CN111984749A

CN111984749A - Method and device for ordering interest points

Info

Publication number: CN111984749A
Application number: CN201910435833.3A
Authority: CN
Inventors: 杨雨然; 孙翔东
Original assignee: Beijing Sogou Technology Development Co Ltd
Current assignee: Beijing Sogou Technology Development Co Ltd
Priority date: 2019-05-23
Filing date: 2019-05-23
Publication date: 2020-11-24

Abstract

The embodiment of the application discloses a method and a device for ordering interest points, which can determine word weights corresponding to participles in keywords when the keywords for searching the interest points are obtained. The keywords generally include one or more participles, the word weight of any participle can identify the degree of correlation between the participle and the search purpose embodied by the keyword, that is, the higher the word weight of a participle is, the more the participle conforms to the search purpose embodied by the keyword, so that in the process of determining the ranking value of the interest point in the search result corresponding to the keyword, the word weight can be used as a basis for adjusting the ranking value of the interest point.

Description

Method and device for ordering interest points

Technical Field

The present application relates to the field of map search, and in particular, to a method and an apparatus for ranking points of interest.

Background

The electronic map configured on the terminal can provide convenient map guide service for the user. For example, the electronic map may show a Point of Interest (POI) corresponding to the keyword to the user as a guide at a corresponding position of the electronic map according to the keyword input by the user.

At present, the ranking order of the POIs corresponding to the keywords is mainly determined according to the character repetition degree, and the higher the repetition degree of the POIs and the keywords is, the higher the priority is to be displayed to the user. For example, the keyword is "beijing university of stamp and technology", the POI corresponding to the keyword includes "beijing university of post and telecommunications", and the POI "beijing university of post and telecommunications" and the character of the keyword have a high repetition degree, and are arranged at a very front display position, so that misdirection may be caused to a user, and poor user search experience is brought.

Therefore, the POI ranking effect in the above manner is not good, and a POI with low relevance may be preferentially displayed to the user.

Disclosure of Invention

In order to solve the technical problem, the application provides a method and a device for ordering the interest points, which preferentially show the interest points with higher possibility of high word weight word segmentation in the keywords to the user, thereby reducing the possibility of misleading the user and improving the search experience of the user.

The embodiment of the application discloses the following technical scheme:

in a first aspect, an embodiment of the present application provides a method for ranking points of interest, where the method includes:

acquiring a keyword for searching for a point of interest;

performing word segmentation on the keywords, and determining word weights corresponding to the word segmentation in the keywords, wherein the word weight corresponding to any word segmentation in the keywords is used for identifying the degree of correlation between the word segmentation and the search purpose embodied by the keywords;

and determining the ranking value of the interest points in the search result corresponding to the keyword according to the word weight corresponding to the participle in the keyword.

Optionally, the determining the word weight corresponding to the segmentation word in the keyword includes:

determining word weights corresponding to the segments in the keywords according to a click model, wherein the click model is obtained by training according to historical search data of a user, and the historical search data of the user comprises historical keywords and interest points selected through the historical keywords.

Optionally, after determining the word weight corresponding to the segmentation word in the keyword, the method further includes:

determining the text relevance of the interest points in the search result and the keywords respectively according to the word weight corresponding to the participles in the keywords;

The determining the ranking value of the interest points in the search result corresponding to the keyword according to the word weight corresponding to the segmentation word in the keyword comprises the following steps:

and determining the ranking values of the interest points in the search results corresponding to the keywords according to the text relevance of the interest points in the search results and the keywords respectively.

Optionally, the method further includes:

determining the type relevance of the interest points in the search results and the keywords respectively; the type relevance of any interest point in the search results and the keyword is used for identifying the degree of relevance of the entity type to which the interest point belongs and the entity type to which the keyword belongs;

the determining the ranking values of the interest points in the search results corresponding to the keywords according to the text relevance between the interest points in the search results and the keywords respectively comprises the following steps:

and determining the ranking value of the interest points in the search result corresponding to the keywords according to the text relevance and the type relevance of the interest points in the search result and the keywords respectively.

Optionally, the target interest point is any interest point in the search result, and the type correlation between the target interest point and the keyword is obtained according to the following method:

Determining a first probability distribution function of the target interest point belonging to each entity type in the entity type set;

determining a second probability distribution function of the keyword belonging to each entity type in the entity type set;

determining a type relevance of the target interest point to the keyword according to the first probability distribution function and the second probability distribution function.

Optionally, the determining that the target point of interest belongs to the first probability distribution function of each entity type in the entity type set includes:

and determining a first probability distribution function of the target interest point belonging to each entity type in the entity type set according to the suffix word characteristics of the target interest point.

Optionally, the determining, according to the text relevance and the type relevance of the interest points in the search result with the keywords, the ranking values of the interest points in the search result corresponding to the keywords includes:

and determining the ranking value of the interest points in the search result corresponding to the keywords according to the text relevance and the type relevance of the interest points in the search result and the keywords, the word weight corresponding to the participles in the keywords and the second probability distribution function.

