acquiring forward feedback data of which the forward feedback characteristic value meets a first preset condition from the plurality of processed search data; the first preset condition is used for indicating that the probability of the processed search data being selected as the forward feedback data corresponds to the forward feedback characteristic value; acquiring negative feedback data with a negative feedback characteristic value meeting a second preset condition from the plurality of processed search data; the second preset condition is used for indicating that the probability that the processed search data is selected as negative feedback data corresponds to the negative feedback characteristic value; and determining that the sample data comprises positive feedback data and negative feedback data.
The embodiment of the disclosure also correspondingly provides a search display device, and fig. 11 shows a schematic structural diagram of the search display device provided by the embodiment of the disclosure. As shown in fig. 11, the search display apparatus may include: an acquisition unit 1101, a determination unit 1102, and a display unit 1103.
An obtaining unit 1101 is configured to obtain a plurality of target search results corresponding to a target search term in response to the target search term input by the account.
A determining unit 1102, configured to determine, according to the semantic model trained in advance, a degree of association between the target search term acquired by the acquiring unit 1101 and a plurality of target search results; the semantic model is obtained by training according to the training method of the semantic model in any one of the above-mentioned fig. 2-fig. 8.
A display unit 1103, configured to display the multiple target search results in order of magnitude of the degree of association between the target search term and the multiple target search results determined by the determination unit 1102.
As described above, the embodiments of the present disclosure may perform functional module division on a semantic model training device or a search display device according to the above method examples. The integrated module can be realized in a hardware form, and can also be realized in a software functional module form. In addition, it should be further noted that the division of the modules in the embodiments of the present disclosure is schematic, and is only a logic function division, and there may be another division manner in actual implementation. For example, the functional blocks may be divided for the respective functions, or two or more functions may be integrated into one processing block.
The specific manner in which each module executes the operation and the beneficial effects of the semantic model training device or the search display device in the foregoing embodiments have been described in detail in the foregoing method embodiments, and are not described again here.
Fig. 12 is a schematic structural diagram of another semantic model training device or search display device provided by the present disclosure. As shown in fig. 12, the determining means 15 may comprise at least one processor 151 and a memory 153 for storing processor-executable instructions. Wherein the processor 151 is configured to execute instructions in the memory 153 to implement the classification method in the above-described embodiments.
In addition, the determination means 15 may further comprise a communication bus 152 and at least one communication interface 154.
Processor 151 may be a Central Processing Unit (CPU), a micro-processing unit, an ASIC, or one or more integrated circuits for controlling the execution of programs in accordance with the disclosed aspects.
The communication bus 152 may include a path that conveys information between the aforementioned components.
Communication interface 154, using any transceiver or the like, may be used to communicate with other devices or communication networks, such as an ethernet, a Radio Access Network (RAN), a Wireless Local Area Network (WLAN), etc.
The memory 153 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), magnetic disc storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to these. The memory may be self-contained and connected to the processing unit by a bus. The memory may also be integrated with the processing unit.
The memory 153 is used for storing instructions for executing the disclosed solution, and is controlled by the processor 151. The processor 151 is configured to execute instructions stored in the memory 153 to implement the functions of the disclosed method.
In particular implementations, processor 151 may include one or more CPUs such as CPU0 and CPU1 of fig. 12 for one embodiment.
In particular implementations, determining means 15 may include a plurality of processors, such as processor 151 and processor 157 in fig. 12, for example, as an embodiment. Each of these processors may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. A processor herein may refer to one or more devices, circuits, and/or processing cores for processing data (e.g., computer program instructions).
In one embodiment, the determining apparatus 15 may further include an output device 155 and an input device 156. The output device 155 is in communication with the processor 151 and may display information in a variety of ways. For example, the output device 155 may be a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display device, a Cathode Ray Tube (CRT) display device, a projector (projector), or the like. The input device 156 is in communication with the processor 151 and can accept user input in a variety of ways. For example, the input device 156 may be a mouse, keyboard, touch screen device, or sensing device, among others.
It will be appreciated by those skilled in the art that the arrangement shown in figure 12 is not intended to be limiting of the determination means 15 and may comprise more or fewer components than those shown, or some components may be combined, or a different arrangement of components may be used.
Additionally, the present disclosure also provides a computer-readable storage medium comprising instructions that, when executed by a processor, cause the processor to perform the semantic model training or search display method as provided by the above embodiments.
Additionally, the present disclosure also provides a computer program product comprising instructions that, when executed by a processor, cause the processor to perform the semantic model training or search display method as provided by the embodiments above.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims (10)

