CN111930954B - Intention recognition method and device, storage medium and electronic equipment - Google Patents
Intention recognition method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN111930954B CN111930954B CN202010996810.2A CN202010996810A CN111930954B CN 111930954 B CN111930954 B CN 111930954B CN 202010996810 A CN202010996810 A CN 202010996810A CN 111930954 B CN111930954 B CN 111930954B
- Authority
- CN
- China
- Prior art keywords
- information
- target
- user
- intention
- search
- 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
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
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- 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)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present disclosure relates to an intention recognition method, apparatus, storage medium, and electronic device, the method comprising: acquiring retrieval information input by a target user; determining first characteristic information of the retrieval information, wherein the first characteristic information is used for representing the degree of correlation between the retrieval information and the target intention in the aspect of user searching behaviors, and determining second characteristic information of the retrieval information, and the second characteristic information is used for representing the degree of correlation between the retrieval information and the target intention in the aspect of semantics; and identifying whether the retrieval information represents the target intention or not according to the first characteristic information and the second characteristic information. By the technical scheme, the intention recognition can be carried out from the two aspects of the user searching behavior and the semantics, the two aspects of the user searching behavior and the semantics are combined for the intention recognition, the accuracy and the recognition success rate of the intention recognition can be effectively improved, the provided searching result is the content meeting the real searching intention of the user, and the accuracy of the searching result provided for the user is ensured.
Description
Technical Field
The present disclosure relates to the field of search technologies, and in particular, to an intention recognition method, an intention recognition apparatus, a storage medium, and an electronic device.
Background
In the search field, a user can input retrieval information for information query and search, and the retrieval information can comprise retrieval words or retrieval sentences. In a search scenario, intention recognition refers to understanding of search information input by a user, which kind or kinds of results the user wants to search is further determined by understanding the search information input by the user to recognize the search intention of the user, and relevant search results are provided to the user according to the recognized search intention.
The same retrieval information may represent a plurality of different intentions, if the search intention of the user cannot be accurately identified, a search result meeting the requirement of the user cannot be provided, and the provided search result may not be the content desired by the user, thereby affecting the accuracy of the search result. Therefore, the recognition of the search intention of the user is an important link in the search field.
Disclosure of Invention
The purpose of the present disclosure is to provide an intention identification method, apparatus, storage medium, and electronic device, which can identify an intention from two aspects of a user search behavior and semantics, and can effectively improve accuracy and success rate of intention identification.
In order to achieve the above object, in a first aspect, the present disclosure provides an intention identifying method, the method including: acquiring retrieval information input by a target user; determining first characteristic information of the retrieval information, wherein the first characteristic information is used for representing the degree of correlation between the retrieval information and the target intention in terms of user search behaviors, and determining second characteristic information of the retrieval information, wherein the second characteristic information is used for representing the degree of correlation between the retrieval information and the target intention in terms of semantics; and identifying whether the retrieval information represents the target intention or not according to the first characteristic information and the second characteristic information.
Optionally, the determining first feature information of the retrieval information includes: inputting the retrieval information and the position information of the target user into a user search intention recognition model, so that the user search intention recognition model determines target related information between the retrieval information and the target intention in terms of user search behaviors according to the retrieval information, the position information and user intention recognition characteristic information of the retrieval information in a target area indicated by the position information, wherein the user intention recognition characteristic information is determined according to historical search behaviors of the user on the retrieval information, and the user intention recognition characteristic information is updated according to a preset period; and under the condition that the target related information is determined by the user search intention recognition model, determining the target related information as the first characteristic information.
Optionally, the method further comprises: and under the condition that the target related information is not determined by the user search intention identification model, determining the first characteristic information as a preset value.
Optionally, the target intent comprises a search merchant; the user intent identification feature information comprises at least one of: the method comprises the following steps of obtaining information related to operation of a user in a target area for a target merchant, obtaining information related to search conditions of users in a plurality of areas for the search information, obtaining correlation information between the search information and merchant feature information of the target merchant, obtaining information related to operation of the user in the target area for a preset merchant, and obtaining information related to entity link of the search information, wherein the target merchant comprises each merchant in the target area determined according to the search information, and the preset merchant comprises chained merchants in the target area.
Optionally, the determining second characteristic information of the retrieval information includes: determining the second characteristic information in one of the following ways: inputting the retrieval information into a semantic intention recognition model to obtain the second characteristic information output by the semantic intention recognition model; and obtaining the second characteristic information of the retrieval information from a model cache result under the condition that the second characteristic information exists in the model cache result, wherein the model cache result is determined according to the historical output result of the semantic intention recognition model.
Optionally, the identifying whether the retrieval information represents the target intention according to the first feature information and the second feature information includes: determining target characteristic information according to the first characteristic information, a first preset weight corresponding to the first characteristic information, the second characteristic information and a second preset weight corresponding to the second characteristic information; and determining that the retrieval information is characterized by the target intention under the condition that the target characteristic information is larger than a preset threshold value.
Optionally, the method further comprises: and determining display information to be displayed to the target user according to the intention identification result of the retrieval information.