Optionally, before displaying the interest points in the search result according to the ranking value, the method further includes:

and screening out the interest points for display according to the difference value of the ranking values of the interest points in the search result.

In a second aspect, an embodiment of the present application provides an interest point ranking device, where the device includes an obtaining unit, a first determining unit, and a second determining unit:

the acquisition unit is used for acquiring keywords for searching the interest points;

the first determining unit is used for segmenting the keywords and determining word weights corresponding to the segments in the keywords, wherein the word weight corresponding to any segment in the keywords is used for identifying the degree of correlation between the segment and the search purpose embodied by the keywords;

and the second determining unit is used for determining the ranking value of the interest points in the search result corresponding to the keyword according to the word weight corresponding to the participle in the keyword.

Optionally, the first determining unit is specifically configured to:

Optionally, the first determining unit is further specifically configured to:

after the word weight corresponding to the segmentation in the keywords is determined, determining the text relevance of the interest points in the search results and the keywords respectively according to the word weight corresponding to the segmentation in the keywords;

the second determining unit is specifically configured to:

Optionally, the first determining unit is further specifically configured to:

the second determining unit is further specifically configured to:

Optionally, the first determining unit is further specifically configured to:

Determining a first probability distribution function of the target interest point belonging to each entity type in the entity type set, wherein the target interest point is any interest point in the search result;

Optionally, the first determining unit is further specifically configured to:

Optionally, the second determining unit is further specifically configured to:

and screening out the interest points for display according to the difference value of the ranking values of the interest points in the search result before displaying the interest points in the search result according to the ranking values.

In a third aspect, an embodiment of the present application provides a point of interest ranking apparatus, including a memory, and one or more programs, where the one or more programs are stored in the memory, and configured to be executed by one or more processors, where the one or more programs include instructions for:

acquiring a keyword for searching for a point of interest;

Optionally, the processor executing the one or more programs includes instructions for:

In a fourth aspect, embodiments of the present application provide a machine-readable medium having stored thereon instructions, which, when executed by one or more processors, cause an apparatus to perform the method of point of interest ranking as described in one or more of the first aspects.

According to the technical scheme, when the keywords for searching the interest points are obtained, the word weights corresponding to the participles in the keywords can be determined. The keywords generally include one or more participles, the word weight of any participle can identify the degree of correlation between the participle and the search purpose embodied by the keyword, that is, the higher the word weight of a participle is, the more the participle conforms to the search purpose embodied by the keyword, so that in the process of determining the ranking value of the interest point in the search result corresponding to the keyword, the word weight can be used as a basis for adjusting the ranking value of the interest point.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.

Fig. 1 is a flowchart of a method for ranking points of interest according to an embodiment of the present disclosure;

fig. 2 is a structural diagram of an interest point ranking device according to an embodiment of the present application;

fig. 3 is a block diagram of a terminal device 300 according to an embodiment of the present disclosure;

fig. 4 is a block diagram of a server according to an embodiment of the present application.

Detailed Description

Embodiments of the present application are described below with reference to the accompanying drawings.

At present, the mode of determining the ordering of the POI corresponding to the keyword according to the character repetition degree is likely to cause misleading to the user and bring about poor user search experience. Therefore, how to improve the ordering effect of the POI is a problem which needs to be solved urgently at present.

Therefore, the embodiment of the application provides a method for ordering points of interest, and the core idea of the method is to use word weights corresponding to participles in keywords as ordering bases of POIs in search results. The word segmentation of the keyword may be a word or a word included in the keyword, for example: the keyword of "Beijing university" may include three segments of "Beijing", "teacher" and "university". The word weight of each word is used for identifying the degree of relevance of the word and the search purpose embodied by the keyword, namely, if the word weight of one word is higher, the word is more consistent with the search purpose embodied by the keyword. For example: for the keyword 'Beijing university of education', the embodied search purpose is the Beijing university of education, but not other Beijing university of post and telecommunications, or Beijing university of science and engineering, etc. Therefore, the term "scope" is more suitable for the search purpose embodied by the keyword "Beijing teacher university" than "Beijing university" and "university", and the word weights corresponding to the term "scope" are respectively higher than those corresponding to the two terms "Beijing" and "university".

The method for ordering the interest points provided by the embodiment of the application specifically comprises the following steps: after the key words used for searching the POI are obtained, determining the word weight corresponding to each participle in the key words; and determining the ranking value of the POI in the search result according to the word weight corresponding to the segmentation word in the keyword. The sorting method can be applied to processing equipment, and the processing equipment can be a terminal, a computer, a server and the like.

When the ranking value of the POI in the search result is determined according to the word weight corresponding to the participle in the keyword, the ranking value of the POI in the search result can be further determined according to the correlation between the POI in the search result and the high-word-weight participle in the keyword. For example: if the relevance between a certain POI in the search result and the high-word-weight participle in the keyword is large, the relevance between the POI and the search purpose embodied by the keyword is high, and the ranking value of the POI can be relatively increased; if the relevance between a certain POI in the search result and the high-word-weight participle in the keyword is small, the relevance between the POI and the search purpose embodied by the keyword is low, and the ranking value of the POI can be relatively reduced.

The following is a description with specific examples: for the keyword of Beijing university, the word weight of the participle of "teacher" is the highest. If the search result POI includes "beijing university", "beijing university of mail and telecommunications", and for the POI "beijing university of stamp and telecommunications", because the high word weight participle of "master" is included therein, it can be indicated that the correlation between the POI and the high word weight participle is relatively large, and the degree of correlation between the POI and the search purpose embodied by the keyword "beijing university of stamp and telecommunications" is relatively high, so that the ranking value of the POI "beijing university of stamp and telecommunications" can be relatively increased; as for the POI "beijing post and telecommunications university" in the search result, because the high word weight participle of "tutor" is not included, the correlation between the POI and the high word weight participle in the keyword can be represented to be small, the correlation degree between the POI and the search purpose embodied by the keyword of "beijing tutor and telecommunications university" is low, and the rank value of the POI of the beijing post and telecommunications university can be relatively reduced.

Or, for example, for the keyword "fried chicken", the high-word-weight segmentation may be "fried chicken", and then, if the searched POI includes "XX chicken house" and "XX beer house", since the POI "XX chicken house" has a greater correlation with the high-word-weight segmentation "fried chicken", it indicates that the POI is more correlated with the search purpose embodied by the keyword "fried chicken", so that the ranking value of the POI can be relatively increased; since the POI "XX brewery" has a smaller correlation with the high-word-weight segmented word "fried chicken", it means that the POI is less correlated with the search purpose embodied by the keyword "fried chicken", and thus the ranking value of the POI can be relatively lowered.

Since the higher the ranking value of the POI is, it indicates that the POI has a large correlation with the high-word-weight participles in the keyword, the more the POI and keyword embody the search purpose. Therefore, the POI can be ranked according to the ranking value of the POI in the search result, so that the POI with the higher ranking value is ranked at the position displayed preferentially. Therefore, the POI displayed preferentially has larger relevance with the high-word-weight word, namely, the POI is more consistent with the search purpose embodied by the keyword, so that the possibility of misleading the user is reduced, and the search experience of the user is improved.

The method for ranking points of interest provided by the embodiment of the present application is described next with reference to fig. 1, where the method includes:

s101: keywords for searching for points of interest are obtained.

The keyword may be any character string that is input into the electronic map by the user and used for embodying the purpose of searching by the user, for example, the keyword may be a character string of "university of beijing". After the user inputs the keywords, the electronic map acquires the keywords for searching the interest points.

S102: and performing word segmentation on the keywords, and determining word weights corresponding to the word segmentation in the keywords.

After the electronic map obtains the keywords, the obtained keywords can be segmented, that is, the keywords are divided into one or more words or words, and the word weight corresponding to each segmented word in the keywords is determined. For the word segmentation obtained after the word segmentation is performed on the keyword, the word may be any word or word included in the keyword, it should be noted that the embodiment of the present application does not limit the granularity of the word segmentation performed on the keyword, and for the keyword in the electronic map, the word may be generally divided by the granularity of a single word, for example: the keyword of "family building of university of Beijing teacher" can be divided into four segments of "Beijing", "teacher", "university" and "family building" at the granularity of a single word.

The word weight corresponding to any participle in the keyword can be used for identifying the degree of correlation between the participle and the search purpose embodied by the keyword. Moreover, if the word weight of a word segmentation is higher, the word segmentation is indicated to be more consistent with the search purpose embodied by the keyword. For example: for the keyword 'Beijing university of education', the embodied search purpose is the Beijing university of education, but not other Beijing university of post and telecommunications, or Beijing university of science and engineering, etc. Therefore, the term "scope" is more suitable for the search purpose embodied by the keyword "Beijing teacher university" than "Beijing university" and "university", and the word weights corresponding to the term "scope" are respectively higher than those corresponding to the two terms "Beijing" and "university".

In a possible implementation manner, the method for determining the word weight score corresponding to each participle in the keyword may be to train a click model in advance, so that the function of determining the word weight of each participle in the keyword for the input keyword can be realized. Thus, according to the click model, the word weight corresponding to each participle in the keyword obtained in step S101 can be determined.

The training method of the click model may be to collect and clean hundred million levels of historical search data of the user, where the historical search data of the user may be historical keywords input by the user and POIs selected by the user through the historical keywords. The historical keywords can be keywords input by the user in a historical mode, the click relation between the historical keywords contained in the historical search data and POI selected by the user through the historical keywords is applied, and the click model is trained by utilizing the generalization ability of the machine learning model.

S103: and determining the ranking value of the interest points in the search result corresponding to the keyword according to the word weight corresponding to the participle in the keyword.

The ranking value of the POI in the search result corresponding to the keyword may be a score indicating a degree of relevance between the POI and the search purpose embodied by the keyword.

In the embodiment of the application, the ranking value of each POI in the search result corresponding to the keyword can be determined according to the word weight corresponding to each participle in the keyword.

The word weight of each participle in the keyword can identify the degree of correlation between the participle and the search purpose embodied by the keyword, that is, the higher the word weight of a participle is, the more the participle is consistent with the search purpose embodied by the keyword. Therefore, in the process of determining the ranking value of the POI in the search result corresponding to the keyword, the word weight can be used as a basis for adjusting the ranking value of the interest point, and the ranking value is used for displaying the order of the positions when the interest point is displayed, so that when the interest point is ranked according to the ranking value determined by the word weight, the preferentially displayed interest point has higher word relevance with high word weight, and is more consistent with the search purpose embodied by the keyword, the possibility of misleading the user is reduced, and the search experience of the user is improved.

It should be noted that, in the embodiment of the present application, a specific manner of determining the ranking value of the POI in the search result corresponding to the keyword according to the word weight corresponding to the participle in the keyword is not limited, and a determination manner of adaptability may be selected and set according to different determination requirements.

With respect to step 103, the embodiment of the present application provides a plurality of possible POI ranking values determining manners, which are described below.

The first determination method: and determining the ranking value of the POI according to the text relevance.

In this embodiment, after step S102 is completed, that is, after the word weight corresponding to the segment in the keyword is determined, the text relevance between each POI in the search result and the keyword can be determined according to the feature of the word weight corresponding to the segment in the keyword. The text relevance of the POI and the key word can be used for representing the association degree between the POI and the key word in terms of text. The higher the text relevance of the POI to the keyword, the higher the degree of association in text between them. Therefore, the ranking value of each POI in the search result is determined according to the text relevance of each POI and the key words in the search result.

The word weight corresponding to the participle in the key word is used as an important basis for determining the association degree, namely the text relevance, of the POI and the key word in the text aspect, so that if one POI has the participle with the high word weight in the key word, the association degree, namely the text relevance, of the POI and the key word in the text aspect is relatively high, and the determined ranking value of the POI is relatively high; if a POI does not include a high-word-weight segmentation word in the keyword, the text relevance, which is the degree of association between the POI and the keyword in terms of text, is relatively low, and the determined ranking value of the POI is relatively low, so that the determined ranking value of the POI can more fully embody the search purpose embodied by the keyword.

In addition, in order to ensure that the text correlation performance between the POI and the keyword can more accurately and comprehensively represent the correlation degree between the keyword and the POI in terms of text, the word weight corresponding to the participle in the keyword can be used as a basis for determining the text correlation between the POI and the keyword, and the text similarity, the semantic similarity, the morpheme similarity, the Longest Common Substring (LCS) similarity and the like of the POI and the keyword can be taken into consideration. That is, the text relevance between a POI and a keyword can be determined according to the word weight corresponding to the segmentation word in the keyword, the text similarity between the POI and the keyword, the semantic similarity, the morpheme similarity, and the LCS similarity.

The text similarity may be the similarity between the POI and the keyword in the text content; semantic similarity may be the similarity of a POI to a keyword in terms of expressed semantics; the morpheme similarity can be the similarity of the POI and the key words in the aspect of morphemes, and the morphemes can be the smallest sound-meaning combination in the language; the LCS similarity may be the degree of similarity of the POI to the keyword in terms of the longest common substring.

In a specific implementation, the text similarity, the semantic similarity, the morpheme similarity, and the LCS similarity between a POI and a keyword may be obtained from a pre-trained corresponding calculation model. Wherein each calculation model can be trained according to the calculation method related to the calculation model. The following description will be given only by taking training of the semantic similarity calculation model as an example: based on a neural network language model, hundred million levels of training linguistic data are used as training samples, multi-layer neurons are used for feature extraction, semantic information is expressed in a Word2Vec multi-dimensional vector mode, accordingly, text content information of the samples is deeply mined, and a semantic similarity calculation model with high semantic understanding degree of the text information is obtained through training. Where Word2Vec may be a model for generating Word vectors.

The method for determining the text relevance between the POI and the keyword according to the word weight corresponding to the participle in the keyword, the text similarity between the POI and the keyword, the semantic similarity, the morpheme similarity and the LCS similarity can be realized by training a text relevance calculation model in advance, wherein the text relevance calculation model can determine the text relevance between the POI and the keyword according to the word weight corresponding to the participle in the input keyword, the text similarity between the POI and the keyword, the semantic similarity, the morpheme similarity and the LCS similarity.

The training method for the text relevance calculation model may be that, in order to reduce complexity of processing short text information by the model when the POI or the keyword is a short text and to cope with diversity of the keyword input by the user, the text relevance calculation model is trained based on a Gradient Boosting Decision Tree (GBDT) model by performing multi-level fitting iteration on sample errors, continuously complementing and adjusting samples, and utilizing strong generalization capability of a neural network model. The text relevance of the POI and the key words is calculated through the text relevance calculation model, and the defects of processing short text information through a traditional characteristic calculation mode can be effectively overcome.

The second determination method comprises: and determining the ranking value of the POI according to the text relevance and the type relevance.

In this embodiment, on the basis of determining the text relevance between each POI in the search result and the keyword, the type relevance between each POI in the search result and the keyword may also be determined. The type correlation between the POI and the keyword can be used to identify the degree of correlation between the entity type to which the POI belongs and the entity type to which the keyword belongs. The entity may be various objects objectively existing in the map data and distinguishable from each other, for example, "university of beijing teachers" may be one entity. And the entity types may be categories into which the entities are classified. The type of entity to which the entity belongs, such as "university of Beijing", may be "school". The embodiment of the present application does not limit the granularity of entity type division, such as: the entity types can be schools, stadiums, companies, residential areas and the like according to the larger granularity division, and the entity types can be primary schools, middle schools, universities, badminton halls, table tennis halls, basketball halls and the like according to the smaller granularity division.

Regarding the type relevance of the POI and the keyword, if the type relevance of the POI and the keyword is higher, the entity type to which the POI belongs is more similar to the entity type to which the keyword belongs. For example: for the keyword of "beijing university", the entity type to which the keyword belongs is "school", and the POI searched out according to the keyword includes "beijing university" and "beijing university family building", then, since the entity type to which the POI "beijing university" belongs is "school", which is the same as the entity type to which the keyword belongs, the type correlation of the POI "beijing university" and the keyword is relatively high, and for the POI "beijing university family building", since the entity type to which the POI belongs is "residential district", that is, is far away from the entity type to which the keyword belongs, the type correlation of the POI "beijing university family building" and the keyword is relatively low.

In a possible implementation manner, the method for determining the type correlation between one POI in the search result and the keyword may be that, when the type correlation between one POI in the search result and the keyword needs to be calculated, the POI may be marked as a target interest point, a first probability distribution function that the target interest point belongs to each entity type in the entity type set is determined, and a second probability distribution function that the keyword belongs to each entity type in the entity type set is determined. The entity type set may be a preset and comprehensive set including entity types, such as: the set of entity types may be one that includes "school, residential, sports stadium". Determining a first probability distribution function of the target interest point belonging to each entity type in the entity type set, namely determining a corresponding probability distribution function when the target interest point belongs to each entity type in the entity type set; for example, it is determined that the probability that the target interest point of Beijing university belongs to the entity type of school in the entity type set is 80%, the probability that the target interest point belongs to the entity type of residential district in the entity type set is 15%, and the probability that the target interest point belongs to the entity type of stadium in the entity type set is 5%. And determining a second probability distribution function of the keyword belonging to each entity type in the entity type set, namely determining the probability distribution function corresponding to the keyword belonging to each entity type in the entity type set.

For the determination that the target interest point belongs to the first probability distribution function of each entity type in the entity type set, in one possible implementation, the determination may be performed according to suffix word features of the target interest point. Wherein, the suffix word feature of the target interest point may be a feature related to the suffix word of the target interest point, such as: the suffix word feature of the "beijing university" may be "university", and the suffix word feature of the "beijing university family building" may be "family building", so it may be understood that, compared with determining the entity type of the target interest point according to the complete feature of the target interest point, the entity type of the target interest point may be determined more efficiently and accurately only according to the suffix word feature of the target interest point. For example, for a target interest point of 'Beijing university Home building', when a computer determines the entity type of the target interest point according to the complete characteristic of 'Beijing university Home building', the entity type of the target interest point possibly belonging to 'school' can be determined according to 'Beijing university' in the 'Beijing university Home building', and the entity type of the target interest point possibly belonging to 'residential district' can be determined according to 'Home building', so that the efficiency of determining the entity type of the 'Beijing university Home building' is reduced because the entity type of the 'Beijing university Home building' specifically belonging to 'school' or 'residential district' needs to be determined; and if the entity type of the family building of Beijing university can be more accurately and efficiently determined according to the suffix word characteristic of the family building, wherein the family building belongs to the residential area. Therefore, the first probability distribution function of the target interest point can be determined more efficiently and accurately according to the suffix word characteristics of the target interest point.

Correspondingly, the second probability distribution function of the keyword can be determined based on the suffix character of the keyword, so that the determination efficiency of the entity type is improved.

In a specific implementation, a probability distribution function calculation model may be trained in advance, so that the function of determining, for an input keyword or target interest point, a probability distribution function that the keyword or target interest point belongs to each entity type in the entity type set may be implemented.

The method for training the probability distribution function calculation model may be, for example, to use a ten million-level POI database as a training sample of a machine learning model, determine a probability distribution function of each POI by mining suffix word features of each POI in the POI database, and using mutual information to calculate a correlation between the suffix word features and each entity type included in the entity type set, and further train the probability distribution function calculation model by using a strong generalization capability of the neural network model. The mutual information may be an amount of information contained in one random variable with respect to another random variable. In this way, the probability distribution function calculation model may determine the second probability distribution function of the keyword and the first probability distribution function of the target interest point according to the suffix feature of the input keyword or the target interest point.

Alternatively, the training method of the probability distribution function calculation model may be, for example: the massive user historical search data is used as a training sample of a machine learning model, the corresponding relation between historical keywords input by the user history and POI selected by the user through the historical keywords is analyzed, the search purpose embodied by the historical keywords is analyzed, the correlation between the historical keywords and each entity type in the entity type set is determined according to the search purpose embodied by the historical keywords, and the probability distribution function calculation model is trained by utilizing the strong generalization capability of the neural network model. In this way, the probability distribution function calculation model can determine the second probability distribution function of the keyword and the first probability distribution function of the target interest point by analyzing the search purpose embodied by the keyword or the target interest point for the input keyword or the target interest point.

In this way, after the first probability distribution function of the target interest point and the second probability distribution function of the keyword are determined, the type correlation between the target interest point and the keyword is determined according to the first probability distribution function of the target interest point and the second probability distribution function of the keyword. The specific determination method of the type correlation between the target interest point and the keyword may be, for example, calculating the first probability distribution function and the second probability distribution function according to a cross entropy manner, thereby determining a distance between the two probability distribution functions, and further determining the type correlation between the target interest point and the keyword. The cross entropy may be a calculation method for measuring the difference information between two probability distributions.

It can be understood that any POI in the search results can be marked as a target point of interest, and according to the above method, the type relevance of each POI to the keyword is determined.

After the type relevance between each POI in the search result and the key words is determined, the ranking value of each POI in the search result corresponding to the key words can be determined according to the text relevance and the type relevance between each POI in the search result and the key words.

The second method for determining the ranking value of the POI uses the type relevance of the POI and the keyword as a dimension characteristic for determining the ranking value of the POI, in addition to the text relevance of the POI and the keyword, that is, when two POIs have the same text relevance with the keyword respectively, if the type relevance of one POI is higher, the POI is closer to the entity type to which the keyword belongs, and the POI is determined to have a higher ranking value; if the lower the type relevance of a POI and the keyword is, the POI is more different from the entity type of the keyword, and the POI is determined to have a lower ranking value. Therefore, the type relevance of the POI and the key words is also used as a basis for determining the POI ranking value, so that the ranking value can more accurately represent the relevance degree of the POI and the search purpose embodied by the key words.

The third determination method: and determining the ranking value of the POI according to the text relevance, the type relevance, the word weight corresponding to the segmentation words in the key words and the second probability distribution function.

In this embodiment, on the basis of determining the text relevance, the type relevance, and the second probability distribution function of each POI and the keyword in the search result, and the word weight corresponding to the participle in the keyword, the ranking value of each POI in the search result corresponding to the keyword can be determined according to the text relevance, the type relevance, and the second probability distribution function of each POI and the keyword in the search result, and the word weight corresponding to the participle in the keyword.

According to the third method for determining the ranking value of the POI, besides the text relevance and the type relevance of the POI and the key words, the word weight corresponding to the participles in the key words and the second probability distribution function of the key words are used as the dimensional characteristics for determining the ranking value of the POI. Because the word weight corresponding to the participle in the keyword can represent the search purpose embodied by the keyword to a large extent, and the second probability distribution function of the keyword can represent the entity type of the search purpose embodied by the keyword to a large extent, that is, both the two parameters can represent the search purpose of the user to a large extent, the two characteristics are used as the dimension characteristics for determining the ranking value of the POI, and the influence of the search purpose of the user in determining the ranking value of the POI can be strengthened.

In the three methods for determining the ranking value of the POI according to the word weight corresponding to the participle in the keyword, the specific implementation modes of the three methods may be that a ranking value determination model is trained in advance, so that the method can effectively fuse one or more dimension characteristics which affect the ranking value of the POI and determine a comprehensive ranking value.

Thus, for the first determination method, the text relevance of each POI in the search result to the keyword can be input into the ranking value determination model, and the ranking value determination model can determine and output the ranking value of each POI according to the text relevance feature of each POI to the keyword. For the second determination method, the text relevance and the type relevance of each POI and the keyword in the search result can be input into the ranking value determination model, and the ranking value determination model can effectively fuse the two dimensional features of the text relevance and the type relevance of each POI, and determine and output a comprehensive ranking value of each POI in the search result. For the third determination method, the text relevance, the type relevance, the second probability distribution function of the key words and the word weights corresponding to the participles in the key words in the search results can be input into the ranking value determination model, the ranking value determination model can effectively fuse the four dimensional features of the text relevance, the type relevance, the second probability distribution function of the key words and the word weights corresponding to the participles in the key words, and determine and output the comprehensive ranking value of each POI in the search results.

The training method of the ranking value determination model may be based on a Learning ranking model of a Learning2Rank machine, and further train the ranking model by using the fitting capability of a neural network model.

In addition, in order to determine the ranking value of each POI in the search result more comprehensively and accurately, other dimensional features influencing the ranking value of the POIs may be added on the basis of the three methods, such as: the dimension characteristics such as the grade, the popularity information and the comment number of the POI and the like related to the attribute of the POI can be obtained; and the user's geographic region, city, distance, etc. dimensional characteristics related to the user's geographic location, etc. Therefore, the dimension characteristics influencing the POI ranking value are effectively fused, and the ranking value of each POI in the search result can be determined more comprehensively and accurately.

After determining the ranking value of each POI in the search result, the electronic map may rank the POIs according to the high-low order of the ranking values of the POIs. It can be understood that, in the POIs searched according to the keyword, a situation that the ranking value of some POIs is higher and the ranking value of some POIs is lower may occur, that is, the degree of correlation between some POIs and the search purpose embodied by the keyword is higher, and the degree of correlation between some POIs and the search purpose embodied by the keyword is lower, so before showing the POIs in the search result, POIs with higher degree of correlation with the search purpose embodied by the keyword, that is, POIs with higher ranking values, may be screened out and shown to the user.

In a possible implementation manner, the method for screening POIs with higher ranking values from the search result may be that, before the POIs in the search result are displayed according to the ranking values of the POIs, POIs for display may be screened according to a difference value of the ranking values of the POIs in the search result. For example: the POIs in the search result include A, B, C, D, E, F, and the determined ranking values of the POIs are respectively: 92. 90, 89, 40, 38, 36, the POIs being ranked according to the ranking value in the order: A. b, C, D, E, F, it can be assumed that the POI whose ranking value differs from the maximum ranking value of POIs by less than 50 is used as the POI for presentation, so that since the difference between the ranking values of POIs a and B, C and the ranking value of POI a is less than 50, the POIs with higher ranking values of A, B, C can be screened out as the POIs for presentation to the user.

The mode of screening the POI with the higher ranking value can dynamically select the threshold of POI display according to different search requirements of the user, and filters the POI which is not in line with the search requirements of the user, so as to simplify the display result, adjust the visual field range of the user and improve the user experience.

In addition, in order to meet the challenges of high concurrency and low response time in map search ranking, the calculation involved in map search ranking is divided into an online real-time calculation part and an offline calculation part in the embodiment of the application, wherein the models which are introduced in the foregoing and realize different functions are trained offline, the current keyword search of the user, the calculation of the POI ranking value and the like are calculated online in real time, and therefore the online calculation amount and time are reduced. Moreover, the decoding programs of the models are optimized, so that a plurality of POI in the search result can be decoded in parallel, and the response time of the system is greatly reduced.

Based on the method for ranking points of interest provided in the embodiment corresponding to fig. 1, the embodiment of the present application provides a device for ranking points of interest, referring to fig. 2, which shows a structure diagram of the device for ranking points of interest provided in the embodiment of the present application, and as shown in fig. 2, the device includes an obtaining unit 201, a first determining unit 202, and a second determining unit 203:

the acquiring unit 201 is configured to acquire a keyword for searching for a point of interest;

the first determining unit 202 is configured to perform word segmentation on the keyword, and determine a word weight corresponding to a word in the keyword, where the word weight corresponding to any word in the keyword is used to identify a degree of correlation between the word and a search purpose embodied by the keyword;

The second determining unit 203 is configured to determine, according to the word weight corresponding to the segmented word in the keyword, the rank value of the interest point in the search result corresponding to the keyword.

Optionally, the first determining unit 202 is further specifically configured to:

the second determining unit 203 is specifically configured to:

The second determining unit is further specifically configured to:

Optionally, the second determining unit 203 is further specifically configured to:

In summary, when a keyword for searching for a point of interest is obtained, word weights corresponding to the participles in the keyword may be determined. The keywords generally include one or more participles, the word weight of any participle can identify the degree of correlation between the participle and the search purpose embodied by the keyword, that is, the higher the word weight of a participle is, the more the participle conforms to the search purpose embodied by the keyword, so that in the process of determining the ranking value of the interest point in the search result corresponding to the keyword, the word weight can be used as a basis for adjusting the ranking value of the interest point.

Based on the foregoing provided interest point ranking method and apparatus, this embodiment provides an interest point ranking device, where the interest point ranking device may be a terminal device, and fig. 3 is a block diagram of a terminal device 300 shown according to an exemplary embodiment. For example, the terminal device 300 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.

Referring to fig. 3, the terminal device 300 may include one or more of the following components: processing component 302, memory 304, power component 306, multimedia component 308, audio component 310, input/output (I/O) interface 312, sensor component 314, and communication component 316.

The processing component 302 generally controls the overall operation of the terminal device 300, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing elements 302 may include one or more processors 320 to execute instructions to perform all or a portion of the steps of the methods described above. Further, the processing component 302 can include one or more modules that facilitate interaction between the processing component 302 and other components. For example, the processing component 302 can include a multimedia module to facilitate interaction between the multimedia component 308 and the processing component 302.

The memory 304 is configured to store various types of data to support operations at the terminal device 300. Examples of such data include instructions for any application or method operating on device 300, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 304 may be implemented by any type or combination of volatile or non-volatile memory devices, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

The power supply component 306 provides power to the various components of the terminal device 300. The power components 306 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the apparatus 300.

The multimedia component 308 comprises a screen providing an output interface between said terminal device 300 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 308 includes a front facing camera and/or a rear facing camera. The front camera and/or the rear camera may receive external multimedia data when the terminal device 300 is in an operation mode, such as a photographing mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

The audio component 310 is configured to output and/or input audio signals. For example, audio component 310 includes a Microphone (MIC) configured to receive external audio signals when apparatus 300 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may further be stored in the memory 304 or transmitted via the communication component 316. In some embodiments, audio component 310 also includes a speaker for outputting audio signals.

The I/O interface 312 provides an interface between the processing component 302 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.

Sensor component 314 includes one or more sensors for providing various aspects of status assessment for terminal device 300. For example, sensor assembly 314 may detect an open/closed state of terminal device 300, the relative positioning of components, such as a display and keypad of terminal device 300, sensor assembly 314 may also detect a change in the position of terminal device 300 or a component of terminal device 300, the presence or absence of user contact with terminal device 300, orientation or acceleration/deceleration of terminal device 300, and a change in the temperature of terminal device 300. Sensor assembly 314 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 314 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 314 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 316 is configured to facilitate communication between the terminal device 300 and other devices in a wired or wireless manner. The terminal device 300 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication section 316 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 316 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the terminal device 300 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described methods.

In an exemplary embodiment, a non-transitory computer readable storage medium comprising instructions, such as the memory 304 comprising instructions, executable by the processor 320 of the terminal device 300 to perform the above-described method is also provided. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

A non-transitory computer readable storage medium having instructions therein which, when executed by a processor of a mobile terminal, enable the mobile terminal to perform a point of interest ranking method, the method comprising:

acquiring a keyword for searching for a point of interest;

Fig. 4 is a schematic structural diagram of a server in an embodiment of the present invention. The server 400 may vary significantly due to configuration or performance, and may include one or more Central Processing Units (CPUs) 422 (e.g., one or more processors) and memory 432, one or more storage media 430 (e.g., one or more mass storage devices) storing applications 442 or data 444. Wherein the memory 432 and storage medium 430 may be transient or persistent storage. The program stored on the storage medium 430 may include one or more modules (not shown), each of which may include a series of instruction operations for the server. Still further, the central processor 422 may be arranged to communicate with the storage medium 430, and execute a series of instruction operations in the storage medium 430 on the server 400.

The server 400 may also include one or more power supplies 426, one or more wired or wireless network interfaces 450, one or more input-output interfaces 458, one or more keyboards 456, and/or one or more operating systems 441, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and so forth.

Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium may be at least one of the following media: various media that can store program codes, such as read-only memory (ROM), RAM, magnetic disk, or optical disk.

It should be noted that, in the present specification, all the embodiments are described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, they are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described embodiments of the apparatus and system are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

The above description is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for ranking points of interest, the method comprising:

acquiring a keyword for searching for a point of interest;

2. The method of claim 1, wherein the determining the word weight corresponding to the segmented word in the keyword comprises:

3. The method of claim 1, wherein after determining the word weight corresponding to the participle in the keyword, the method further comprises:

4. The method of claim 3, further comprising:

5. The method of claim 4, wherein a target interest point is any interest point in the search results, and the type relevance of the target interest point to the keyword is obtained according to the following method:

6. The method of claim 5, wherein the determining that the target point of interest belongs to a first probability distribution function for each entity type in the set of entity types comprises:

7. The method according to claim 5, wherein determining the ranking value of the interest points in the search result corresponding to the keyword according to the text relevance and the type relevance of the interest points in the search result to the keyword respectively comprises:

8. An interest point ranking device, characterized in that the device comprises an acquisition unit, a first determination unit and a second determination unit:

9. A point of interest ranking apparatus comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured for execution by one or more processors the one or more programs comprising instructions for:

acquiring a keyword for searching for a point of interest;

10. A machine-readable medium having stored thereon instructions, which when executed by one or more processors, cause an apparatus to perform the method of point of interest ranking as recited in one or more of claims 1-7.