1. A semantic model training method is characterized by comprising the following steps:
acquiring a plurality of search data; search data of the plurality of search data includes: search terms and search results corresponding to the search terms;
selecting sample data from the plurality of search data; the characteristic value of the search result in the sample data meets a preset condition; the characteristic value of the search result is determined according to the interactive operation performed on the search result by an account;
training to obtain a semantic model based on the sample data; the semantic model is used for determining the association degree between the search word and the search result.
2. A search display method, comprising:
responding to a target search word input by an account, and acquiring a plurality of target search results corresponding to the target search word;
determining the association degree between the target search word and the plurality of target search results according to a semantic model trained in advance; the semantic model is obtained by training according to the semantic model training method of claim 1;
and displaying the target search results according to the magnitude sequence of the association degrees between the target search terms and the target search results.
3. A semantic model training apparatus, comprising: the device comprises an acquisition unit, a determination unit and a training unit;
the acquisition unit is used for acquiring a plurality of search data; search data of the plurality of search data includes: search terms and search results corresponding to the search terms;
the determining unit is configured to select sample data from the plurality of search data acquired by the acquiring unit; the characteristic value of the search result in the sample data meets a preset condition; the characteristic value of the search result is determined according to the interactive operation performed on the search result by an account;
the training unit is used for training to obtain a semantic model based on the sample data determined by the determining unit; the semantic model is used for determining the association degree between the search word and the search result.
4. A search display device, comprising: the device comprises an acquisition unit, a determination unit and a display unit;
the acquisition unit is used for responding to a target search word input by an account and acquiring a plurality of target search results corresponding to the target search word;
the determining unit is used for determining the association degrees between the target search words acquired by the acquiring unit and the plurality of target search results according to a semantic model trained in advance; the semantic model is obtained by training according to the semantic model training method of claim 1;
the display unit is configured to display the plurality of target search results according to the magnitude order of the association degrees between the target search terms and the plurality of target search results determined by the determination unit.
5. A semantic model training device, comprising:
a processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to execute the instructions to implement the semantic model training method of claim 1.
6. A computer-readable storage medium having instructions stored thereon, wherein the instructions in the computer-readable storage medium, when executed by a processor of a computer, enable the computer to perform the semantic model training method of claim 1.
7. A computer program product comprising a computer program, characterized in that the computer program realizes the semantic model training method of claim 1 when executed by a processor.
8. A search display device, characterized in that the search display device comprises:
a processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to execute the instructions to implement the search display method of claim 2.
9. A computer-readable storage medium having instructions stored thereon, wherein the instructions in the computer-readable storage medium, when executed by a processor of a computer, enable the computer to perform the search display method of claim 2.
10. A computer program product comprising a computer program, characterized in that the computer program realizes the search display method of claim 2 when executed by a processor.
CN202110320886.8A 2021-03-25 2021-03-25 Semantic model training and search display method, device, equipment and storage medium Active CN113076395B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110320886.8A CN113076395B (en) 2021-03-25 2021-03-25 Semantic model training and search display method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110320886.8A CN113076395B (en) 2021-03-25 2021-03-25 Semantic model training and search display method, device, equipment and storage medium

Publications (2)

Publication Number Publication Date
CN113076395A true CN113076395A (en) 2021-07-06
CN113076395B CN113076395B (en) 2024-03-26