In a second aspect, the present disclosure provides an intent recognition apparatus, the apparatus comprising: the retrieval information acquisition module is configured to acquire retrieval information input by a target user; a feature information determination module configured to determine first feature information of the retrieval information, wherein the first feature information is used for representing a degree of correlation between the retrieval information and a target intention in terms of user search behavior, and determine second feature information of the retrieval information, wherein the second feature information is used for representing a degree of correlation between the retrieval information and the target intention in terms of semantics; a target intention identifying module configured to identify whether the retrieval information represents the target intention according to the first characteristic information and the second characteristic information.
Optionally, the characteristic information determining module includes: a first input sub-module configured to input the retrieval information and the location information of the target user into a user search intention recognition model to determine, by the user search intention recognition model, target related information between the retrieval information and the target intention in terms of user search behavior according to the retrieval information, the location information, and user intention recognition feature information of the retrieval information in a target area indicated by the location information, wherein the user intention recognition feature information is determined according to historical search behavior of a user for the retrieval information, and the user intention recognition feature information is updated according to a preset period; a first determination sub-module configured to determine the target-related information as the first feature information if the target-related information is determined by the user search intention recognition model.
Optionally, the apparatus further comprises: a determination module configured to determine the first feature information as a preset value if the target related information is not determined by the user search intention recognition model.
Optionally, the feature information determining module determines the second feature information by using one of a second input submodule and an obtaining submodule: a second input submodule configured to input the retrieval information into a semantic intention recognition model, resulting in the second feature information output by the semantic intention recognition model; an obtaining sub-module configured to obtain the second feature information from a model cache result if the second feature information of the retrieval information exists in the model cache result, the model cache result being determined according to a historical output result of the semantic intent recognition model.
Optionally, the target intent recognition module includes: a second determining submodule configured to determine target feature information according to the first feature information, a first preset weight corresponding to the first feature information, the second feature information, and a second preset weight corresponding to the second feature information; a third determining sub-module configured to determine that the retrieval information is characterized by the target intention if the target characteristic information is greater than a preset threshold.
Optionally, the apparatus further comprises: and the display information determining module is configured to determine display information to be displayed to the target user according to the intention recognition result of the retrieval information.
In a third aspect, the present disclosure provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method provided by the first aspect of the present disclosure.
In a fourth aspect, the present disclosure provides an electronic device comprising: a memory having a computer program stored thereon; a processor for executing the computer program in the memory to implement the steps of the method provided by the first aspect of the present disclosure.
Through the technical scheme, after the retrieval information input by the target user is acquired, the first characteristic information of the retrieval information can be determined, and the second characteristic information of the retrieval information can be determined, wherein the first characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of user searching behaviors, and the second characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of semantics. Meanwhile, whether the retrieval information representation is a target intention is identified according to the first characteristic information and the second characteristic information, and intention identification can be carried out from two aspects of user search behavior and semantics. The method has the advantages that the influence of the semantics of the retrieval information is small in the aspect of user searching behaviors, the searching intention of the target user can be accurately identified under the condition that the searching amount of the retrieval information is large, and the searching intention of the target user can be accurately identified based on the semantics of the retrieval information under the condition that the searching amount of the retrieval information is small or the retrieval information is not searched. Therefore, the limitation problem caused by identifying the search intention of the user only from the aspect of user search behavior or only from the aspect of semantics is avoided, the intention identification is carried out by combining the two aspects of user search behavior and semantics, the accuracy and the identification success rate of the intention identification can be effectively improved, the provided search result is the content meeting the real search intention of the user, the accuracy of the search result provided for the user is ensured, and the user experience is improved.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure without limiting the disclosure.
FIG. 1 is a flow chart illustrating a method of intent recognition according to an example embodiment.
Fig. 2 is a flowchart illustrating a method of determining first characteristic information of retrieved information according to an example embodiment.
FIG. 3 is a flow chart illustrating a method of intent recognition according to another exemplary embodiment.
FIG. 4 is a block diagram illustrating an intent recognition apparatus according to an example embodiment.
FIG. 5 is a block diagram illustrating an electronic device in accordance with an example embodiment.
Detailed Description
The following detailed description of specific embodiments of the present disclosure is provided in connection with the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present disclosure, are given by way of illustration and explanation only, not limitation.
FIG. 1 is a flow chart illustrating an intent recognition method that may be applied to an electronic device having processing capabilities, such as a server, as shown in FIG. 1, and that may include S101-S103, according to an exemplary embodiment.
In S101, retrieval information input by a target user is acquired.
The search information may be a search word or a search sentence that is input when the target user wants to perform an information search. The target user can input the retrieval information through the terminal equipment such as a mobile phone and a computer used by the target user, and when the method provided by the disclosure is applied to the server, the terminal equipment can send the retrieval information input by the target user to the server, so that the server can obtain the retrieval information.
In S102, first characteristic information of the search information is determined, and second characteristic information of the search information is determined.
The first feature information may be used to characterize a degree of correlation between the retrieval information and the target intention in terms of user search behavior, and the second feature information may be used to characterize a degree of correlation between the retrieval information and the target intention in terms of semantics.
Wherein the target intent may refer to a purpose of the search. For example, at a shopping platform or a takeaway platform, the target intent may include searching for merchants, searching for items, searching for meals, and the like. By way of further example, in searching for media files, the target intent may include searching for songs, searching for videos, searching for singers, and so forth.
The degree of correlation between the search information and the target intention may refer to a probability or likelihood that the target user inputs the search information for the target intention. For example, the target intention is a search merchant, and the degree of correlation between the retrieved information and the target intention may refer to the possibility that the target user inputs the purpose of the retrieved information for searching the merchant.
The user searching behavior can comprise the searching condition of the user on the searching information, the clicking behavior of the user on the result searched according to the searching information and the like, wherein the user can refer to the user who has performed the searching behavior and can comprise the target user, the user can click and look up the result which is interested in the user and meets the requirement of the user, and the characteristics obtained by statistics and analysis based on the user searching behavior can reflect the searching intention and the searching requirement of most users. The first characteristic information determined by the server can be used for representing the degree of correlation between the retrieval information and the target intention in terms of the user searching behavior, and the degree of correlation can refer to the probability or possibility that the retrieval information is represented as the target intention from the perspective of the user searching behavior.
The semantic meaning refers to a literal meaning or a literal meaning represented by the retrieval information, and the second characteristic information determined by the server can be used for representing the correlation degree between the retrieval information and the target intention in the aspect of the semantic meaning, wherein the correlation degree can refer to the probability or possibility that the retrieval information is represented as the target intention from the semantic perspective.
In S103, it is identified whether the search information represents the target intention or not based on the first feature information and the second feature information.
The inventor finds that the intention recognition only from the aspect of user search behaviors or only from the aspect of semantics has certain limitations in the research process. The intention recognition is performed from the aspect of user search behavior, and the influence of the semantics of the retrieval information is small. For example, taking a target intention as a search merchant as an example, the retrieval information input by the target user includes a word a, the word a is not a merchant but rather the user searches many known merchants, if the word a is identified only from the semantic aspect, it may not be identified that the target user is a merchant that the target user wants to search, and from the aspect of user search behavior, based on the characteristics of the historical search behavior statistics of the user, it may be identified that the word a input by the target user is intended to search the merchant, so that the search intention of the target user may be accurately identified.
However, in the intention recognition from the aspect of user search behavior, depending on the features counted according to the historical search behavior of the user, the search information needs to appear a certain frequency, that is, the search amount of the search information is enough to have statistical significance, and if the search amount of the search information is small or the search information is never searched, the corresponding features may not be accurately counted according to the historical search behavior of the user on the search information. Therefore, if the search volume of the retrieved information is small or not searched, it may not be possible to accurately identify whether the retrieved information represents the target intention from the aspect of the user search behavior.
Therefore, the method only identifies the search intention of the user from the aspect of user search behavior or only identifies the search intention of the user from the aspect of semantics, has certain limitations, and cannot accurately identify whether the retrieval information representation input by the target user is the target intention.
In view of this, in the present disclosure, it is possible to identify whether or not the retrieval information input by the target user is characterized as the target intention, based on both the first feature information for characterizing the degree of correlation between the retrieval information in terms of the user search behavior and the target intention and the second feature information for characterizing the degree of correlation between the retrieval information in terms of the semantics and the target intention, from both the user search behavior and the semantics. The method has the advantages that the influence of the semantics of the retrieval information is small in the aspect of user searching behaviors, and the searching intention of a target user can be accurately identified under the condition that the searching amount of the retrieval information is large; in terms of semantics, in the case where the search amount of the retrieval information is small or the retrieval information is not searched, the search intention of the target user can be accurately identified based on the semantics of the retrieval information. Therefore, the intention recognition is carried out by combining two aspects of user searching behaviors and semantics, and the accuracy and the recognition success rate of the intention recognition can be effectively improved.
Through the technical scheme, after the retrieval information input by the target user is acquired, the first characteristic information of the retrieval information can be determined, and the second characteristic information of the retrieval information can be determined, wherein the first characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of user searching behaviors, and the second characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of semantics. Meanwhile, whether the retrieval information representation is a target intention is identified according to the first characteristic information and the second characteristic information, and intention identification can be carried out from two aspects of user search behavior and semantics. The method has the advantages that the influence of the semantics of the retrieval information is small in the aspect of user searching behaviors, the searching intention of the target user can be accurately identified under the condition that the searching amount of the retrieval information is large, and the searching intention of the target user can be accurately identified based on the semantics of the retrieval information under the condition that the searching amount of the retrieval information is small or the retrieval information is not searched. Therefore, the limitation problem caused by identifying the search intention of the user only from the aspect of user search behavior or only from the aspect of semantics is avoided, the intention identification is carried out by combining the two aspects of user search behavior and semantics, the accuracy and the identification success rate of the intention identification can be effectively improved, the provided search result is the content meeting the real search intention of the user, the accuracy of the search result provided for the user is ensured, and the user experience is improved.
In the present disclosure, an exemplary embodiment of determining the first feature information of the search information in S102 may include S201 to S204 as shown in fig. 2.
In S201, the retrieval information and the location information of the target user are input into the user search intention recognition model to determine target related information between the retrieval information and the target intention in terms of the user search behavior by the user search intention recognition model according to the retrieval information, the location information, and the user intention recognition feature information of the retrieval information in the target area indicated by the location information.
The location information of the target user may be identification information of a target area where the target user is located. For example, the area is divided by cities, that is, one city is one area, the target area where the target user is located may be the city where the target user is located, and the location information may be a city ID of the city where the target user is located.
When the server acquires the retrieval information input by the target user and the position information of the target user, the retrieval information and the position information can be input into the user search intention recognition model. For example, the user search intention recognition model may be trained in advance, and any model may be adopted, for example, an XGBoost (eXtreme Gradient Boosting) model.
In the present disclosure, for each piece of information searched by the user, the user intention identification feature information of the information searched by the user in each area may be counted in advance and stored in the storage module of the server. The user intention recognition feature information may be determined according to historical search behaviors of the users of the area for the information, for example, according to the search behaviors of the users in the last preset number of days. Moreover, the user intention identification feature information can be updated according to a preset period, and the preset period can be preset, for example, one day or two days, so as to ensure the timeliness of updating.
The user search intention identification model can acquire user intention identification characteristic information of the search information in the target area counted in advance from the storage module, and determine target related information between the search information and the target intention in terms of user search behaviors according to the search information, the position information and the user intention identification characteristic information.
In the present disclosure, the user intention recognition feature information may be multi-dimensional feature information to ensure comprehensiveness of the user intention recognition feature information. For example, taking the target intent as including a search merchant as an example, retrieving the user intent identification feature information of the information in the target area may include at least one of: the information processing method comprises the steps of obtaining information related to operation of users in a target area for a target merchant, information related to search conditions of users in multiple areas for the retrieved information, correlation information between the retrieved information and merchant feature information of the target merchant, information related to operation of the users in the target area for a preset merchant, and information related to entity link of the retrieved information.
The target merchants may include merchants in a target area determined according to the retrieval information, for example, the retrieval information includes a word a, and the target merchants may include merchants related to the word a in a city where the target user is located. The operation of the user in the target area for the target business may include a click operation, a browsing operation, and the like of the user in the city where the target user is located for the target business. For example, the information related to the operation of the user of the target area for the target merchant may include: the click rate and the click occupation ratio corresponding to the target merchants with the top M points of click amount of the user, the sum of the click rate and the click occupation ratio corresponding to all the target merchants, the kini coefficient and the information entropy calculated according to the target merchants clicked by the user, and the like. M may be a positive integer greater than or equal to 1, and M may have a plurality of values, for example, the target merchants with the top M bits of the user click rate may include the target merchant with the first click rate, the target merchants with the top three bits, and the target merchants with the top fifteen bits, for example. By counting the information related to the operation of the user in the target area for the target business, the concentration degree of the user clicks and the degree of the target business concerned by the user can be represented, and the search intention of most users can be reflected.
The information related to the search situation of the retrieval information by the users of the plurality of areas may include: the number of areas in which the user searched for the search information, the sum of the search amounts for the search information by the users of the plurality of areas, the average search amount for the search information by the users of the plurality of areas, the sum of the click amounts for the search results corresponding to the search information by the users of the plurality of areas, the average click amount for the search results corresponding to the search information by the users of the plurality of areas, and so on.
The merchant characteristic information may include a merchant name, a commodity name provided by the merchant, a service name provided by the merchant, and the like. The correlation information between the retrieved information and the merchant characteristic information of the target merchant may include: the information may include, for example, text-related information between the retrieved information and the merchant name, whether the retrieved information is included in a product name provided by the merchant, whether the retrieved information is included in a service name provided by the merchant, whether the retrieved information is included in a landmark of a location where the merchant is located, whether the retrieved information is included in a political region name of the location where the merchant is located, the number of times the retrieved information appears in merchant characteristic information, and so forth. Wherein retrieving textual relevant information between the information and the merchant name may include: the method comprises the steps of retrieving the character length of information, removing brackets of merchant names and the average character length after description in the brackets, taking the longest continuous sub-string of the retrieved information into the character length of the retrieved information, taking the longest continuous sub-string of the merchant names into the character length of the merchant names, taking the longest continuous sub-string of the retrieved information out of the preset number of characters, taking the longest continuous sub-string of the merchant names into the preset number of characters, taking the number of repeated characters in the retrieved information into the character length of the retrieved information, taking the number of repeated characters in the merchant names into the character length of the merchant names.
The preset merchants may include chained merchants in the target area, such as merchants with a chained pattern in a city where the target user is located. The information related to the operation of the user of the target area for the preset merchant may include: the sum of the click amounts of the user to all the preset merchants, the sum of the click amounts of the user to the preset merchants with the click amounts being the preset digits, the click amount and the click ratio of the user to the preset merchant with the largest click amount, the number of the preset merchants with the click ratio being greater than the preset percentage, and the like.
Information related to the entity link from which the information is retrieved may include: the entity link is whether to link the retrieved information to the target merchant, the target merchant's rating information, and so on.
In the disclosure, the user intention identification characteristic information is multidimensional information about user operation behaviors, entity connection and the like, so that the considered user intention identification characteristic information is more comprehensive, and the user search intention identification model can more accurately determine the target related information between the retrieval information and the target intention according to the characteristic information. And the multi-dimensional user intention identification characteristic information can be updated according to a preset period, so that the updating timeliness is ensured, and the current search requirement and search intention of the user can be better fitted.
It should be noted that the above description is exemplified by the purpose of including the search merchant, but the embodiments of the disclosure are not limited thereto, and for example, a search for media files, a search for literature, and the like are also applicable to the disclosure. In addition, the preset numerical values, such as preset digits, preset characters, preset percentages, and the like, can be pre-calibrated, and the value of the numerical value is not specifically limited in the present disclosure.
In S202, it is determined whether the user search intention recognition model determines the target related information. In case of yes, S203 is executed; in the case of no, S204 is executed.
If the search amount of the user for the retrieval information is large, the relevant user intention identification characteristic information can be counted in advance according to the historical search behavior of the user for the retrieval information, and the user search intention identification model can accurately determine the target relevant information between the retrieval information and the target intention in the aspect of the user search behavior according to the characteristic information. If the search amount of the user for the retrieval information is small, or the retrieval information is not searched, the related user intention identification characteristic information is small, and the user search intention identification model may not be capable of accurately determining the target related information.
In S203, the target related information is determined as the first feature information.
In the case where the user search intention recognition model determines the target related information, the user search intention recognition model may determine the target related information as the first feature information.
In S204, the first feature information is determined as a preset value.
In the case where the target related information is not determined by the user search intention recognition model, the first feature information may be determined as a preset value, and the preset value may be determined in advance according to a test.
Through the technical scheme, the characteristic information for identifying the user intention can be multi-dimensional information, and the characteristic information is more comprehensive. The user search intention identification model can accurately determine target related information between the retrieval information and the target intention in the aspect of user search behaviors according to the multi-dimensional user intention identification characteristic information, and the target related information is used as first characteristic information of the retrieval information. Under the condition that the target related information is not determined by the user search intention recognition model, the first characteristic information can be determined as a preset value, and the first characteristic information can be effectively determined.
In this disclosure, the determining the second characteristic information of the search information in S102 may include: the second characteristic information is determined in one of two embodiments as follows.
In one embodiment, the search information is input into the semantic intention recognition model, and second feature information output by the semantic intention recognition model is obtained.
The semantic intent recognition model may be pre-trained, and the semantic intent recognition model may adopt any model, such as a BERT (Bidirectional Encoder retrieval from transforms) model. After acquiring the retrieval information input by the target user, the server can input the retrieval information into the semantic intention recognition model, and the semantic intention recognition model can output second characteristic information of the retrieval information through calculation.
In another embodiment, in the case that second feature information of the retrieval information exists in the model cache result, the second feature information is obtained from the model cache result, wherein the model cache result is determined according to the historical output result of the semantic intent recognition model.
In this embodiment, the historical output result of the semantic intention recognition model may be cached, for example, the second feature information of the search information is calculated before the semantic intention recognition model, and after the server acquires the search information input by the target user this time, the server may directly obtain the second feature information from the model cache result, so that repeated calculation of the semantic intention recognition model is not required, the model calling efficiency is improved, and the intention recognition efficiency is improved.
FIG. 3 is a flow chart illustrating a method of intent recognition, which may include S301-S304, as shown in FIG. 3, according to another exemplary embodiment. Wherein S103 may include S303 and S304.
In S301 (101), search information input by the target user is acquired.
In S302 (102), first characteristic information of the search information is determined, and second characteristic information of the search information is determined.
The correlation operation in S301 is consistent with the correlation operation mentioned in S101, and the correlation operation in S302 is consistent with the correlation operation mentioned in S102, which are not described herein again.
In S303, the target feature information is determined according to the first feature information, the first preset weight corresponding to the first feature information, the second feature information, and the second preset weight corresponding to the second feature information.
The first preset weight corresponding to the first characteristic information and the second preset weight corresponding to the second characteristic information can be determined in advance. For example, the target feature information may be determined by the following formula (1), and the target feature information may characterize the probability or likelihood that the retrieval information is the target intention.
Y=α*X1+β*X2 （1）
Wherein Y represents object feature information, X1The first characteristic information is represented by a first characteristic information,αrepresenting a first predetermined weight, X2The first characteristic information is represented by a first characteristic information,βrepresenting a second preset weight.
In S304, in the case that the target feature information is greater than the preset threshold, it is determined that the retrieval information represents the target intention.
The target characteristic information is larger than a preset threshold value, which can indicate that the possibility that the purpose of inputting the retrieval information by the user is the target intention is high, and the target intention represented by the retrieval information can be determined. Wherein the predetermined threshold value can be calibrated in advance. Illustratively, taking the target intention as the search merchant as an example, for example, the retrieval information includes a word a, and if it is determined that the target feature information corresponding to the word a is greater than the preset threshold, it may be determined that the intention of the target user is to search for merchants related to the word a.
In addition, the intention identifying method provided by the present disclosure may further include: and determining display information to be displayed to the target user according to the intention identification result of the retrieval information.
Wherein the intent recognition result may include a target intent characterized by the search information or a non-target intent characterized by the search information. For example, taking the target intention including the search merchant as an example, for example, it is identified that the retrieval information represents that the target user is to search for the merchant, then information of the merchant related to the retrieval information may be determined as the presentation information to be presented to the target user, for example, information of the merchant including the retrieval information in the name of the merchant, and information of the merchant including the retrieval information in the name of the product may be determined as the presentation information. If it is identified that the retrieved information is not representative of a searching merchant, i.e., the user is not intended to search for a merchant by retrieving information, then information for the merchant that is relevant to the retrieved information may not be determined to be the presentation information.
Therefore, the intention identification result can directly influence the search result, the display information to be displayed to the target user is determined according to the intention identification result of the retrieval information, the accuracy of the search result can be ensured, and the determined display information is information meeting the requirements of the user.
By the scheme, the intention recognition is carried out by combining two aspects of user searching behaviors and semantics, the accuracy and the recognition success rate of the intention recognition can be effectively improved, the provided searching result is the content meeting the real searching intention of the user, the accuracy of the searching result provided for the user is ensured, and the user experience is improved.
Based on the same inventive concept, the present disclosure also provides an intention recognition apparatus, and fig. 4 is a block diagram illustrating an intention recognition apparatus according to an exemplary embodiment, and as shown in fig. 4, the apparatus 400 may include:
a retrieval information obtaining module 401 configured to obtain retrieval information input by a target user;
a feature information determination module 402 configured to determine first feature information of the retrieval information, wherein the first feature information is used for characterizing a degree of correlation between the retrieval information and a target intention in terms of user search behavior, and determine second feature information of the retrieval information, wherein the second feature information is used for characterizing a degree of correlation between the retrieval information and the target intention in terms of semantics;
a target intention identifying module 403 configured to identify whether the retrieval information represents the target intention according to the first feature information and the second feature information.
Through the technical scheme, after the retrieval information input by the target user is acquired, the first characteristic information of the retrieval information can be determined, and the second characteristic information of the retrieval information can be determined, wherein the first characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of user searching behaviors, and the second characteristic information can be used for representing the degree of correlation between the retrieval information and the target intention in the aspect of semantics. Meanwhile, whether the retrieval information representation is a target intention is identified according to the first characteristic information and the second characteristic information, and intention identification can be carried out from two aspects of user search behavior and semantics. The method has the advantages that the influence of the semantics of the retrieval information is small in the aspect of user searching behaviors, the searching intention of the target user can be accurately identified under the condition that the searching amount of the retrieval information is large, and the searching intention of the target user can be accurately identified based on the semantics of the retrieval information under the condition that the searching amount of the retrieval information is small or the retrieval information is not searched. Therefore, the limitation problem caused by identifying the search intention of the user only from the aspect of user search behavior or only from the aspect of semantics is avoided, the intention identification is carried out by combining the two aspects of user search behavior and semantics, the accuracy and the identification success rate of the intention identification can be effectively improved, the provided search result is the content meeting the real search intention of the user, the accuracy of the search result provided for the user is ensured, and the user experience is improved.
Optionally, the feature information determining module 402 may include: a first input sub-module configured to input the retrieval information and the location information of the target user into a user search intention recognition model to determine, by the user search intention recognition model, target related information between the retrieval information and the target intention in terms of user search behavior according to the retrieval information, the location information, and user intention recognition feature information of the retrieval information in a target area indicated by the location information, wherein the user intention recognition feature information is determined according to historical search behavior of a user for the retrieval information, and the user intention recognition feature information is updated according to a preset period; a first determination sub-module configured to determine the target-related information as the first feature information if the target-related information is determined by the user search intention recognition model.
Optionally, the apparatus 400 may further include: a determination module configured to determine the first feature information as a preset value if the target related information is not determined by the user search intention recognition model.
Optionally, the feature information determining module 402 determines the second feature information by using one of the following second input submodule and obtaining submodule: a second input submodule configured to input the retrieval information into a semantic intention recognition model, resulting in the second feature information output by the semantic intention recognition model; an obtaining sub-module configured to obtain the second feature information from a model cache result if the second feature information of the retrieval information exists in the model cache result, the model cache result being determined according to a historical output result of the semantic intent recognition model.
Optionally, the target intention identifying module 403 may include: a second determining submodule configured to determine target feature information according to the first feature information, a first preset weight corresponding to the first feature information, the second feature information, and a second preset weight corresponding to the second feature information; a third determining sub-module configured to determine that the retrieval information is characterized by the target intention if the target characteristic information is greater than a preset threshold.
Optionally, the apparatus 400 may further include: and the display information determining module is configured to determine display information to be displayed to the target user according to the intention recognition result of the retrieval information.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Fig. 5 is a block diagram illustrating an electronic device 500 in accordance with an example embodiment. For example, the electronic device 500 may be provided as a server. Referring to fig. 5, the electronic device 500 comprises a processor 522, which may be one or more in number, and a memory 532 for storing computer programs executable by the processor 522. The computer programs stored in memory 532 may include one or more modules that each correspond to a set of instructions. Further, the processor 522 may be configured to execute the computer program to perform the above-described intention identifying method.
Additionally, the electronic device 500 may also include a power component 526 and a communication component 550, the power component 526 may be configured to perform power management of the electronic device 500, and the communication component 550 may be configured to enable communication, e.g., wired or wireless communication, of the electronic device 500. In addition, the electronic device 500 may also include input/output (I/O) interfaces 558. The electronic device 500 may operate based on an operating system, such as Windows Server, stored in the memory 532TM，Mac OS XTM，UnixTM，LinuxTMAnd so on.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the above-described intent recognition method is also provided. For example, the computer readable storage medium may be the memory 532 described above including program instructions that are executable by the processor 522 of the electronic device 500 to perform the intent recognition method described above.
In another exemplary embodiment, a computer program product is also provided, which comprises a computer program executable by a programmable apparatus, the computer program having code portions for performing the above-mentioned intention identification method when executed by the programmable apparatus.
The preferred embodiments of the present disclosure are described in detail with reference to the accompanying drawings, however, the present disclosure is not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solution of the present disclosure within the technical idea of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
It should be noted that, in the foregoing embodiments, various features described in the above embodiments may be combined in any suitable manner, and in order to avoid unnecessary repetition, various combinations that are possible in the present disclosure are not described again.
In addition, any combination of various embodiments of the present disclosure may be made, and the same should be considered as the disclosure of the present disclosure, as long as it does not depart from the spirit of the present disclosure.
Claims (8)
1. An intent recognition method, the method comprising:
acquiring retrieval information input by a target user;
determining first characteristic information of the retrieval information, and determining second characteristic information of the retrieval information, wherein the first characteristic information is used for characterizing the degree of correlation between the retrieval information and a target intention in terms of user search behaviors, the second characteristic information is used for characterizing the degree of correlation between the retrieval information and the target intention in terms of semantics, and the degree of correlation between the retrieval information and the target intention is used for characterizing the probability that the target user inputs the retrieval information for the purpose of the target intention;
identifying whether the retrieval information represents the target intention or not according to the first characteristic information and the second characteristic information;
the determining the first characteristic information of the retrieval information comprises:
inputting the retrieval information and the position information of the target user into a user search intention recognition model, so that the user search intention recognition model determines target related information between the retrieval information and the target intention in terms of user search behaviors according to the retrieval information, the position information and user intention recognition characteristic information of the retrieval information in a target area indicated by the position information, wherein the user intention recognition characteristic information is determined according to historical search behaviors of the user on the retrieval information, and the user intention recognition characteristic information is updated according to a preset period;
determining the target related information as the first characteristic information under the condition that the target related information is determined by the user search intention recognition model;
the target intent comprises a search merchant; the user intent identification feature information comprises at least one of:
the method comprises the following steps of obtaining information related to operation of a user in a target area for a target merchant, obtaining information related to search conditions of users in a plurality of areas for the search information, obtaining correlation information between the search information and merchant feature information of the target merchant, obtaining information related to operation of the user in the target area for a preset merchant, and obtaining information related to entity link of the search information, wherein the target merchant comprises each merchant in the target area determined according to the search information, and the preset merchant comprises chained merchants in the target area.
2. The method of claim 1, further comprising:
and under the condition that the target related information is not determined by the user search intention identification model, determining the first characteristic information as a preset value.
3. The method of claim 1, wherein the determining second characteristic information of the retrieved information comprises:
determining the second characteristic information in one of the following ways:
inputting the retrieval information into a semantic intention recognition model to obtain the second characteristic information output by the semantic intention recognition model;
and obtaining the second characteristic information of the retrieval information from a model cache result under the condition that the second characteristic information exists in the model cache result, wherein the model cache result is determined according to the historical output result of the semantic intention recognition model.
4. The method of claim 1, wherein identifying whether the retrieved information is indicative of the target intent based on the first characteristic information and the second characteristic information comprises:
determining target characteristic information according to the first characteristic information, a first preset weight corresponding to the first characteristic information, the second characteristic information and a second preset weight corresponding to the second characteristic information;
and determining that the retrieval information is characterized by the target intention under the condition that the target characteristic information is larger than a preset threshold value.
5. The method of claim 1, further comprising:
and determining display information to be displayed to the target user according to the intention identification result of the retrieval information.
6. An intent recognition apparatus, characterized in that the apparatus comprises:
the retrieval information acquisition module is configured to acquire retrieval information input by a target user;
a feature information determination module configured to determine first feature information of the retrieval information, and determine second feature information of the retrieval information, wherein the first feature information is used for characterizing a degree of correlation between the retrieval information and a target intention in terms of user search behavior, the second feature information is used for characterizing a degree of correlation between the retrieval information and the target intention in terms of semantics, and the degree of correlation between the retrieval information and the target intention is used for characterizing a probability that a purpose of the retrieval information input by the target user is the target intention;
a target intention identifying module configured to identify whether the retrieval information represents the target intention according to the first characteristic information and the second characteristic information;
the characteristic information determination module includes: a first input sub-module configured to input the retrieval information and the location information of the target user into a user search intention recognition model to determine, by the user search intention recognition model, target related information between the retrieval information and the target intention in terms of user search behavior according to the retrieval information, the location information, and user intention recognition feature information of the retrieval information in a target area indicated by the location information, wherein the user intention recognition feature information is determined according to historical search behavior of a user for the retrieval information, and the user intention recognition feature information is updated according to a preset period;
a first determination sub-module configured to determine the target-related information as the first feature information if the target-related information is determined by the user search intention recognition model;
the target intent comprises a search merchant; the user intent identification feature information comprises at least one of: the method comprises the following steps of obtaining information related to operation of a user in a target area for a target merchant, obtaining information related to search conditions of users in a plurality of areas for the search information, obtaining correlation information between the search information and merchant feature information of the target merchant, obtaining information related to operation of the user in the target area for a preset merchant, and obtaining information related to entity link of the search information, wherein the target merchant comprises each merchant in the target area determined according to the search information, and the preset merchant comprises chained merchants in the target area.
7. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 5.
8. An electronic device, comprising:
a memory having a computer program stored thereon;
a processor for executing the computer program in the memory to carry out the steps of the method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010996810.2A CN111930954B (en) | 2020-09-21 | 2020-09-21 | Intention recognition method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010996810.2A CN111930954B (en) | 2020-09-21 | 2020-09-21 | Intention recognition method and device, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111930954A CN111930954A (en) | 2020-11-13 |
| CN111930954B true CN111930954B (en) | 2021-01-29 |
Family
ID=73335308
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010996810.2A Active CN111930954B (en) | 2020-09-21 | 2020-09-21 | Intention recognition method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111930954B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105404680A (en) * | 2015-11-25 | 2016-03-16 | 百度在线网络技术（北京）有限公司 | Searching recommendation method and apparatus |
| CN111274787A (en) * | 2020-02-21 | 2020-06-12 | 支付宝(杭州)信息技术有限公司 | User intention prediction method and system |
| CN111353836A (en) * | 2018-12-20 | 2020-06-30 | 百度在线网络技术（北京）有限公司 | Commodity recommendation method, device and equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160078012A1 (en) * | 2014-09-11 | 2016-03-17 | Bmc Software, Inc. | Systems and methods for formless information technology and social support mechanics |
-
2020
- 2020-09-21 CN CN202010996810.2A patent/CN111930954B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105404680A (en) * | 2015-11-25 | 2016-03-16 | 百度在线网络技术（北京）有限公司 | Searching recommendation method and apparatus |
| CN111353836A (en) * | 2018-12-20 | 2020-06-30 | 百度在线网络技术（北京）有限公司 | Commodity recommendation method, device and equipment |
| CN111274787A (en) * | 2020-02-21 | 2020-06-12 | 支付宝(杭州)信息技术有限公司 | User intention prediction method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111930954A (en) | 2020-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110069610B (en) | Solr-based retrieval method, solr-based retrieval device, solr-based retrieval equipment and storage medium | |
| US7756868B2 (en) | Method for providing search results list based on importance information and system thereof | |
| US8983971B2 (en) | Method, apparatus, and system for mobile search | |
| KR101644817B1 (en) | Generating search results | |
| CN111797210A (en) | Information recommendation method, device and equipment based on user portrait and storage medium | |
| CN109918594B (en) | Information display method and device | |
| CN109299235B (en) | Knowledge base searching method, device and computer readable storage medium | |
| CN110990533B (en) | Method and device for determining standard text corresponding to query text | |
| CN107885717B (en) | Keyword extraction method and device | |
| CN114330329A (en) | Service content searching method and device, electronic equipment and storage medium | |
| CN110245357B (en) | Main entity identification method and device | |
| CN110008396B (en) | Object information pushing method, device, equipment and computer readable storage medium | |
| CN112487159B (en) | Search method, search device, and computer-readable storage medium | |
| CN112506864B (en) | File retrieval method, device, electronic equipment and readable storage medium | |
| CN111930954B (en) | Intention recognition method and device, storage medium and electronic equipment | |
| CN116881432A (en) | Text pushing method, text pushing device, electronic equipment and storage medium | |
| CN114579883A (en) | Address query method, method for obtaining address vector representation model and corresponding device | |
| CN116610853A (en) | Search recommendation method, search recommendation system, computer device, and storage medium | |
| CN114282119A (en) | Scientific and technological information resource retrieval method and system based on heterogeneous information network | |
| CN105159899B (en) | Searching method and device | |
| CN112749258A (en) | Data searching method and device, electronic equipment and storage medium | |
| CN116610782B (en) | Text retrieval method, device, electronic equipment and medium | |
| CN117648444B (en) | Patent clustering method and system based on graph convolution attribute aggregation | |
| CN112307312A (en) | Article recommendation method and device | |
| CN117573959B (en) | General method for obtaining news text based on web page xpath |
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 |