CN110390054B - Interest point recall method, device, server and storage medium - Google Patents
Interest point recall method, device, server and storage medium Download PDFInfo
- Publication number
- CN110390054B CN110390054B CN201910677499.2A CN201910677499A CN110390054B CN 110390054 B CN110390054 B CN 110390054B CN 201910677499 A CN201910677499 A CN 201910677499A CN 110390054 B CN110390054 B CN 110390054B
- Authority
- CN
- China
- Prior art keywords
- candidate
- search request
- recall
- target
- target 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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a method, a device, a server and a storage medium for recalling interest points. The method comprises the following steps: determining candidate interest point information of a target search request; extracting candidate recall domain information from the candidate interest point information; determining the correlation degree between the target search request and the candidate interest points according to the target search request and the candidate recall domain information; and determining a target interest point from the candidate interest points according to the correlation degree between the target search request and the candidate interest points. According to the embodiment of the invention, through fine-grained matching of the information of each recall domain and the influence of the search request on the importance degree of the recall domain, the determination of the relevance is directly embodied in the understanding of the search request, more reliable relevance characteristics are provided for the recall of the interest points, the recall accuracy of the interest points is improved, and the search requirements of users are better met.
Description
Technical Field
The embodiment of the invention relates to the technical field of retrieval, in particular to a method, a device, a server and a storage medium for recalling interest points.
Background
With the rapid development of retrieval technology and the internet, a user can retrieve and obtain Point of Interest (POI) information to be queried by inputting a search request.
Currently, a search for a point of interest is usually performed based on a keyword, and the point of interest information including all words in a search request is used as a final search result, that is, a recall result of the point of interest is determined according to a text similarity between the search request and a retrieved point of interest. Or the searched word list and the regular expression are adopted, the retrieved interest point information is evaluated based on factors such as geographic position, heat degree and the like, and the effective interest points evaluated based on the factors are recalled.
However, in actual situations, the search demand of the user varies, and for example, when searching a building, the user may search for a perfect match based on the name, may search for an address, may search for a route, or the like. Therefore, in the prior art, the search requirements of the user for the points of interest are completely ignored, the accuracy of understanding the search request and positioning the requirements of the user is insufficient, the understanding information of the search request is not fully utilized, the search requirement identification is not flexible enough, and the accuracy of recalling the point of interest information is low.
Disclosure of Invention
The embodiment of the invention provides a point of interest recall method, a point of interest recall device, a server and a storage medium, which can improve the accuracy of point of interest recall.
In a first aspect, an embodiment of the present invention provides a method for recalling a point of interest, including:
determining candidate interest point information of a target search request;
extracting candidate recall domain information from the candidate interest point information;
determining the correlation degree between the target search request and the candidate interest points according to the target search request and the candidate recall domain information;
and determining a target interest point from the candidate interest points according to the correlation degree between the target search request and the candidate interest points.
In a second aspect, an embodiment of the present invention provides a point of interest recall apparatus, including:
the interest point searching module is used for determining candidate interest point information of the target searching request;
a recall domain information extraction module for extracting candidate recall domain information from the candidate interest point information;
the relevancy determining module is used for determining the relevancy between the target search request and the candidate interest points according to the target search request and the candidate recall domain information;
and the target interest point determining module is used for determining a target interest point from the candidate interest points according to the correlation between the target search request and the candidate interest points.
In a third aspect, an embodiment of the present invention provides a server, including:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a point of interest recall method according to any embodiment of the present invention.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a point of interest recall method according to any embodiment of the present invention.
After global search is carried out according to a target search request to obtain candidate interest point information, the interest point information is split to carry out fine-grained matching on each piece of split recall domain information and the search request, and according to the hit condition of each piece of recall domain information and the influence of the search request on the importance degree of the recall domain, the correlation degree between the search request and the interest point is comprehensively obtained, so that the target interest point is determined from the candidate interest points according to the correlation degree. According to the embodiment of the invention, through fine-grained matching of the information of each recall domain and the influence of the search request on the importance degree of the recall domain, the determination of the relevance is directly embodied in the understanding of the search request, more reliable relevance characteristics are provided for the recall of the interest points, the recall accuracy of the interest points is improved, and the search requirements of users are better met.
Drawings
Fig. 1 is a flowchart of a point of interest recall method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a point of interest recall method according to a second embodiment of the present invention;
fig. 3 is an exemplary diagram of a point of interest recall process according to a second embodiment of the present invention;
fig. 4 is a schematic structural diagram of a point of interest recall apparatus according to a third embodiment of the present invention;
fig. 5 is a schematic structural diagram of a server according to a fourth embodiment of the present invention.
Detailed Description
The embodiments of the present invention will be described in further detail with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the invention and do not delimit the invention. It should be further noted that, for convenience of description, only some structures related to the embodiments of the present invention are shown in the drawings, not all of them.
It should be further noted that, for the convenience of description, only some but not all of the matters relating to the present application are shown in the drawings. Before discussing exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart may describe the operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently or simultaneously. In addition, the order of the operations may be re-arranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figure. The processes may correspond to methods, functions, procedures, subroutines, and the like.
Example one
Fig. 1 is a flowchart of a point of interest recall method according to an embodiment of the present invention, where the embodiment is applicable to a case where a target point of interest satisfying a user requirement is recalled from candidate points of interest obtained through searching, and the method may be implemented by a point of interest recall apparatus, which may be implemented in software and/or hardware, and is preferably configured on a server carrying a point of interest searching function. The method specifically comprises the following steps:
and S110, determining candidate interest point information of the target search request.
In the specific embodiment of the present invention, the target search request refers to a search request currently initiated by a user, and is used to perform a map interest point search from a map to obtain at least one candidate interest point, where each candidate interest point has corresponding candidate interest point information. The search request may include, but is not limited to, name information, address information, category information, attribute information, brand information, and the like of the point of interest to be searched, and is used to limit the point of interest to be searched in different dimensions. Therefore, the restriction information of different dimensions contained in the target search request initiated by the user currently can reflect the search requirement of the user. For example, if the target search request is "a building", the search requirement of the user may be an accurate point search; if the target search request is "a certain way N number", the search requirement of the user may be an address search requirement. Correspondingly, the candidate interest point information may include, but is not limited to, name information, address information, category information, attribute information, brand information, and the like of the candidate interest point, and is used to describe the candidate interest point from different dimensions.
After receiving a target search request initiated by a user, the present embodiment may perform a global search based on the target search request to obtain at least one candidate interest point. In this embodiment, the search mode of the candidate interest point is not limited, and any mode capable of realizing the interest point search may be applied to this embodiment. For example, matching is performed based on text, and the interest points containing all information or key information in the target search request are determined as candidate interest points. The content matched between the target search request and the candidate interest point information may be respectively located in text regions where the target search request and the candidate interest point information are correspondingly consistent, for example, both located in the name information; the matched contents may also be located in different text regions between the target search request and the candidate point of interest information, respectively, such as in the target search request name information and in the address information of the candidate point of interest information, respectively.
And S120, extracting candidate recall domain information from the candidate interest point information.
In the embodiment of the present invention, the recall domain refers to a text region in the text description of the candidate interest point, where the candidate interest point is described from different dimensions. Namely, the content categories are split for the text description of the candidate interest point information, and the text description area corresponding to one category is called a recall domain. Accordingly, the recall domain includes, but is not limited to, at least one of a name domain, an address domain, a topological relation domain, an attribute tag domain, a category tag domain, and a brand tag domain. The name field is used for describing name information of the interest points; the address field is used for describing the address information of the interest point; the topological relation domain is used for describing a road connection relation between at least two interest points; the attribute tag field is used for identifying the attribute of the interest point, such as business hours and other information; the category label field is used for identifying categories of the interest points, such as restaurants, clothes or banks and the like; the brand label field is used to identify brand information for the point of interest.
In this embodiment, since the restriction information of different dimensions included in the target search request may reflect the search requirement of the user, the recall domain included in the candidate interest point information may be used as the candidate recall domain, or all the recall domains may be used as the candidate recall domain, so as to extract the candidate recall domain information located in the candidate recall domain from the candidate interest point information, that is, split the candidate interest point information into fine-grained information of different dimensions, thereby facilitating the matching with the user requirement and the determination of the importance of each piece of granular information.
For example, assuming that the candidate interest point is kendeki, and the candidate interest point information is "kendeki (new hua avenue store) -fast food-tongzhou district new hua avenue 56 number 1 floor-everyone 40 yuan", the name field information may be extracted from the candidate interest point information as kendeki, the address field information as tongzhou district new hua avenue 56 number 1 floor, the category label field information as fast food, the brand label field information as new hua avenue store, and the like.
S130, determining the correlation degree between the target search request and the candidate interest points according to the target search request and the candidate recall domain information.
In an embodiment of the present invention, for a single candidate point of interest, a single domain correlation between the target search request and each candidate recall domain information in the candidate point of interest information may be determined. And comprehensively determining the correlation degree between the target search request and the candidate interest points based on the target search request and the single domain correlation degree between the target search request and each candidate recall domain information.
In this embodiment, a deep learning model of an Attention mechanism may be introduced, and a historical search request and interest point information matched with the historical search request are learned in advance to obtain a relevance calculation model. And inputting the target search request and the candidate interest point information into a relevance calculation model, and identifying the importance degree of each candidate recall domain under the condition of the current search request by the reference of an Attention mechanism. And finally, comprehensively obtaining the correlation between the target search request and the candidate interest points based on the single domain correlation between the target search request and each candidate recall domain information and the importance degree of each candidate recall domain.
In this embodiment, on the basis of the identification of the importance degree of the candidate recall domain, according to the importance degree of the candidate recall domain, if the importance degree of a certain candidate recall domain is higher and the candidate recall domain information is matched with the target search request, the relevance degree scoring processing may be performed; on the contrary, if the importance degree of a certain candidate recall domain is lower, even if the candidate recall domain information is matched with the target search request, the relevance degree reduction processing can be carried out; if the importance degree of a certain candidate recall domain is low and the candidate recall domain information is not matched with the target search request, the relevance can be greatly reduced.
In this embodiment, a deep learning-based classification model may also be introduced, a user click behavior of a search result of a historical search request is predetermined, click behavior distribution of the historical search request is determined according to the user click behavior, and a search type of the historical search request is learned; and/or determining key components capable of reflecting search requirements in the historical search requests in advance, and learning the search types of the historical search requests according to the key components of the search requirements to obtain a deep learning-based classification model. And classifying the target search request based on the classification model obtained by training so as to determine the target search type to which the target search request belongs. Wherein the search type includes but is not limited to at least one of a precision point search type, an address search type, a route search type, an attribute search type, a category search type and a brand search type, and the target search type can further reflect the search requirement of the user. And then the target search type is used as one of the calculation bases of the relevance calculation model introduced into the Attention mechanism, and the guide model identifies the importance degree of each candidate recall domain by combining the specific content in the target search request under the condition of the target search type. And finally, comprehensively obtaining the correlation between the target search request and the candidate interest points based on the single domain correlation between the target search request and each candidate recall domain information and the importance degree of each candidate recall domain.
Illustratively, a target search type to which the target search request belongs is determined, the weight of the candidate recall domain is determined according to the target search request and the target search type, and the correlation degree between the target search request and the candidate interest point is determined according to the target search request, the candidate recall domain information and the weight of the candidate recall domain. Before determining the target search type to which the target search request belongs, determining click behavior distribution of the historical search request according to user click behavior of search results of the historical search request, and determining the search type of the historical search request according to the click behavior distribution of the historical search request. And determining the search type of the historical search request matched with the target search request as the target search type of the target search request according to the similarity between the target search request and the historical search request. Or performing semantic segmentation on the target search request, determining key text components in the target search request, and determining the search type to which the key text components belong as the target search type of the target search request. And then determining the correlation between the target search request and the candidate interest points according to the target search request, the candidate recall domain information and the weights of the candidate recall domains, which may include determining at least one candidate recall domain information of the candidate interest points, the matching degree between the candidate recall domain information and the target search request, and performing an operation, such as weighted summation, according to the weights of the candidate recall domains and the matching degree between the candidate recall domain information and the target search request to determine the correlation between the target search request and the candidate interest points.
In the embodiment, the candidate interest point information is split into the recall domain information, the correlation degree of the target search request and each candidate recall domain information is determined, and the fine-grained matching of the candidate interest points is realized in the interest point recall process. And the method and the device take a recall domain capable of reflecting the search requirements of the user as a main part, comprehensively determine the correlation degree between the candidate interest points and the target search request, directly reflect the determination of the correlation on the understanding of the search request and provide more reliable correlation characteristics for the interest point recall.
And S140, determining the target interest points from the candidate interest points according to the correlation between the target search request and the candidate interest points.
In the specific embodiment of the invention, according to the correlation between the target search request and the candidate interest points, the candidate interest points with the highest correlation can be selected as the target interest points, and the candidate interest points can be ranked according to the correlation and a plurality of interest points with higher correlation can be selected as the target interest points, so that the interest points can be effectively recalled.
According to the technical scheme of the embodiment, after global search is carried out according to a target search request to obtain candidate interest point information, the interest point information is split to carry out fine-grained matching on each piece of split recall domain information and the search request, and according to the hit condition of each piece of recall domain information and the influence of the search request on the importance degree of the recall domain, the correlation degree between the search request and the interest point is comprehensively obtained, so that the target interest point is determined from the candidate interest points according to the correlation degree. According to the embodiment of the invention, through fine-grained matching of the information of each recall domain and the influence of the search request on the importance degree of the recall domain, the determination of the relevance is directly embodied in the understanding of the search request, more reliable relevance characteristics are provided for the recall of the interest points, the recall accuracy of the interest points is improved, and the search requirements of users are better met.
Example two
On the basis of the first embodiment, the present embodiment provides a preferred implementation of the method for recalling points of interest, which can comprehensively determine the correlation between the target search request and the candidate points of interest according to the target search type to which the target search request belongs. Fig. 2 is a flowchart of a point of interest recall method according to a second embodiment of the present invention, as shown in fig. 2, the method specifically includes the following steps:
s210, candidate interest point information of the target search request is determined.
And S220, extracting candidate recall domain information from the candidate interest point information.
And S230, determining the target search type to which the target search request belongs.
In an embodiment of the present invention, the search request is classified into types based on the search requirement of the user, and the search type includes, but is not limited to, at least one of a precise point search type, an address search type, a route search type, an attribute search type, a category search type, and a brand search type. The precise point search type, the address search type and the route search type can belong to precise search, and the attribute search type, the category search type and the brand search type can belong to general search. And carrying out space semantic analysis on the target search request, and determining a target search type reflecting the user search requirement of the target search request.
In this embodiment, a deep learning based classification model may be introduced. Optionally, before determining the target search type to which the target search request belongs, the method further includes: determining click behavior distribution of historical search requests according to user click behaviors of search results of the historical search requests; and determining the search type of the historical search request according to the click behavior distribution of the historical search request. And/or performing text component division based on the semantics of the historical search request, determining a key text component capable of reflecting the search requirement of the user, and determining the search type of the key text component. Therefore, at least the click behavior distribution of the historical search request and the key text components are used as training samples, and the classification model is obtained through deep learning of the training samples.
Accordingly, the present embodiment inputs the target search request into the classification model, so as to obtain the target search type to which the target search request belongs. Optionally, determining similarity between the target search request and the historical search request according to text information in the search request; and determining the search type of the historical search request matched with the target search request as the target search type of the target search request according to the similarity between the target search request and the historical search request. In this embodiment, the calculation method of the similarity is not limited, and any calculation method capable of achieving the similarity may be applied to this embodiment, for example, the similarity calculation based on the text. Optionally, semantic segmentation can be performed on the target search request, and a key text component in the target search request is determined; and determining the search type to which the key text component belongs as the target search type of the target search request.
S240, determining the weight of the candidate recall domain according to the target search request and the target search type.
In the embodiment of the present invention, the weight of the candidate recall domain reflects the applicability of matching of each candidate recall domain in the case of a target search request and a target search type to which the target search request belongs. For example, under the exact point search type, it is usually appropriate to match in the name domain. Notably, there is no absolute association between the search type and the weight of the candidate recall domain, which is also limited by the content of the search request itself.
In this embodiment, a deep learning model of an Attention mechanism may be introduced, and a historical search request and interest point information matched with the historical search request are learned in advance to obtain a relevance calculation model. Therefore, the target search request and the candidate interest point information are input into the relevance calculation model, the importance degree of each candidate recall domain is identified through the reference of the Attention mechanism, and the weight is distributed to each candidate recall domain.
S250, determining the matching degree between at least one candidate recall domain information in the candidate interest points and the target search request.
In the embodiment of the present invention, through the determination of the matching degree between each candidate recall domain information and the target search request in the candidate interest points, that is, the single domain relevance degree described above, it can be determined whether each candidate recall domain information contains the text content of the target search request. In this embodiment, the determination method of the matching degree is not limited, and any determination method of the matching degree may be applied to this embodiment, for example, similarity matching based on text.
S260, determining the correlation degree between the target search request and the candidate interest points according to the weight of the candidate recall domain and the matching degree between the candidate recall domain information and the target search request.
In a specific embodiment of the present invention, the weights of the candidate recall regions identify the applicability of each recall region in the current search scenario. For example, precision point search is more suitable for name domain matching, route search is more suitable for topological relation domain matching, and the like. And each candidate recall domain information under the candidate interest points has a certain matching degree with the target search request. And then combining the weight of the candidate recall domain and the matching degree of the candidate recall domain information and the target search request to comprehensively determine the correlation degree between the target search request and the candidate interest point. In this embodiment, the comprehensive determination method of the correlation is not limited, and a function calculation, for example, a weighted sum, may be performed based on the weight of the candidate recall domain and the matching degree of the candidate recall domain information and the target search request to obtain the correlation between the target search request and the candidate interest point.
S270, sorting the candidate interest points according to the correlation degree between the target search request and the candidate interest points.
In the embodiment of the present invention, the higher the correlation between the target search request and the candidate interest point is, the more the candidate interest point meets the search requirement of the user. Therefore, the candidate interest points are ranked according to the relevance between the target search request and the candidate interest points, and preferably, the relevance is ranked from high to low, so that the interest points with high relevance can be conveniently extracted.
S280, determining a target interest point from the candidate interest points according to the sorting result of the candidate interest points.
In the specific embodiment of the invention, the candidate interest points with the highest relevance can be selected as the target interest points, the ranking can be carried out according to the relevance, and a plurality of interest points with high relevance are selected as the target interest points, so that the effective recall of the interest points is realized.
Fig. 3 is an exemplary diagram of a point of interest recall process. As shown in fig. 3, a global search is performed based on the target search request to obtain at least one candidate interest point. And determining the target search type to which the target search request belongs based on the classification model. And determining the information between each candidate recall domain in the candidate interest points and the target search request and the weight of the candidate recall domain based on a deep learning model added into an Attention mechanism, and finally determining the correlation between the target search request and the candidate interest points. And determining target interest points matched with the user search requirements of the target search request based on the correlation degree, and feeding the target interest points serving as the finally recalled interest points back to the user.
For another example, assume that the target search request is "Quiko science and technology building"; and assuming that the at least three candidate interest points obtained by the search are POI _1 "quasichi science and technology building (hai lake district information hayfood 9)", POI _2 "quasichi science and technology building-siemens (hai lake district information hayfood 9 siemens)", and POI _3 "great permanent new era science and technology limited company (shang di information haba No. 9 No. 3 building quasichi science and technology building 2 layer)", respectively. The target search request is first classified based on the classification model, and the target search type of the target search request may be determined to be an exact point search. Secondly, inputting a target search request 'Quiku science and technology building' and a target search type 'accurate point search' into a deep learning model of an Attention mechanism, and calculating the weight of each recall domain. Since all the three pieces of interest point information at least include a name field and an address field, it is assumed that the weight of the calculated name field is 0.8 and the weight of the calculated address field is 0.3. Then, taking POI _1 as an example, name domain information of POI _1 is extracted as "quasicity science and technology building" and address domain information is "hai-lake region information road deck No. 9". And matching the target search request 'Quiki science and technology building' with name domain information 'Quiki science and technology building' and address domain information 'Haihe regional information Lujia 9' of the POI _1 respectively. The target search request is completely matched with the POI _1 name field, and the matching degree between the target search request and the POI _1 name field is assumed to be 1, and the matching degree between the target search request and the POI _1 address field is assumed to be 0.2. Finally, the single domain correlation degree between the target search request and the name domain of the POI _1 is obtained to be (0.8 × 1), the single domain correlation degree between the target search request and the address domain of the POI _1 is obtained to be (0.3 × 0.2), and the correlation degree between the target search request and the POI _1 is obtained to be f (0.8 × 1+0.3 × 0.2) comprehensively. Where f may be a weighted sum function. Similarly, the target search request does not match the name field of POI _2 completely, but is correlated with it; the target search request hits the address field of the POI _3, but the target search type of the target search request is precise point search, which is more suitable for matching of name field information, so that the relevance of the POI _3 is the lowest. And then POI _1 can be determined as the target point of interest.
According to the technical scheme of the embodiment, at least one candidate interest point is obtained based on a target search request, each candidate recall domain information in the candidate interest point information is extracted, a target search type to which the target search request belongs is determined, the weight of a candidate recall domain is determined according to the target search request and the target search type, the matching degree between at least one candidate recall domain information in the candidate interest points and the target search request is determined, the weight of the candidate recall domain and the matching degree between the candidate recall domain information and the target search request are synthesized, the correlation degree between the target search request and the candidate interest points is determined, and therefore the target interest point is determined from the candidate interest points according to the correlation degree. The embodiment of the invention determines the importance degree of the recall domain by determining the search type of the search request and analyzing the space semantics of the search request based on the search request content and the search type, thereby determining the correlation degree between the search request and the interest points according to the fine-grained matching of the information of each recall domain, directly reflecting the determination of the correlation on the understanding of the search request, providing more reliable correlation characteristics for the interest point recall, improving the accuracy of the interest point recall and better meeting the search requirements of users.
EXAMPLE III
Fig. 4 is a schematic structural diagram of an interest point recall apparatus according to a third embodiment of the present invention, which is applicable to a case where a target interest point meeting a user requirement is recalled from candidate interest points obtained through searching. The device specifically comprises the following steps:
an interest point searching module 410, configured to determine candidate interest point information of the target search request;
a recall domain information extracting module 420, configured to extract candidate recall domain information from the candidate interest point information;
a relevance determining module 430, configured to determine a relevance between the target search request and a candidate interest point according to the target search request and the candidate recall domain information;
a target interest point determining module 440, configured to determine a target interest point from the candidate interest points according to a correlation between the target search request and the candidate interest points.
Optionally, the correlation determination module 430 is specifically configured to:
a search type determining unit 4301 configured to determine a target search type to which the target search request belongs;
a recall domain weight determination unit 4302, configured to determine weights of candidate recall domains according to the target search request and the target search type;
a correlation determination unit 4303, configured to determine a correlation between the target search request and a candidate interest point according to the target search request, the candidate recall domain information, and a weight of the candidate recall domain.
Optionally, the correlation determination module 430 further includes a behavior analysis unit 4304, specifically configured to:
before the target search type to which the target search request belongs is determined, determining click behavior distribution of historical search requests according to user click behaviors of search results of the historical search requests;
and determining the search type of the historical search request according to the click behavior distribution of the historical search request.
Optionally, the correlation determination module 430 is further specifically configured to:
determining the similarity between the target search request and the historical search request according to text information in the search request;
and determining the search type of the historical search request matched with the target search request as the target search type of the target search request according to the similarity between the target search request and the historical search request.
Optionally, the search type determining unit 4301 is specifically configured to:
performing semantic segmentation on the target search request, and determining key text components in the target search request;
and determining the search type to which the key text component belongs as the target search type of the target search request.
Optionally, the correlation determination unit 4303 is specifically configured to:
determining the matching degree between at least one candidate recall domain information in the candidate interest points and the target search request;
and determining the correlation degree between the target search request and the candidate interest points according to the weight of the candidate recall domain and the matching degree between the candidate recall domain information and the target search request.
Optionally, the target interest point determining module 440 is specifically configured to:
sorting the candidate interest points according to the correlation degree between the target search request and the candidate interest points;
and determining a target interest point from the candidate interest points according to the sequencing result of the candidate interest points.
Optionally, the recall domain includes at least one of a name domain, an address domain, a topological relation domain, an attribute tag domain, a category tag domain, and a brand tag domain.
Optionally, the search type includes at least one of a precision point search type, an address search type, a route search type, an attribute search type, a category search type, and a brand search type.
According to the technical scheme of the embodiment, through mutual cooperation of all functional modules, the functions of receiving a search request, determining the search type, searching candidate interest points, extracting candidate recall domain information, calculating the relevance of a single domain, determining the weight of a recall domain, calculating the relevance of the search request and the interest points, selecting target interest points and the like are achieved. The embodiment of the invention determines the importance degree of the recall domain by determining the search type of the search request and analyzing the space semantics of the search request based on the search request content and the search type, thereby determining the correlation degree between the search request and the interest points according to the fine-grained matching of the information of each recall domain, directly reflecting the determination of the correlation on the understanding of the search request, providing more reliable correlation characteristics for the interest point recall, improving the accuracy of the interest point recall and better meeting the search requirements of users.
Example four
Fig. 5 is a schematic structural diagram of a server according to a fourth embodiment of the present invention, and fig. 5 shows a block diagram of an exemplary server suitable for implementing the embodiments of the present invention. The server shown in fig. 5 is only an example, and should not bring any limitation to the function and the scope of use of the embodiments of the present invention.
The server 500 shown in fig. 5 is only an example, and should not bring any limitation to the function and the use range of the embodiment of the present invention.
As shown in FIG. 5, server 500 is in the form of a general purpose computing device. The components of server 500 may include, but are not limited to: one or more processors 510, a system memory 520, and a bus 530 that couples the various system components (including the system memory 520 and the processors 510).
The system memory 520 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 521 and/or cache memory 522. The server 500 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 523 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 5 and commonly referred to as a "hard disk drive"). Although not shown in FIG. 5, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 530 by one or more data media interfaces. System memory 520 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 525 having a set (at least one) of program modules 524 may be stored, for example, in system memory 520, such program modules 524 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may include an implementation of a network environment. Program modules 524 generally perform the functions and/or methodologies of embodiments described herein.
The server 500 may also communicate with one or more external devices 540 (e.g., keyboard, pointing device, display 541, etc.), one or more devices that enable a user to interact with the server 500, and/or any device (e.g., network card, modem, etc.) that enables the server 500 to communicate with one or more other computing devices. Such communication may occur via input/output (I/O) interfaces 550. Also, server 500 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet) via network adapter 560. As shown, the network adapter 560 communicates with the other modules of the server 500 over a bus 530. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the server 500, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
The processor 510 executes various functional applications and data processing, such as implementing a point of interest recall method provided by an embodiment of the present invention, by executing programs stored in the system memory 520.
EXAMPLE five
An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program (or referred to as computer-executable instructions) is stored, where the computer program is used for executing a point of interest recall method when executed by a processor, and the method includes:
determining candidate interest point information of a target search request;
extracting candidate recall domain information from the candidate interest point information;
determining the correlation degree between the target search request and the candidate interest points according to the target search request and the candidate recall domain information;
and determining a target interest point from the candidate interest points according to the correlation degree between the target search request and the candidate interest points.
Computer storage media for embodiments of the present invention may take the form of any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for embodiments of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the embodiments of the present invention have been described in more detail through the above embodiments, the embodiments of the present invention are not limited to the above embodiments, and many other equivalent embodiments may be included without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A point of interest recall method, comprising:
determining candidate interest point information of a target search request;
extracting candidate recall domain information from the candidate interest point information;
determining a target search type to which the target search request belongs;
inputting the target search request and the target search type into a deep learning model of an attention mechanism to obtain the weight of a candidate recall domain;
determining a matching degree between at least one candidate recall domain information in the candidate interest points and the target search request;
performing function calculation based on the weight of the candidate recall domain and the matching degree of the candidate recall domain information and the target search request to obtain the correlation degree between the target search request and the candidate interest points; the weights of the candidate recall domains are used for identifying the applicability degree of each recall domain in the current search scene;
and determining a target interest point from the candidate interest points according to the correlation degree between the target search request and the candidate interest points.
2. The method of claim 1, further comprising, prior to the determining a target search type to which the target search request belongs:
determining click behavior distribution of historical search requests according to user click behaviors of search results of the historical search requests;
and determining the search type of the historical search request according to the click behavior distribution of the historical search request.
3. The method of claim 2, wherein the determining the target search type to which the target search request belongs comprises:
determining the similarity between the target search request and the historical search request according to text information in the search request;
and determining the search type of the historical search request matched with the target search request as the target search type of the target search request according to the similarity between the target search request and the historical search request.
4. The method of claim 1, wherein the determining a target search type to which the target search request belongs comprises:
performing semantic segmentation on the target search request, and determining key text components in the target search request;
and determining the search type to which the key text component belongs as the target search type of the target search request.
5. The method of claim 1, wherein determining a target point of interest from the candidate points of interest based on a degree of correlation between the target search request and the candidate points of interest comprises:
sorting the candidate interest points according to the correlation degree between the target search request and the candidate interest points;
and determining a target interest point from the candidate interest points according to the sequencing result of the candidate interest points.
6. The method of claim 1, wherein the recall domain comprises at least one of a name domain, an address domain, a topological relation domain, an attribute tag domain, a category tag domain, and a brand tag domain.
7. The method of claim 1, wherein the search type comprises at least one of a precision point search type, an address search type, a route search type, an attribute search type, a category search type, and a brand search type.
8. A point of interest recall apparatus, comprising:
the interest point searching module is used for determining candidate interest point information of the target searching request;
the recall domain information extraction module is used for extracting candidate recall domain information from the candidate interest point information;
the correlation determination module is specifically configured to:
a search type determining unit, configured to determine a target search type to which the target search request belongs;
a recall domain weight determining unit, configured to input the target search request and the target search type into a deep learning model of an attention mechanism, so as to obtain a weight of a candidate recall domain;
a relevancy determining unit, configured to determine a matching degree between at least one candidate recall domain information in the candidate interest points and the target search request; performing function calculation based on the weight of the candidate recall domain and the matching degree of the candidate recall domain information and the target search request to obtain the correlation degree between the target search request and the candidate interest points; the weights of the candidate recall domains are used for identifying the applicability of each recall domain in the current search scene;
and the target interest point determining module is used for determining a target interest point from the candidate interest points according to the correlation between the target search request and the candidate interest points.
9. A server, comprising:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement a point of interest recall method as recited in any of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out a point of interest recall method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910677499.2A CN110390054B (en) | 2019-07-25 | 2019-07-25 | Interest point recall method, device, server and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910677499.2A CN110390054B (en) | 2019-07-25 | 2019-07-25 | Interest point recall method, device, server and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110390054A CN110390054A (en) | 2019-10-29 |
CN110390054B true CN110390054B (en) | 2022-11-18 |
Family
ID=68287489
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910677499.2A Active CN110390054B (en) | 2019-07-25 | 2019-07-25 | Interest point recall method, device, server and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110390054B (en) |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110851726B (en) * | 2019-11-18 | 2022-07-26 | 百度在线网络技术(北京)有限公司 | Interest point selection method and device and electronic equipment |
CN113139136B (en) * | 2020-01-17 | 2023-09-26 | 百度在线网络技术(北京)有限公司 | Address searching method and device, electronic equipment and medium |
CN113221025B (en) * | 2020-01-21 | 2024-04-02 | 百度在线网络技术(北京)有限公司 | Point-of-interest recall method, device, equipment and medium |
CN113254756B (en) * | 2020-02-12 | 2024-03-26 | 百度在线网络技术(北京)有限公司 | Advertisement recall method, device, equipment and storage medium |
CN113282690B (en) * | 2020-02-19 | 2024-04-02 | 百度在线网络技术(北京)有限公司 | Method, device, equipment and storage medium for ordering recall of interest points |
CN111753195B (en) * | 2020-06-17 | 2024-01-09 | 百度在线网络技术(北京)有限公司 | Label system construction method, device, equipment and storage medium |
CN111767478B (en) * | 2020-06-22 | 2023-08-15 | 北京百度网讯科技有限公司 | Association relation construction method, device, equipment and storage medium |
CN112115222B (en) * | 2020-09-08 | 2024-03-08 | 北京百度网讯科技有限公司 | Point of interest display method, device, electronic equipment and storage medium |
CN112559876B (en) * | 2020-12-22 | 2024-05-28 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for showing map search results |
CN113239295A (en) * | 2021-05-18 | 2021-08-10 | 北京百度网讯科技有限公司 | Search method, search device, electronic equipment and storage medium |
CN114168756B (en) * | 2022-01-29 | 2022-05-13 | 浙江口碑网络技术有限公司 | Query understanding method and device for search intention, storage medium and electronic device |
CN116304384B (en) * | 2023-03-02 | 2023-10-27 | 车巴达(苏州)网络科技有限公司 | Point-of-interest searching method, device, computer equipment and storage medium |
CN117725324B (en) * | 2024-02-08 | 2024-05-24 | 腾讯科技(深圳)有限公司 | Map searching method and device, electronic equipment, storage medium and program product |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102867031A (en) * | 2012-08-27 | 2013-01-09 | 百度在线网络技术(北京)有限公司 | Method and system for optimizing point of interest (POI) searching results, mobile terminal and server |
CN104077324A (en) * | 2013-03-29 | 2014-10-01 | 百度在线网络技术(北京)有限公司 | Map searching method and system |
KR20150057472A (en) * | 2013-11-19 | 2015-05-28 | 한국과학기술연구원 | Method for solving ambiguity for extraction of a POI, Method for extracting a POI from a document and Apparatus for extracting a POI |
CN109508361A (en) * | 2018-11-12 | 2019-03-22 | 百度在线网络技术(北京)有限公司 | Method and apparatus for output information |
-
2019
- 2019-07-25 CN CN201910677499.2A patent/CN110390054B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102867031A (en) * | 2012-08-27 | 2013-01-09 | 百度在线网络技术(北京)有限公司 | Method and system for optimizing point of interest (POI) searching results, mobile terminal and server |
CN104077324A (en) * | 2013-03-29 | 2014-10-01 | 百度在线网络技术(北京)有限公司 | Map searching method and system |
KR20150057472A (en) * | 2013-11-19 | 2015-05-28 | 한국과학기술연구원 | Method for solving ambiguity for extraction of a POI, Method for extracting a POI from a document and Apparatus for extracting a POI |
CN109508361A (en) * | 2018-11-12 | 2019-03-22 | 百度在线网络技术(北京)有限公司 | Method and apparatus for output information |
Also Published As
Publication number | Publication date |
---|---|
CN110390054A (en) | 2019-10-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110390054B (en) | Interest point recall method, device, server and storage medium | |
CN110008300B (en) | Method and device for determining alias of POI (Point of interest), computer equipment and storage medium | |
CN109783490B (en) | Data fusion method and device, computer equipment and storage medium | |
CN109033385B (en) | Picture retrieval method, device, server and storage medium | |
CN110968654B (en) | Address category determining method, equipment and system for text data | |
CN109241319B (en) | Picture retrieval method, device, server and storage medium | |
CN109325108B (en) | Query processing method, device, server and storage medium | |
JP7451747B2 (en) | Methods, devices, equipment and computer readable storage media for searching content | |
CN110543592A (en) | Information searching method and device and computer equipment | |
CN110688434B (en) | Method, device, equipment and medium for processing interest points | |
CN108491543A (en) | Image search method, image storage method and image indexing system | |
CN116917887A (en) | Query processing using attention-based ranking system | |
US20230376691A1 (en) | Method for point-of-interest information management, electronic device, and storage medium | |
CN110598122B (en) | Social group mining method, device, equipment and storage medium | |
CN104615620A (en) | Map search type identification method and device and map search method and system | |
CN109165283B (en) | Resource recommendation method, device, equipment and storage medium | |
CN111723165A (en) | Address interest point determining method, device and system | |
CN115827864A (en) | Processing method for automatic classification of bulletins | |
CN115017385A (en) | Article searching method, device, equipment and storage medium | |
De Rouck et al. | Georeferencing Wikipedia pages using language models from Flickr | |
US11842165B2 (en) | Context-based image tag translation | |
CN113515687B (en) | Logistics information acquisition method and device | |
CN114328844A (en) | Text data set management method, device, equipment and storage medium | |
CN114385891A (en) | Data searching method and device, electronic equipment and storage medium | |
CN112328630A (en) | Data query method, device, equipment and storage medium |
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 |