Family

ID=76610319

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110320886.8A Active CN113076395B (en) 2021-03-25 2021-03-25 Semantic model training and search display method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN113076395B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113988157A (en) * 2021-09-30 2022-01-28 北京百度网讯科技有限公司 Semantic retrieval network training method and device, electronic equipment and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170300828A1 (en) * 2016-04-14 2017-10-19 Yahoo! Inc. Method and system for distributed machine learning
CN109508394A (en) * 2018-10-18 2019-03-22 青岛聚看云科技有限公司 A kind of training method and device of multi-medium file search order models
CN110162593A (en) * 2018-11-29 2019-08-23 腾讯科技（深圳）有限公司 A kind of processing of search result, similarity model training method and device
CN110909182A (en) * 2019-11-29 2020-03-24 北京达佳互联信息技术有限公司 Multimedia resource searching method and device, computer equipment and storage medium
CN111881343A (en) * 2020-07-07 2020-11-03 Oppo广东移动通信有限公司 Information pushing method and device, electronic equipment and computer readable storage medium

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170300828A1 (en) * 2016-04-14 2017-10-19 Yahoo! Inc. Method and system for distributed machine learning
CN109508394A (en) * 2018-10-18 2019-03-22 青岛聚看云科技有限公司 A kind of training method and device of multi-medium file search order models
CN110162593A (en) * 2018-11-29 2019-08-23 腾讯科技（深圳）有限公司 A kind of processing of search result, similarity model training method and device
CN110909182A (en) * 2019-11-29 2020-03-24 北京达佳互联信息技术有限公司 Multimedia resource searching method and device, computer equipment and storage medium
CN111881343A (en) * 2020-07-07 2020-11-03 Oppo广东移动通信有限公司 Information pushing method and device, electronic equipment and computer readable storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113988157A (en) * 2021-09-30 2022-01-28 北京百度网讯科技有限公司 Semantic retrieval network training method and device, electronic equipment and storage medium
CN113988157B (en) * 2021-09-30 2023-10-13 北京百度网讯科技有限公司 Semantic retrieval network training method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
CN113076395B (en) 2024-03-26

Similar Documents

Publication Publication Date Title
CN107609152B (en) Method and apparatus for expanding query expressions
CN107256267B (en) Query method and device
CN107784010B (en) Method and equipment for determining popularity information of news theme
CN110647614A (en) Intelligent question and answer method, device, medium and electronic equipment
KR102281186B1 (en) Animated snippets for search results
US20150278359A1 (en) Method and apparatus for generating a recommendation page
CN110069698B (en) Information pushing method and device
CN106407377B (en) Search method and device based on artificial intelligence
CN114861889B (en) Deep learning model training method, target object detection method and device
US20240086479A1 (en) Identification and Issuance of Repeatable Queries
US20200159765A1 (en) Performing image search using content labels
CN110008396B (en) Object information pushing method, device, equipment and computer readable storage medium
CN111105294A (en) VR navigation method, system, client, server and storage medium thereof
CN111159431A (en) Knowledge graph-based information visualization method, device, equipment and storage medium
CN114036322A (en) Training method for search system, electronic device, and storage medium
CN112328889A (en) Method and device for determining recommended search terms, readable medium and electronic equipment
CN111538830B (en) French searching method, device, computer equipment and storage medium
CN113806660A (en) Data evaluation method, training method, device, electronic device and storage medium
KR101931859B1 (en) Method for selecting headword of electronic document, method for providing electronic document, and computing system performing the same
US9152698B1 (en) Substitute term identification based on over-represented terms identification
CN113076395B (en) Semantic model training and search display method, device, equipment and storage medium
CN112231444A (en) Processing method and device for corpus data combining RPA and AI and electronic equipment
CN106570003B (en) Data pushing method and device
CN111970327A (en) News spreading method and system based on big data processing
CN110659419A (en) Method for determining target user and related device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant