CN113536156A - Search result ordering method, model construction method, device, equipment and medium - Google Patents

Search result ordering method, model construction method, device, equipment and medium Download PDF

Info

Publication number
CN113536156A
CN113536156A CN202010287348.9A CN202010287348A CN113536156A CN 113536156 A CN113536156 A CN 113536156A CN 202010287348 A CN202010287348 A CN 202010287348A CN 113536156 A CN113536156 A CN 113536156A
Authority
CN
China
Prior art keywords
result
query
ranking
search result
attribute
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010287348.9A
Other languages
Chinese (zh)
Other versions
CN113536156B (en
Inventor
钟贤德
王康
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Baidu Online Network Technology Beijing Co Ltd
Original Assignee
Baidu Online Network Technology Beijing Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Baidu Online Network Technology Beijing Co Ltd filed Critical Baidu Online Network Technology Beijing Co Ltd
Priority to CN202010287348.9A priority Critical patent/CN113536156B/en
Publication of CN113536156A publication Critical patent/CN113536156A/en
Application granted granted Critical
Publication of CN113536156B publication Critical patent/CN113536156B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—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
    • G06F16/9538—Presentation of query results
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0603—Catalogue ordering

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Accounting & Taxation (AREA)
  • General Engineering & Computer Science (AREA)
  • Finance (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The embodiment of the application discloses a search result ordering method, a model construction method, a device, equipment and a medium, and relates to an intelligent search technology. The search result ordering method comprises the following steps: determining a search result according to the user query; constructing at least one ordering attribute of the search result by utilizing the correlation between the user query language and the search result; a ranking result of the search result is determined based on at least one ranking attribute of the search result using the ranking model. According to the method and the device, the ranking effect of the search results can be improved on the basis of not increasing the calculation amount, the matching relevance of user search is improved, and the search response efficiency is ensured.

Description

Search result ordering method, model construction method, device, equipment and medium
Technical Field
The embodiment of the application relates to a computer technology, in particular to an intelligent search technology, and particularly relates to a search result ordering method, a model construction method, a device, equipment and a medium.
Background
In the e-commerce field of Business-to-Business (B2B), aiming at commodity search, in order to ensure the retrieval ranking effect of recalled commodities, the commonly adopted retrieval ranking method for each service party comprises: an ordering method based on entity identification, an ordering method based on word sequence matching, an ordering method based on semantic matching and the like.
However, the sorting method based on entity identification can only be applied in a scenario where the commodity materials are accurately labeled with categories or corresponding entity names, and the sorting accuracy is limited by the entity or category identification accuracy of a user query (query); the sorting method based on word sequence matching has great dependence on the sequence of word occurrence in user query and a commodity title (title), and greatly influences the judgment accuracy of the word sequence when the same word appears in the query or the title for multiple times, thereby influencing the sorting effect; although the sorting effect is good to a certain extent based on the semantic matching sorting method, the related calculation amount is large, and the method is still limited to be used in some scenes with high requirements on search response time.
Disclosure of Invention
The embodiment of the application discloses a search result sorting method, a model construction method, a device, equipment and a medium, so that on the basis of not increasing the calculated amount, the sorting effect of the search results is improved, the matching relevance of user search is improved, and the search response efficiency is ensured.
In a first aspect, an embodiment of the present application discloses a search result ranking method, including:
determining a search result according to the user query;
Constructing at least one ordering attribute of the search result by utilizing the relevance of the user query language and the search result;
determining, using a ranking model, a ranking result for the search result based on at least one ranking attribute of the search result.
In a second aspect, an embodiment of the present application further discloses a method for constructing a ranking model, including:
determining a sample result according to the sample query;
constructing at least one ordering attribute of the sample result by utilizing the correlation between the sample query and the sample result;
and constructing a ranking model by using at least one ranking attribute of the sample result.
In a third aspect, an embodiment of the present application further discloses a search result ranking device, including:
the search result determining module is used for determining a search result according to the user query;
the ranking attribute construction module is used for constructing at least one ranking attribute of the search result by utilizing the correlation between the user query language and the search result;
a ranking result determination module to determine a ranking result of the search result based on at least one ranking attribute of the search result using a ranking model.
In a fourth aspect, an embodiment of the present application further discloses a ranking model building apparatus, including:
the sample result determining module is used for determining a sample result according to the sample query;
the ranking attribute construction module is used for constructing at least one ranking attribute of the sample result by utilizing the correlation between the sample query language and the sample result;
and the sequencing model building module is used for building a sequencing model by utilizing at least one sequencing attribute of the sample result.
In a fifth aspect, an embodiment of the present application further discloses an electronic device, including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a search result ranking method as described in any of the embodiments of the present application or to perform a ranking model construction method as described in any of the embodiments of the present application.
In a sixth aspect, an embodiment of the present application further discloses a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the search result ranking method according to any of the embodiments of the present application, or execute the ranking model constructing method according to any of the embodiments of the present application.
According to the technical scheme of the embodiment of the application, at least one ordering attribute of the search result is constructed by utilizing the correlation between the user query and the search result, then the constructed ordering attribute is used as the input of a pre-trained ordering model, and the ordering result of the search result is obtained through model processing, wherein the constructed ordering attribute can reflect the correlation between the search result and the user query from different angles, so that the ordering effect of the search result is improved on the basis of not increasing the calculated amount, the matching correlation of the search of the user is improved, and the search response efficiency is ensured.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present application, nor do they limit the scope of the present application. Other features of the present application will become apparent from the following description.
Drawings
The drawings are included to provide a better understanding of the present solution and are not intended to limit the present application. Wherein:
FIG. 1 is a flow chart of a method of ranking search results according to an embodiment of the present application;
FIG. 2 is a flow chart of another search result ranking method disclosed in accordance with an embodiment of the present application;
FIG. 3 is a flow chart of another search result ranking method disclosed in accordance with an embodiment of the present application;
FIG. 4 is a flow chart of another search result ranking method disclosed in accordance with an embodiment of the present application;
FIG. 5 is a flow chart of a method of constructing a ranking model according to an embodiment of the present application;
FIG. 6 is a schematic structural diagram of a search result sorting apparatus according to an embodiment of the present application;
FIG. 7 is a schematic structural diagram of an ordering model constructing apparatus according to an embodiment of the present application;
fig. 8 is a block diagram of an electronic device disclosed according to an embodiment of the present application.
Detailed Description
The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
Fig. 1 is a flowchart of a search result ranking method disclosed in an embodiment of the present application, which may be applied to a case of ranking search results for recall, and which may be performed by a search result ranking apparatus. The search result ranking means may be implemented in software and/or hardware, and may be integrated on any electronic device with computing capabilities, such as a server.
As shown in fig. 1, the search result ranking method disclosed in the embodiment of the present application may include:
s101, determining a search result according to the user query.
The user query (query) is used for characterizing the search requirement of the user, and specifically, the query may be in the form of a word or a phrase. The search engine may recall a set number of search results in the background using a search algorithm based on the user query. Taking the commodity retrieval scenario as an example, the search engine may determine a recalled commodity list according to the user query.
S102, constructing at least one ordering attribute of the search result by utilizing the correlation between the user query language and the search result.
In the embodiment of the application, for each search result of the user query, a corresponding at least one ranking attribute is constructed by using a preset attribute calculation mode. The at least one ranking attribute may reflect the relevance of the search results to the user query from different perspectives. The relevance of the user query to the search results includes, but is not limited to: the matching condition of the user query language and the search result description information in terms of contained words, the matching condition of the category attributes corresponding to the user query language and the search results, the matching condition of the text formats of the user query language and the search result description information and the like. The matching condition of the user query language and the search result description information in terms of contained terms may include, but is not limited to: the matching condition between the respective core terms (or main terms), whether the final word of the search result description information is contained in the user query language, the matching condition between the expansion word based on the user query language and the search result description information, and the like. The search result description information may be generally title or subject information of the search result, the expansion word of the user query may be generally synonym or synonym of the word in the user query, the core word in the user query may be used for reflecting the core of the user search requirement, and the core word in the search result description information may be used for reflecting the core of the search result. The search result corresponds to at least one dimension of the ranking attribute in each matching aspect of the user query.
In addition, in the embodiment of the application, at least one sorting attribute of the search result is an attribute which is extracted and screened out based on the statistical rule and has higher discrimination to the search result, so that the accuracy of sorting of the subsequent search result can be ensured.
Illustratively, the constructing at least one ranking attribute of the search result by utilizing the relevance of the user query and the search result includes: respectively determining query keywords of a user query and result keywords of search result description information; and constructing at least one ordering attribute of the search result according to the matching processing result of the query keyword and the result keyword.
The query keywords and the result keywords can be obtained based on any available word segmentation technology, the specific implementation of the word segmentation process is not limited in the embodiment of the application, and in the word segmentation process, the words in the query keywords and the result keywords can be distinguished based on the importance degree of each word in the user query language and the search result description information, and main query words (or core query words) in the query keywords and main result words (or core result words) in the result keywords are respectively screened out. The matching processing result of the query keyword and the result keyword includes but is not limited to: the matching result between the main query word and the main result word, the matching result between the main query word and the result keyword, the matching result between the main result word and the query keyword, and the like. Regarding matching between words, any available word matching algorithm may be used to implement matching between words, such as word similarity calculation, and the present embodiment is not limited in particular.
Illustratively, the constructing at least one ranking attribute of the search result by using the relevance of the user query and the search result further comprises: if the user query language is a mixed text consisting of letters and numbers, matching the search result description information by using a preset regular expression comprising the user query language; and if the matching is successful, recording the result successfully matched as the sorting attribute of the search result.
The format of the preset regular expression including the user query language may include, but is not limited to: /[ ^0-9a-zA-Z ] query [ ^0-9a-zA-Z ]/. If the regular matching fails, the similarity between the text format of the search result description information and the text format of the user query language is small, and the ranking attribute of the dimension can be set to be in a default state.
In some fine search scenarios, mixed text of letters and numbers may be used to represent information such as the model of the item. At this time, by considering the regular matching and determining a ranking attribute of the search result, the phenomenon that the recalled search result contains similar type articles and the ranking of the search result is inaccurate when the query language of the user contains information of the type of the article can be avoided. For example, when a user searches for an item of model LM358, an item of model ALM358 may appear in the recalled search results, and although the item models are very similar, the search results are not what the user really desires. By determining a ranking attribute of the search result based on the regular matching result, the relevance contribution of the dimension attribute to the search result and the user query can be reduced when the ranking model is subsequently utilized to rank the search result, and the influence of the dimension attribute on the accuracy of the ranking result is reduced.
S103, determining a sorting result of the search result based on at least one sorting attribute of the search result by using the sorting model.
The ranking model is pre-trained based on the ranking attributes of the training samples. The construction process of the ordering attribute of the training sample is the same as the construction process of the ordering attribute of the search result in the implementation logic, and the dimensions of the ordering attribute of the training sample are the same as the dimensions of the ordering attribute of the search result. In the process of model construction or training, influence factors or contribution amounts of the ranking attributes of all dimensions on the final ranking of the search results can be learned, so that at least one ranking attribute of each current search result is input into the training model, and the final ranking result of each search result can be accurately output. After the sorting result of the search result is determined, the search result can be issued to the user terminal for the user to check.
On the basis of the foregoing technical solution, optionally, the search result ranking method disclosed in the embodiment of the present application may further include:
determining a sample result according to a sample query, wherein the sample query and the sample result can be collectively referred to as a training sample;
constructing at least one ordering attribute of the sample result by utilizing the correlation between the sample query language and the sample result;
And constructing a ranking model by using at least one ranking attribute of the sample result. Furthermore, the constructed ranking model can be used in the ranking process of the search results recalled in the search scene.
For example, at least one ranking attribute of each sample result may be used as an input of model training, and a ranking labeling result of each sample result may be used as an output of the model training, so as to obtain a ranking model through training. The ranking model may employ any available ranking algorithm, including but not limited to LTR (ranking to rank) algorithm, and the like, such as LambdaMat LTR algorithm, and the like.
According to the technical scheme of the embodiment of the application, at least one ordering attribute of each search result is constructed by utilizing the correlation between the user query and each search result, and then the ordering result of each search result is determined by utilizing the ordering model based on the constructed ordering attribute, wherein the ordering attribute of each search result can reflect the correlation between the search result and the user query from different angles, so that the optimization of the existing search result ordering effect can be realized, the ordering effect is improved, and the problem of poor search result ordering effect in the existing scheme is solved.
In addition, the ranking attribute of the search result in the embodiment of the application does not depend on the query language of the user and the entity identification result of the search result, so that the search ranking effect is not influenced by the accuracy of the entity identification, and even aiming at the search result without the entity mark, the embodiment of the application can also ensure better ranking effect; secondly, the ordering attribute of the search result in the embodiment of the application does not depend on the word order matching result between the user query word and the search result description information, and the appearance order of the words in the user query word and the search result description information does not influence the ordering accuracy of the search result.
Finally, compared with most existing sorting modes, the calculation amount involved in the sorting attribute construction and sorting stages is not obviously increased, and especially compared with a sorting method based on semantic matching, the calculation amount involved in the technical scheme of the embodiment of the application is small, the sorting attribute of the search result can be constructed on line in real time, a large number of user search requirements can be borne, the search response efficiency is ensured, the problem that the search response time is long due to the fact that the sorting calculation amount is large in the existing scheme is solved, and the method and the device can be suitable for scenes with high search response time requirements.
Fig. 2 is a flowchart of another search result ranking method disclosed in the embodiment of the present application, which is further optimized and expanded based on the above technical solution, and can be combined with the above optional embodiments. As shown in fig. 2, the method may include:
s201, determining a search result according to the user query.
S202, respectively determining a query keyword of a user query language and a result keyword of search result description information.
S203, determining the ranking attribute of the search result according to the matching result between the main query word in the query keyword and the main result word in the result keyword.
The main query words and the main result words are words with higher importance degrees in the user query words and the search result description information which are screened in advance respectively. The higher the matching degree between the main query word and the main result word is, the higher the correlation between the search result and the query language of the user to a certain extent is, so that in the process of determining the ranking attribute of the search result, the matching result between the main query word and the main result word is preferentially considered, the effectiveness of the ranking attribute construction can be improved, and the accuracy of the ranking result is further ensured.
For example, the matching result between the main query term and the main result term may include the number of the matching terms therebetween and the occurrence position of the matching terms in the search result description information, that is, the ranking attribute of the search result may be determined according to the number of the matching terms or the occurrence position of the matching terms in the search result description information. The more the number of the matching words is, or the more the positions of the matching words in the description information of the search result are (for example, the Chinese language habit, the fixed phrase word with modification generally precedes and the subject term generally succeeds), it is stated that the correlation between the search result and the query language of the user may be larger, and further, the determined ordering attribute may also contribute a larger amount to the ordering of the search result.
Optionally, determining the ranking attribute of the search result according to the matching result between the main query term in the query keyword and the main result term in the result keyword includes:
determining matching words between the main result words and the main query words;
and if the matching word is determined to be the ending word of the search result description information, recording the determination result of the ending word belonging to the matching word as the ordering attribute of the search result. Taking the habit of Chinese language as an example, generally, the fixed phrase word with modification action is generally in front of the main phrase word, and the subject phrase word is generally behind the main phrase word, if the matching word between the main result word and the main query word belongs to the final phrase of the search result description information, it indicates that the correlation between the search result and the user query word is larger, the probability that the search result meets the search requirement of the user is also larger, and further, the contribution amount of the ordering attribute determined based on the final phrase word to the ordering of the search result is also larger, which is beneficial to improving the ordering accuracy. If the matching term is not the end of the search result description information, the attribute for the dimension may be set to a default state.
Further, determining the matching word as an end word of the search result description information includes:
performing space replacement processing on the text of the set type in the search result description information;
And if the text formed by adding a space after the matching word is determined is the ending text of the search result description information, determining that the matching word is the ending word of the search result description information.
Wherein, the text of the set type includes special symbols, specific adjectives, words representing suppliers, words representing production scale, words representing sales modes, and the like, for example,; "," \ t "," | "," manufacturer "," factory "," sales "," price "," wholesale "," price "," generation "," quality ", and" direct sale ", etc. The texts with the set types do not have substantial influence on the topics of the search results, so that the texts can be subjected to space replacement processing, the effectiveness of judging the relevance between the search results and the query words of the user is improved, and the accurate construction of the ranking attributes is facilitated.
For example, the title of the search result is "high-quality sofa leather sofa manufacturer", the matching word between the main result word and the main query word is "sofa", and the blank is replaced by the "manufacturer" in the title, then the "sofa" is a combined text formed by adding the blank as the ending text of the title of the search result, and the matching word "sofa" can be determined as the ending word of the title of the search result.
S204, determining a first target word with the similarity to the query keyword being smaller than a first preset threshold value in the main result words in the result keywords.
In the embodiment of the application, the ranking attribute of the search result can be constructed under the condition that the word matching is successful, and the ranking attribute of the search result can also be constructed under the condition that the word matching is unsuccessful, so that the multi-angle embodiment of the relevance of the ranking attribute to the search result and the query language of the user is fully shown.
Optionally, determining, in the main result words in the result keywords, a first target word whose similarity to the query keyword is smaller than a first preset threshold includes:
and if candidate words with the similarity smaller than a first preset threshold value with the query keyword exist in the main result words, and a combined text formed by the query word of the user and the candidate words appears in the description information of the search results, determining the candidate words as first target words.
For example, if the user query is "corn", the search result title is "corn shredder, electric shredder", the main result word in the search result title is "shredder", first, "shredder" and "corn" are two words that do not match, the "shredder" is a candidate word, the combined text formed by the user query and the candidate word is "corn shredder", and the combined text appears in the search result title, then "shredder" is determined as the current first target word.
The current first target word is identified again by using the combined text formed by the user query and the candidate words, so that the accuracy of determining the first target word in the result keywords can be improved, and the accuracy of constructing the current ranking attribute is further ensured.
S205, determining the ranking attribute of the search result based on the weight of the first target word in the main result words.
The first target word is also the word in the main result word that does not match the query keyword. The weight of the first target word can be determined in the process of segmenting the search result description information by utilizing a word segmentation technology. If the determined first target word is multiple, the ranking attribute of the search result may be determined based on the weights of the multiple first target words.
For example, the ranking attribute of the search result may be determined based on the weight of at least one first target word in the primary result words and the sum of the weights corresponding to the result keywords (i.e., the sum of the weights of the words), for example, the weights of at least one first target word are first summed and then divided by the sum of the weights corresponding to the result keywords, and the resulting quotient is used as one ranking attribute of the search result.
S206, determining synonyms of the main query words in the query keywords. Namely, the query keywords are expanded moderately, and the matching range of the query keywords and the result keywords is improved.
S207, determining a second target word with the similarity to the result keywords smaller than a second preset threshold in the synonyms and the main query words.
S208, determining the ranking attribute of the search result based on the synonym and the weight of the second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
The second target word is also the synonym of the main query word and the word in the main query word which is not matched with the result keyword. The number of second target words may also be at least one. Illustratively, the ranking attributes of the search results are calculated based on the weight of the synonym and at least one second target word in the main query word, and the corresponding weight sum of the query keywords (i.e., the sum of the weights of the terms). For example, the weights of at least one second target word are summed first, and then divided by the sum of weights corresponding to the query keywords, and the resulting quotient is used as a ranking attribute of the search results.
Operations S203, S204-S205, and S206-S208 are all related to the construction of the search result ranking attribute, and the three operations are not limited by a strict execution order, and may be executed in parallel or in series.
S209, determining a ranking result of the search result based on at least one ranking attribute of the search result by using the ranking model.
According to the technical scheme of the embodiment of the application, the multi-dimensional ranking attribute of the search result is firstly constructed, then the ranking result of the search result is determined by using the ranking model based on the constructed ranking attribute, wherein the constructed ranking attribute can reflect the relevance between the search result and the query language of the user from different angles, the ranking effect of the search result is improved on the basis of not increasing the calculated amount, the matching relevance of the search of the user is improved, and the search response efficiency is ensured; in addition, in the construction process of the search result ranking attribute, the matching result between the main query word and the main result word is preferentially considered, so that the effectiveness of the ranking attribute construction can be improved, and the accuracy of the ranking result is further ensured; moreover, the method and the device for searching the query language can construct the ranking attribute of the search result under the condition that the word matching is successful, and can also be used for constructing the ranking attribute of the search result under the condition that the word matching is unsuccessful, so that the multi-angle embodiment of the relevance of the ranking attribute to the search result and the query language of the user is fully shown.
It should be noted that, the first preset threshold, the second preset threshold, the first target word and the second target word mentioned above are not limited in any order, and are only used for distinguishing expression. The specific values of the first preset threshold and the second preset threshold can be set adaptively.
Fig. 3 is a flowchart of another search result ranking method disclosed in the embodiment of the present application, which is further optimized and expanded based on the above technical solution, and can be combined with the above optional embodiments. As shown in fig. 3, the method may include:
s301, determining a search result according to the user query.
S302, respectively determining a query keyword of a user query language and a result keyword of search result description information.
S303, determining the ranking attribute of the search result according to the matching result between the main query word in the query keyword and the main result word in the result keyword.
S304, determining a first target word with the similarity to the query keyword being smaller than a first preset threshold value in the main result words in the result keywords.
S305, determining the ranking attribute of the search result based on the weight of the first target word in the main result words.
S306, determining synonyms of the main query words in the query keywords.
S307, determining a second target word with the similarity to the result keywords being smaller than a second preset threshold in the synonyms and the main query words.
S308, determining the ranking attribute of the search result based on the weight of the synonym and the weight of the second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
S309, if the preset word list is utilized, it is determined that the query keywords and the result keywords are not matched with the words belonging to the same category in the preset word list, and the result of the matching failure is recorded as the ordering attribute of the search result.
And the preset word list comprises words for describing components or accessories of the search results. For example, the preset vocabulary or referred to as sensitive word may be a storage form of KV vocabulary, where key is a word, value is a category to which the word belongs, and the specific category division may be determined according to the existing industry classification, for example, [ accessory- >1,2,3] [ wiper- >1] [ paint- >2], where the numbers 1,2,3 respectively represent different categories of articles. Words belonging to a category such as "fitting", "wiper", "tail light", "headlight", "engine", "muffler"; words belonging to another category, such as "dye", "paint", "wax", "glue", etc.
If the query keyword and the result keyword are determined not to be matched with the words belonging to the same category in the preset word list, the possibility that the correlation between the search result and the query language of the user is low is shown, the result of failed matching is recorded as the ranking attribute of the search result, the ranking of the search result is participated in, the situation that the user searches for a certain type of articles and searches for components or accessories which are fed back to the articles of the user can be avoided, and the ranking effect of the search result is improved. For example, it is inappropriate for the user to search for "excavator" and feed back to the user "bucket".
S310, determining category attributes matched with the query language and the search result of the user respectively by utilizing the query key words and the result key words.
S311, determining the ranking attribute of the search result according to the matching result of the category attributes between the user query and the search result and the weight of the category attributes.
Category attributes are used to represent categories to which an item belongs, such as clothing, appliances, automobiles, hardware, and so forth, and may typically include three category levels. For example, the category attribute matching the user query and the search result may be determined by using any available category attribute recognition technology, for example, a pre-trained category recognition model for recognizing the category to which the item belongs. In determining the attributes of each category, the weights of the attributes of each category, or the prediction probabilities, may be determined simultaneously. Furthermore, a sort attribute of the search result can be determined according to a matching result of the category attributes between the user query and the search result, that is, whether the corresponding category attributes are consistent or not, and a difference between the weights of the category attributes. Under the condition that the category attributes corresponding to the user query and the search result are initially consistent, the correlation between the search result and the user query is relatively large, the probability that the search result meets the search requirement of the user is relatively high, the current ranking attribute determined based on the category attribute contributes relatively large to the ranking of the search result, and the ranking accuracy is improved.
Optionally, determining the ranking attribute of the search result according to the matching result of the category attributes between the user query and the search result and the weight of the category attributes, including:
traversing category attributes of the user query words, and determining the maximum weight of the category attributes;
and if the category attributes of all levels of the user query language respectively correspond to the category attributes of all levels of the search result, calculating the ordering attribute of the search result based on the determined maximum weight and the weight of the category attributes of all levels of the search result.
Illustratively, the weights corresponding to the category attributes of the three levels of the user query language in sequence are V respectively1、V2、V3Wherein the maximum value Vmax is V1The weight corresponding to the category attributes of the three levels of the search results in sequence is M1、M2、M3When the category attributes of each level of the user query language respectively correspond to the category attributes of each level of the search result, namely the category attributes of each level are consistent, the Vmax and M can be used1、M2、M3Computing a ranking attribute of the search results. For example, will [ Vmax-M1，Vmax-M2，Vmax-M3]As a sort attribute of the search results.
If at least one category attribute in the category attributes of each level of the user query language and the category attributes of each level of the search result are not corresponding, namely inconsistent, the probability that the search result meets the search requirement of the user is low, and at this time, the sequencing attribute related to the category attributes can be determined as a default state.
And S312, if the user query is a mixed text consisting of letters and numbers, matching the search result description information by using a preset regular expression comprising the user query.
And S313, if the matching is successful, recording the result successfully matched as the sorting attribute of the search result.
If the regular matching fails, the similarity between the text format of the search result description information and the text format of the user query language is small, and the ranking attribute of the dimension can be set to be in a default state.
Mixed text of letters and numbers can be used to represent information such as the model number of the article. At this time, by considering the regular matching and determining a ranking attribute of the search result, the phenomenon that the recalled search result comprising similar models is not ranked accurately when the user query language comprises information of the model type of the article can be avoided.
Operations S303, S304-S305, S306-S308, S309, S310-S311, and S312-S313 are all included in the search result ranking attribute, and are not limited to a strict execution order, and may be executed in parallel or in series.
S314, determining a sorting result of the search result based on at least one sorting attribute of the search result by using the sorting model.
On the basis of the above technical solution, optionally, the method disclosed in the embodiment of the present application further includes:
counting the number of target search results of which the category attributes are matched with the set category attributes, wherein the set category attributes are set according to categories of which the similarity with the category attributes of the user query language is greater than a third preset threshold; the specific value of the third preset threshold can be set adaptively; the term "third" is used herein without any sequential limitation, and is used for descriptive purposes only;
and if the statistical quantity meets the attribute clearing condition, determining the ranking attribute determined according to the matching result of the category attributes between the user query language and the search result and the weight of the category attributes as an invalid attribute. The attribute clearing condition may include, but is not limited to, that the statistical number exceeds a number threshold, or that a ratio of the statistical number to the total number of the recalled search results exceeds a proportion threshold, and the like, and may be flexibly set as needed. The specific values of the quantity threshold and the proportion threshold can also be set adaptively.
For example, assuming that the currently set category attribute is an electronic chip class, in the recalled search product, the number of target search results of which the category attribute matches the electronic chip class is greater than the quantity threshold, that is, the category attribute of the currently recalled search result is considered to be generally deviated from the actual search requirement of the user, and therefore, the ranking attribute of the search result in terms of the category attribute may not be considered. By adding the posterior judgment about the category attribute, the construction accuracy and effectiveness of the sequencing attribute are improved.
According to the technical scheme of the embodiment of the application, the multi-dimensional ranking attribute of the search result is firstly constructed, then the ranking result of the search result is determined by using the ranking model based on the constructed ranking attribute, wherein the constructed ranking attribute can reflect the relevance between the search result and the query language of the user from different angles, the ranking effect of the search result is improved on the basis of not increasing the calculated amount, the matching relevance of the search of the user is improved, and the search response efficiency is ensured.
Fig. 4 is a flowchart of another search result ranking method disclosed in the embodiment of the present application, which is further optimized and expanded based on the above technical solution, and can be combined with the above optional embodiments. As shown in fig. 4, the method may include:
s401, determining a search result according to the user query.
S402, performing word segmentation processing on the user query language to obtain query keywords and weights of the query keywords.
S403, determining a main query word in the query keywords according to the weight of the query keywords and the position of the query keywords in the query language of the user.
The word segmentation processing is carried out on the user query language, any available word segmentation technology (or word segmentation technology) in the prior art can be adopted for realizing the word segmentation processing, and in the word segmentation process, the weight of each word can be obtained simultaneously. In the embodiment of the application, the position of the keyword in the query language of the user can be queried, and the weight of the keyword is dynamically adjusted. For example, terms with weights exceeding a first weight threshold in the query keywords may be determined as candidate query terms, and then the positions of the candidate query terms in the user query language are compared, and the weights of the candidate query terms are adjusted according to human expression habits. Taking the habit of Chinese language as an example, generally, the fixed-language word with modification function is generally in front of the fixed-language word, and the subject word is generally behind the fixed-language word, so that the positions of the candidate query words in the query language of the user can be compared pairwise, and the weight of the candidate query words with behind positions is increased by a preset proportion, for example, by 10% of the current weight; and finally, determining the terms exceeding the second weight threshold as the main query terms based on the adjusted weights of the candidate query terms. Of course, the weight adjustment method can be adaptively changed for the case where the phrase habits are different from those of the chinese language. The first weight threshold and the second weight threshold can be adaptively set.
The main query word in the query keyword is determined according to the weight of the query keyword and the position of the query keyword in the query language of the user, so that the search requirement of the user can be accurately grasped, a foundation is laid for subsequently constructing the ranking attribute of the search result based on the main query word, and the accuracy and the effectiveness of the ranking attribute construction are ensured.
S404, performing word segmentation processing on the search result description information to obtain result keywords and weights of the result keywords.
S405, determining main result words in the result keywords according to the weight of the result keywords and the occurrence frequency of the result keywords in the search result description information.
The search result description information includes title or subject information of the gathered result. The word segmentation processing is performed on the search result description information, and any available word segmentation technology (or word segmentation technology) in the prior art can be adopted to implement the word segmentation processing, and in the word segmentation process, the weights of all words can be obtained simultaneously. The greater the weight of the result keyword, the greater the number of times the result keyword appears in the description information, and the greater the importance of the word can be explained. For example, words with weights exceeding a third weight threshold in the result keywords may be determined as candidate result words first, and then words with occurrence times exceeding a frequency threshold in the candidate result words may be determined as main result words. In addition, the limitation on the length of the word character string can be added in the process of determining the main result word, namely, the main result word can be a word of which the occurrence number exceeds a threshold number of times and the length of the character string exceeds a length threshold. The third weight threshold, the number threshold, and the length threshold mentioned above may all be flexibly set, and the embodiment of the present application is not particularly limited.
The main result words in the result keywords are determined according to the weight of the result keywords and the occurrence frequency of the result keywords in the description information, so that the core of the search result description information can be accurately extracted, a foundation is laid for subsequently constructing the ranking attributes of the search results based on the main result words, and the accuracy and the effectiveness of the ranking attribute construction are ensured.
In addition, the operations S402 to S403 and the operations S404 to S405 are not limited to a strict execution order, and may be executed in parallel or in series.
S406, constructing at least one ranking attribute of the search result according to the matching processing result of the query keyword and the result keyword.
The matching processing result of the query keyword and the result keyword may include, but is not limited to: the matching result between the main query word and the main result word, the matching result between the main query word and the result keyword, the matching result between the main result word and the query keyword, and the like. Regarding matching between words, any available word matching algorithm may be used to implement matching between words, such as word similarity calculation, and the present embodiment is not limited in particular.
S407, determining a ranking result of the search result based on at least one ranking attribute of the search result by using the ranking model.
In the embodiment of the application, a main query word in the query keyword is determined according to the weight of the query keyword and the position of the query keyword in the query language of the user, a main result word in the result keyword is determined according to the weight of the result keyword and the occurrence frequency of the result keyword in the description information, a foundation is laid for subsequently constructing the ranking attribute of the search result by accurately extracting the main query word and the main result word, and the accuracy and the effectiveness of the ranking attribute construction are ensured; moreover, the constructed ranking attributes can reflect the relevance of the search results and the query language of the user from different angles, so that the ranking effect of the search results is improved, the matching relevance of the search of the user is improved, and the search response efficiency is ensured on the basis of not increasing the calculated amount.
In addition, it should be noted that, in the sorting process of the search results, the available sorting attributes are not limited to the dimensional attributes mentioned in the above technical solutions, and the combined usage manner between the dimensional sorting attributes is not limited to the examples given in fig. 2, fig. 3, and fig. 4. The combination of the ranking attributes can be adjusted by those skilled in the art according to the requirement, for example, some attributes are added or reduced in the specific ranking process of the search results. The computation of other ranking attributes for search results may include, but is not limited to: calculating a relevance score between the description information of the search result and the query language of the user by using a bm25 algorithm, wherein the relevance score is used as a sorting attribute of the search result; calculating the similarity between the description information of the search result and the query language of the user by using an offset function as the ordering attribute of the search result; and calculating the edit distance between the description information of the search result and the query language of the user by using an edit distance calculation formula, wherein the edit distance is used as the ordering attribute of the search result, and other available text matching attributes are calculated.
Fig. 5 is a flowchart of a ranking model construction method disclosed in an embodiment of the present application, which may be applied to how to construct a ranking model that may be used for ranking search results in a retrieval or search scenario. The sequencing model construction method disclosed by the embodiment of the application can be executed by a sequencing model construction device, and the device can be realized by adopting software and/or hardware and can be integrated on any electronic equipment with computing capability.
It should be noted that the ranking model construction method disclosed in the embodiment of the present application and the aforementioned search result ranking method belong to an inventive concept, and are different in that the ranking model construction method aims at a preamble preparation or training phase of the ranking model, and the search result ranking method aims at a use phase of the ranking model. With regard to the content of how to construct at least one ranking attribute of a sample result, which is not described in detail in the embodiments of the present application, reference may be made to the explanation in the above technical solution, that is, an implementation process of constructing at least one ranking attribute of a sample result is the same as an implementation process of constructing at least one ranking attribute of a search result, except that processing objects are different, and processing logic is the same.
As shown in fig. 5, the method for constructing a ranking model disclosed in the embodiment of the present application may include:
s501, determining a sample result according to the sample query.
The sample query and the sample result may be collectively referred to as a training sample. The embodiments of the present application are not specifically implemented as to how the training samples are determined. In the practical application process, the statistics of the sample query words and the sample results can be carried out according to the requirements.
S502, constructing at least one ordering attribute of the sample result by utilizing the correlation between the sample query language and the sample result.
S503, constructing a sequencing model by using at least one sequencing attribute of the sample result.
For example, at least one ranking attribute of each sample result may be used as an input of model training, and a ranking labeling result of each sample result may be used as an output of the model training, so as to obtain a ranking model through training. The ranking marking result of the sample result can be realized by adopting relevance scores of different grades. The ranking model may employ any available ranking algorithm, including but not limited to LTR (ranking to rank) algorithm, and the like, such as LambdaMat LTR algorithm, and the like.
On the basis of the above technical solution, optionally, constructing at least one ranking attribute of the sample result by using the correlation between the sample query and the sample result, including:
Respectively determining sample query keywords of the sample query language and sample result keywords of the sample result description information;
and determining at least one sequencing attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword.
Optionally, constructing at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword, including:
and determining the ordering attribute of the sample result according to the matching result between the main query word in the sample query keyword and the main result word in the sample result keyword.
Optionally, determining the ranking attribute of the sample result according to the matching result between the main query term in the sample query keyword and the main result term in the sample result keyword includes:
determining matching words between the main result words and the main query words;
and if the matching word is determined to be the ending word of the sample result description information, recording the determination result of the ending word belonging to the matching word as the ordering attribute of the sample result.
Optionally, determining that the matching word is an end word of the sample result description information includes:
performing space replacement processing on the text of the set type in the sample result description information;
And if the text formed by adding a space after the matching word is determined is the ending text of the sample result description information, determining that the matching word is the ending word of the sample result description information.
Optionally, constructing at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword, including:
determining a first target word with the similarity to the sample query keyword being smaller than a first preset threshold value in main result words in the sample result keywords;
based on the weight of the first target word in the primary result words, the ranking attributes of the sample results are determined.
Optionally, determining, in the main result words in the sample result keywords, a first target word whose similarity to the sample query keyword is smaller than a first preset threshold includes:
and if candidate words with the similarity smaller than a first preset threshold value with the sample query keyword exist in the main result words in the sample result keywords, and a combined text formed by the user query words and the candidate words appears in the sample result description information, determining the candidate words as first target words.
Optionally, constructing at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword, including:
Determining synonyms of the main query words in the sample query keywords;
determining a second target word with the similarity to the sample result keywords being smaller than a second preset threshold in the synonyms and the main query words;
and determining the ranking attribute of the sample result based on the weight of the synonym and the weight of the second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
Optionally, determining the sample query keyword of the sample query language includes:
performing word segmentation processing on the user query language to obtain sample query keywords and weights of the sample query keywords;
and determining a main query word in the sample query keywords according to the weight of the sample query keywords and the position of the sample query keywords in the user query language.
Optionally, determining a sample result keyword of the sample result description information includes:
performing word segmentation processing on the sample result description information to obtain sample result keywords and weights of the sample result keywords;
and determining the main result words in the sample result keywords according to the weights of the sample result keywords and the times of the sample result keywords appearing in the sample result description information.
Optionally, constructing at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword, including:
If the preset word list is utilized to determine that the sample query keywords and the sample result keywords are not matched with the words belonging to the same category in the preset word list, recording the result of failed matching as the ordering attribute of the sample result;
and the preset word list comprises words for describing components or accessories of the sample result.
Optionally, the method for constructing a ranking model disclosed in the embodiment of the present application may further include:
if the sample query language is a mixed text consisting of letters and numbers, matching the sample query language with the sample result description information by using a preset regular expression comprising the sample query language;
and if the matching is successful, recording the result of successful matching as the sorting attribute of the sample result.
Optionally, constructing at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword, including:
respectively determining category attributes matched with the sample query words and the sample results by using the sample query keywords and the sample result keywords;
and determining the ordering attribute of the sample result according to the matching result of the category attributes between the sample query and the sample result and the weight of the category attributes.
Optionally, determining the ranking attribute of the sample result according to the matching result of the category attributes between the sample query and the sample result and the weight of the category attributes, including:
traversing the category attribute of the sample query language, and determining the maximum weight of the category attribute;
and if the category attributes of all levels of the sample query language respectively correspond to the category attributes of all levels of the sample result, calculating the ordering attribute of the sample result based on the determined maximum weight and the weight of the category attributes of all levels of the sample result.
Optionally, the method for constructing a ranking model disclosed in the embodiment of the present application may further include:
counting the number of target sample results of which the category attributes are matched with the set category attributes, wherein the set category attributes are set according to categories of which the similarity with the category attributes of the sample query words is greater than a third preset threshold;
and if the statistical quantity meets the attribute clearing condition, determining the ordering attribute determined according to the matching result of the category attributes between the sample query language and the sample result and the weight of the category attributes as an invalid attribute.
According to the technical scheme of the embodiment of the application, at least one sequencing attribute of the sample result is constructed by utilizing the correlation between the sample query language and the sample result and is used in the construction process of the sequencing model, wherein the constructed sequencing attribute can reflect the correlation between the sample result and the sample query language from different angles, and the construction of the sequencing attribute cannot be matched with the entity matching result of the sample query language and the sample result and is irrelevant to the word sequence matching result of the sample query language and the sample result description information, so that in the use stage after the construction of the sequencing model is completed, the sequencing effect of the search result is optimized on the basis of not increasing the calculated amount, the matching correlation of the search of a user is improved, the search response efficiency is ensured, and the problems of poor sequencing effect and large calculated amount of the search result in the existing scheme are solved.
Fig. 6 is a schematic structural diagram of a search result ranking apparatus disclosed in an embodiment of the present application, which may be applied to a case of ranking search results for recall. The search result ranking device disclosed in the embodiment of the present application may be implemented by software and/or hardware, and may be integrated on any electronic device with computing capability, such as a server.
As shown in fig. 6, the search result ranking apparatus 600 disclosed in the embodiment of the present application may include a search result determining module 601, a ranking attribute constructing module 602, and a ranking result determining module 603, where:
a search result determining module 601, configured to determine a search result according to the user query;
a ranking attribute constructing module 602, configured to construct at least one ranking attribute of the search result by using the relevance between the user query and the search result;
a ranking result determining module 603 configured to determine a ranking result of the search result based on at least one ranking attribute of the search result using the ranking model.
Optionally, the sorting attribute building module 602 includes:
a keyword determining unit for determining a query keyword of a user query and a result keyword of search result description information, respectively;
And the ranking attribute construction unit is used for constructing at least one ranking attribute of the search result according to the matching processing result of the query keyword and the result keyword.
Optionally, the sorting attribute constructing unit includes:
and the first ordering attribute determining subunit is used for determining the ordering attribute of the search result according to the matching result between the main query word in the query keyword and the main result word in the result keyword.
Optionally, the first ordering attribute determining subunit includes:
a matching word determining subunit, configured to determine a matching word between the main result word and the main query word;
and the attribute recording subunit is used for recording the determined result of the ending word of the matching word as the ordering attribute of the search result if the matching word is determined to be the ending word of the search result description information.
Optionally, the attribute recording subunit includes:
the replacing subunit is used for carrying out space replacing processing on the text of the set type in the description information of the search result;
the determining subunit is used for determining the matching word as the ending word of the search result description information if the text formed by adding a blank space after the matching word is determined is the ending text of the search result description information;
and the recording subunit is used for recording the determination result that the matching word belongs to the final word as the sequencing attribute of the search result.
Optionally, the sorting attribute constructing unit includes:
the first target word determining subunit is used for determining a first target word of which the similarity with the query keyword is smaller than a first preset threshold in the main result words in the result keywords;
and the second ordering attribute determining subunit is used for determining the ordering attribute of the search result based on the weight of the first target word in the main result words.
Optionally, the first target word determining subunit is specifically configured to:
and if candidate words with the similarity smaller than a first preset threshold value with the query keyword exist in the main result words, and a combined text formed by the query word of the user and the candidate words appears in the description information of the search results, determining the candidate words as first target words.
Optionally, the sorting attribute constructing unit includes:
a synonym determining subunit, configured to determine synonyms of the main query term in the query keyword;
the second target word determining subunit is used for determining a second target word of which the similarity with the result keyword is smaller than a second preset threshold in the synonym and the main query word;
and the third ordering attribute determining subunit is used for determining the ordering attribute of the search result based on the weight of the synonym and the weight of the second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
Optionally, the keyword determining unit includes a query keyword determining subunit and a result keyword determining subunit, where:
a query keyword determining subunit, configured to determine a query keyword of a user query;
a result keyword determination subunit operable to determine a result keyword of the search result description information;
wherein, the query keyword determination subunit includes:
the keyword and weight determining subunit is used for performing word segmentation processing on the query language of the user to obtain query keywords and weights of the query keywords;
and the main query term determining subunit is used for determining the main query terms in the query keywords according to the weights of the query keywords and the positions of the query keywords in the user query language.
Optionally, the keyword determining unit includes a query keyword determining subunit and a result keyword determining subunit, where:
a query keyword determining subunit, configured to determine a query keyword of a user query;
a result keyword determination subunit operable to determine a result keyword of the search result description information;
wherein the result keyword determination subunit includes:
the keyword and weight determining subunit is used for performing word segmentation processing on the search result description information to obtain result keywords and weights of the result keywords;
And the main result word determining subunit is used for determining the main result word in the result keywords according to the weight of the result keywords and the occurrence frequency of the result keywords in the search result description information.
Optionally, the sorting attribute constructing unit includes:
a fourth ordering attribute determining subunit, configured to, if it is determined that the query keyword and the result keyword are not matched to terms belonging to the same category in the preset vocabulary by using the preset vocabulary, record a result of failure in matching as an ordering attribute of the search result;
and the preset word list comprises words for describing components or accessories of the search results.
Optionally, the sorting attribute building module 602 further includes:
the regular matching unit is used for matching the search result description information by using a preset regular expression comprising the user query language if the user query language is a mixed text consisting of letters and numbers;
and the regular matching result recording unit is used for recording the result successfully matched as the sorting attribute of the search result if the matching is successful.
Optionally, the sorting attribute constructing unit includes:
a category attribute determining subunit, configured to determine, by using the query keyword and the result keyword, a category attribute that matches the query language of the user and the search result, respectively;
And the fifth ordering attribute determining subunit is used for determining the ordering attribute of the search result according to the matching result of the category attributes between the user query and the search result and the weight of the category attributes.
Optionally, the fifth ordering attribute determining subunit includes:
the category attribute traversal subunit is used for traversing the category attributes of the user query language and determining the maximum weight of the category attributes;
and the ranking attribute calculating subunit is used for calculating the ranking attributes of the search results based on the maximum weight and the weights of the category attributes of each level of the search results if the category attributes of each level of the user query language respectively correspond to the category attributes of each level of the search results.
Optionally, the search result ranking apparatus disclosed in the embodiment of the present application further includes:
the quantity counting module is used for counting the quantity of the target search results of which the category attributes are matched with the set category attributes, wherein the set category attributes are set according to categories of which the similarity with the category attributes of the user query language is greater than a third preset threshold;
and the attribute invalidation module is used for determining the sorting attribute determined according to the matching result of the category attributes between the user query and the search result and the weight of the category attributes as the invalid attribute if the statistical number meets the attribute clearing condition.
The search result ranking device 600 disclosed in the embodiment of the present application can execute any search result ranking method disclosed in the embodiment of the present application, and has functional modules and beneficial effects corresponding to the execution method. Reference may be made to the description in any method embodiment of the present application for details not explicitly described in the embodiments of the present application.
Fig. 7 is a schematic structural diagram of a ranking model constructing apparatus disclosed according to an embodiment of the present application, which may be applied to how to construct a ranking model that may be used for ranking search results in a retrieval or search scenario. The sequencing model building device can be implemented by software and/or hardware, and can be integrated on any electronic equipment with computing capability, such as a server and the like.
As shown in fig. 7, the ranking model constructing apparatus 700 disclosed in the embodiment of the present application may include a sample result determining module 701, a ranking attribute constructing module 702, and a ranking model constructing module 703, where:
a sample result determining module 701, configured to determine a sample result according to the sample query;
a ranking attribute constructing module 702, configured to construct at least one ranking attribute of the sample result by using the correlation between the sample query and the sample result;
A ranking model constructing module 703, configured to construct a ranking model using at least one ranking attribute of the sample result.
Optionally, the sorting attribute building module 702 includes:
the keyword determining unit is used for respectively determining sample query keywords of the sample query language and sample result keywords of the sample result description information;
and the ranking attribute construction unit is used for determining at least one ranking attribute of the sample result according to the matching processing result of the sample query keyword and the sample result keyword.
Optionally, the sorting attribute constructing unit includes:
and the first ordering attribute determining subunit is used for determining the ordering attribute of the sample result according to the matching result between the main query word in the sample query keyword and the main result word in the sample result keyword.
Optionally, the first ordering attribute determining subunit includes:
a matching word determining subunit, configured to determine a matching word between the main result word and the main query word;
and the attribute recording subunit is used for recording the determined result of the end word to which the matching word belongs as the sequencing attribute of the sample result if the matching word is determined to be the end word of the sample result description information.
Optionally, the attribute recording subunit includes:
The replacing subunit is used for performing space replacing processing on the text of the set type in the sample result description information;
the determining subunit is used for determining that the matching word is an ending word of the sample result description information if a text formed by adding a blank space after the matching word is determined is an ending text of the sample result description information;
and the recording subunit is used for recording the determination result that the matching word belongs to the final word as the sequencing attribute of the sample result.
Optionally, the sorting attribute constructing unit includes:
the first target word determining subunit is used for determining a first target word of which the similarity with the sample query keyword is smaller than a first preset threshold in the main result words in the sample result keywords;
and the second ordering attribute determining subunit is used for determining the ordering attribute of the sample result based on the weight of the first target word in the main result word.
Optionally, the first target word determining subunit is specifically configured to:
and if candidate words with the similarity smaller than a first preset threshold value with the sample query keyword exist in the main result words in the sample result keywords, and a combined text formed by the user query words and the candidate words appears in the sample result description information, determining the candidate words as first target words.
Optionally, the sorting attribute constructing unit includes:
the synonym determining subunit is used for determining synonyms of the main query words in the sample query keywords;
the second target word determining subunit is used for determining a second target word of which the similarity with the sample result keywords is smaller than a second preset threshold in the synonym and the main query word;
and the third ordering attribute determining subunit is used for determining the ordering attribute of the sample result based on the weight of the synonym and the weight of the second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
Optionally, the keyword determining unit includes a query keyword determining subunit and a result keyword determining subunit, where:
a query keyword determining subunit, configured to determine a sample query keyword of a sample query;
a result keyword determination subunit, configured to determine a sample result keyword of the sample result description information;
wherein, the query keyword determination subunit includes:
the keyword and weight determining subunit is used for performing word segmentation processing on the user query language to obtain sample query keywords and weights of the sample query keywords;
and the main query term determining subunit is used for determining the main query terms in the sample query terms according to the weights of the sample query terms and the positions of the sample query terms in the user query terms.
Optionally, the keyword determining unit includes a query keyword determining subunit and a result keyword determining subunit, where:
a query keyword determining subunit, configured to determine a sample query keyword of a sample query;
a result keyword determination subunit, configured to determine a sample result keyword of the sample result description information;
wherein the result keyword determination unit includes:
the keyword and weight determining subunit is used for performing word segmentation processing on the sample result description information to obtain a sample result keyword and the weight of the sample result keyword;
and the main result word determining subunit is used for determining the main result words in the sample result keywords according to the weights of the sample result keywords and the times of the sample result keywords appearing in the sample result description information.
Optionally, the sorting attribute constructing unit includes:
a fourth ordering attribute determining subunit, configured to, if it is determined that the sample query keyword and the sample result keyword do not match a word belonging to the same category in the preset word list by using the preset word list, record a result of failure in matching as an ordering attribute of the sample result;
and the preset word list comprises words for describing components or accessories of the sample result.
Optionally, the sorting attribute constructing module 702 further includes:
the regular matching unit is used for matching the sample result description information by utilizing a preset regular expression comprising the sample query language if the sample query language is a mixed text consisting of letters and numbers;
and the regular matching result recording unit is used for recording the result which is successfully matched as the sorting attribute of the sample result if the matching is successful.
Optionally, the sorting attribute constructing unit includes:
the category attribute determining subunit is used for determining category attributes which are respectively matched with the sample query words and the sample results by respectively utilizing the sample query keywords and the sample result keywords;
and the fifth ordering attribute determining subunit is used for determining the ordering attribute of the sample result according to the matching result of the category attributes between the sample query language and the sample result and the weight of the category attributes.
Optionally, the fifth ordering attribute determining subunit includes:
the category attribute traversal subunit is used for traversing the category attributes of the sample query language and determining the maximum weight of the category attributes;
and the ordering attribute calculating subunit is used for calculating the ordering attribute of the sample result based on the determined maximum weight and the weight of each level of category attribute of the sample result if each level of category attribute of the sample query language respectively corresponds to each level of category attribute of the sample result.
Optionally, the ranking model constructing apparatus disclosed in the embodiment of the present application may further include:
the quantity counting module is used for counting the quantity of target sample results of which the category attributes are matched with the set category attributes, wherein the set category attributes are set according to categories of which the similarity with the category attributes of the sample query words is greater than a third preset threshold;
and the attribute invalidation module is used for determining the sorting attribute determined according to the matching result of the category attributes between the sample query language and the sample result and the weight of the category attributes as the invalid attribute if the statistical quantity meets the attribute clearing condition.
The ranking model constructing apparatus 700 disclosed in the embodiment of the present application can execute any ranking model constructing method disclosed in the embodiment of the present application, and has functional modules and beneficial effects corresponding to the execution method. Reference may be made to the description in any method embodiment of the present application for details not explicitly described in the embodiments of the present application.
According to an embodiment of the present application, an electronic device and a readable storage medium are also provided.
As shown in fig. 8, fig. 8 is a block diagram of an electronic device for implementing a search result ranking method or a ranking model building method in the embodiment of the present application. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of embodiments of the present application described and/or claimed herein.
As shown in fig. 8, the electronic apparatus includes: one or more processors 801, memory 802, and interfaces for connecting the various components, including a high speed interface and a low speed interface. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions for execution within the electronic device, including instructions stored in or on the memory to display Graphical information for a Graphical User Interface (GUI) on an external input/output device, such as a display device coupled to the Interface. In other embodiments, multiple processors and/or multiple buses may be used, along with multiple memories and multiple memories, as desired. Also, multiple electronic devices may be connected, with each device providing portions of the necessary operations, e.g., as a server array, a group of blade servers, or a multi-processor system. Fig. 8 illustrates an example of a processor 801.
The memory 802 is a non-transitory computer readable storage medium provided by the embodiments of the present application. The memory stores instructions executable by at least one processor, so that the at least one processor executes the search result ranking method or the ranking model construction method provided by the embodiment of the application. The non-transitory computer-readable storage medium of the embodiments of the present application stores computer instructions for causing a computer to execute the search result ranking method or the ranking model building method provided by the embodiments of the present application.
The memory 802 is a non-transitory computer readable storage medium, and can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the search result ranking method or the ranking model building method in the embodiments of the present application, for example, the search result determining module 601, the ranking attribute building module 602, and the ranking result determining module 603 shown in fig. 6, or for example, the sample result determining module 701, the ranking attribute building module 702, and the ranking model building module 703 shown in fig. 7. The processor 801 executes various functional applications and data processing of the electronic device by running non-transitory software programs, instructions, and modules stored in the memory 802, that is, implements the search result ranking method or the ranking model building method in the above-described method embodiments.
The memory 802 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the electronic device, and the like. Further, the memory 802 may include high speed random access memory and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 802 optionally includes a memory remotely disposed from the processor 801, and the remote memory may be connected to an electronic device for implementing a search result ranking method or a ranking model building method in the embodiments of the present application via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The electronic device for implementing the search result ranking method or the ranking model building method in the embodiment of the present application may further include: an input device 803 and an output device 804. The processor 801, the memory 802, the input device 803, and the output device 804 may be connected by a bus or other means, and are exemplified by a bus in fig. 8.
The input device 803 may receive input numeric or character information and generate key signal inputs related to user settings and function controls of an electronic apparatus for implementing the search result ranking method or the ranking model building method in the embodiments of the present application, such as an input device of a touch screen, a keypad, a mouse, a track pad, a touch pad, a pointing stick, one or more mouse buttons, a track ball, a joystick, or the like. The output device 804 may include a display apparatus, an auxiliary lighting device such as a Light Emitting Diode (LED), a tactile feedback device, and the like; the tactile feedback device is, for example, a vibration motor or the like. The Display device may include, but is not limited to, a Liquid Crystal Display (LCD), an LED Display, and a plasma Display. In some implementations, the display device can be a touch screen.
Various implementations of the systems and techniques described here can be realized in digital electronic circuitry, Integrated circuitry, Application Specific Integrated Circuits (ASICs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
These computer programs, also known as programs, software applications, or code, include machine instructions for a programmable processor, and may be implemented using high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or Device for providing machine instructions and/or data to a Programmable processor, such as a magnetic disk, optical disk, memory, Programmable Logic Device (PLD), including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device for displaying information to a user, for example, a Cathode Ray Tube (CRT) or an LCD monitor; and a keyboard and a pointing device, such as a mouse or a trackball, by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here, or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), the internet, and blockchain networks.
The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
According to the technical scheme of the embodiment of the application, at least one ordering attribute of the search result is constructed by utilizing the correlation between the user query and the search result, then the constructed ordering attribute is used as the input of a pre-trained ordering model, and the ordering result of the search result is obtained through model processing, wherein the constructed ordering attribute can reflect the correlation between the search result and the user query from different angles, so that the ordering effect of the search result is improved on the basis of not increasing the calculated amount, the matching correlation of the search of the user is improved, and the search response efficiency is ensured.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present application can be achieved, and the present invention is not limited herein.
The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims (20)

1. A method for ranking search results, comprising:
determining a search result according to the user query;
constructing at least one ordering attribute of the search result by utilizing the relevance of the user query language and the search result;
determining, using a ranking model, a ranking result for the search result based on at least one ranking attribute of the search result.
2. The method of claim 1, wherein constructing at least one ranking attribute of the search result using the relevance of the user query to the search result comprises:
respectively determining query keywords of the user query and result keywords of the search result description information;
and constructing at least one ranking attribute of the search result according to the matching processing result of the query keyword and the result keyword.
3. The method of claim 2, wherein constructing at least one ranking attribute of the search results from the matching process of the query keyword and the result keyword comprises:
and determining the ranking attribute of the search result according to the matching result between the main query word in the query keyword and the main result word in the result keyword.
4. The method of claim 3, wherein determining the ranking attributes of the search results according to the matching results between the main query term in the query keyword and the main result term in the result keyword comprises:
determining a matching term between the primary result term and the primary query term;
and if the matching word is determined to be the ending word of the search result description information, recording the determination result that the matching word belongs to the ending word as the ordering attribute of the search result.
5. The method of claim 4, wherein determining that the matching word is an end word of the search result description information comprises:
performing space replacement processing on the text of the set type in the search result description information;
And if the text formed by adding a space after the matching word is determined is the ending text of the search result description information, determining that the matching word is the ending word of the search result description information.
6. The method of claim 2, wherein constructing at least one ranking attribute of the search results from the matching process of the query keyword and the result keyword comprises:
determining a first target word with the similarity to the query keyword being smaller than a first preset threshold value in main result words in the result keywords;
determining a ranking attribute of the search result based on a weight of a first target word in the primary result words.
7. The method according to claim 6, wherein determining, among main result words in the result keywords, a first target word having a similarity smaller than a first preset threshold with the query keyword comprises:
and if a candidate word with the similarity to the query keyword being smaller than the first preset threshold exists in the main result word and a combined text formed by the user query word and the candidate word appears in the search result description information, determining the candidate word as the first target word.
8. The method of claim 2, wherein constructing at least one ranking attribute of the search results from the matching process of the query keyword and the result keyword comprises:
determining synonyms of main query words in the query keywords;
determining a second target word with the similarity to the result keyword being smaller than a second preset threshold in the synonym and the main query word;
and determining the ranking attribute of the search result based on the weight of the synonym and the weight of a second target word in the main query word, wherein the weight of the synonym is the same as the weight of the corresponding original word.
9. The method of any of claims 2-8, wherein determining query terms for the user query comprises:
performing word segmentation processing on the user query language to obtain the query keyword and the weight of the query keyword;
and determining a main query word in the query keywords according to the weight of the query keywords and the position of the query keywords in the user query language.
10. The method of any of claims 2-8, wherein determining the result keyword of the search result description information comprises:
Performing word segmentation processing on the search result description information to obtain the result keywords and the weight of the result keywords;
and determining main result words in the result keywords according to the weights of the result keywords and the times of the result keywords appearing in the search result description information.
11. The method of claim 2, wherein constructing at least one ranking attribute of the search results from the matching process of the query keyword and the result keyword comprises:
if the preset word list is utilized to determine that the query keyword and the result keyword are not matched with the words belonging to the same category in the preset word list, recording the result of failed matching as the ordering attribute of the search result;
and the preset word list comprises words for describing components or accessories of the search results.
12. The method of claim 2, wherein constructing at least one ranking attribute of the search result using relevance of the user query and the search result further comprises:
if the user query language is a mixed text consisting of letters and numbers, matching the search result description information by using a preset regular expression comprising the user query language;
And if the matching is successful, recording the result successfully matched as the ranking attribute of the search result.
13. The method of claim 2, wherein constructing at least one ranking attribute of the search results from the matching process of the query keyword and the result keyword comprises:
respectively utilizing the query key words and the result key words to determine category attributes matched with the query words of the user and the search results respectively;
and determining the ordering attribute of the search result according to the matching result of the category attributes between the user query and the search result and the weight of the category attributes.
14. The method of claim 13, wherein determining the ranking attributes of the search results according to the matching results of the category attributes between the user query and the search results and the weights of the category attributes comprises:
traversing the category attribute of the user query language, and determining the maximum weight of the category attribute;
and if the category attributes of all levels of the user query language respectively correspond to the category attributes of all levels of the search result, calculating the ordering attribute of the search result based on the maximum weight and the weight of the category attributes of all levels of the search result.
15. The method of claim 13, further comprising:
counting the number of target search results of which the category attributes are matched with set category attributes, wherein the set category attributes are set according to categories of which the similarity with the category attributes of the user query language is greater than a third preset threshold;
and if the statistical quantity meets the attribute clearing condition, determining the ranking attribute determined according to the matching result of the category attributes between the user query language and the search result and the weight of the category attributes as an invalid attribute.
16. A method for constructing a ranking model, comprising:
determining a sample result according to the sample query;
constructing at least one ordering attribute of the sample result by utilizing the correlation between the sample query and the sample result;
and constructing a ranking model by using at least one ranking attribute of the sample result.
17. A search result ranking apparatus, comprising:
the search result determining module is used for determining a search result according to the user query;
the ranking attribute construction module is used for constructing at least one ranking attribute of the search result by utilizing the correlation between the user query language and the search result;
A ranking result determination module to determine a ranking result of the search result based on at least one ranking attribute of the search result using a ranking model.
18. An order model construction apparatus, comprising:
the sample result determining module is used for determining a sample result according to the sample query;
the ranking attribute construction module is used for constructing at least one ranking attribute of the sample result by utilizing the correlation between the sample query language and the sample result;
and the sequencing model building module is used for building a sequencing model by utilizing at least one sequencing attribute of the sample result.
19. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a search result ranking method according to any one of claims 1-15 or to perform a ranking model construction method according to claim 16.
20. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the search result ranking method of any one of claims 1-15 or the ranking model building method of claim 16.
CN202010287348.9A 2020-04-13 2020-04-13 Search result ordering method, model building method, device, equipment and medium Active CN113536156B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010287348.9A CN113536156B (en) 2020-04-13 2020-04-13 Search result ordering method, model building method, device, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010287348.9A CN113536156B (en) 2020-04-13 2020-04-13 Search result ordering method, model building method, device, equipment and medium

Publications (2)

Publication Number Publication Date
CN113536156A true CN113536156A (en) 2021-10-22
CN113536156B CN113536156B (en) 2024-05-28

Family

ID=78087955

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010287348.9A Active CN113536156B (en) 2020-04-13 2020-04-13 Search result ordering method, model building method, device, equipment and medium

Country Status (1)

Country Link
CN (1) CN113536156B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114491318A (en) * 2021-12-16 2022-05-13 北京百度网讯科技有限公司 Method, device and equipment for determining target information and storage medium
CN114548100A (en) * 2022-03-01 2022-05-27 深圳市医未医疗科技有限公司 Clinical scientific research auxiliary method and system based on big data technology

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110218989A1 (en) * 2009-09-23 2011-09-08 Alibaba Group Holding Limited Information Search Method and System
CN102591876A (en) * 2011-01-14 2012-07-18 阿里巴巴集团控股有限公司 Sequencing method and device of search results
CN103514178A (en) * 2012-06-18 2014-01-15 阿里巴巴集团控股有限公司 Searching and sorting method and device based on click rate
CN104077306A (en) * 2013-03-28 2014-10-01 阿里巴巴集团控股有限公司 Search engine result sequencing method and search engine result sequencing system
CN104239440A (en) * 2014-09-01 2014-12-24 百度在线网络技术（北京）有限公司 Method and device for representing search results
CN104778201A (en) * 2015-01-23 2015-07-15 湖南科技大学 Multi-query result combination-based prior art retrieval method
CN104899322A (en) * 2015-06-18 2015-09-09 百度在线网络技术（北京）有限公司 Search engine and implementation method thereof
CN105488068A (en) * 2014-09-19 2016-04-13 阿里巴巴集团控股有限公司 Methods and apparatuses for searching music and establishing index, and search result judgment method
CN106446122A (en) * 2016-09-19 2017-02-22 华为技术有限公司 Information retrieval method and device and computation device
CN106933947A (en) * 2017-01-20 2017-07-07 北京三快在线科技有限公司 A kind of searching method and device, electronic equipment
CN107832432A (en) * 2017-11-15 2018-03-23 北京百度网讯科技有限公司 A kind of search result ordering method, device, server and storage medium
US20180181573A1 (en) * 2016-12-27 2018-06-28 Beijing Baidu Netcom Science And Technology Co., Ltd. Search method and device for asking type query based on deep question and answer
CN110162593A (en) * 2018-11-29 2019-08-23 腾讯科技（深圳）有限公司 A kind of processing of search result, similarity model training method and device

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110218989A1 (en) * 2009-09-23 2011-09-08 Alibaba Group Holding Limited Information Search Method and System
CN102591876A (en) * 2011-01-14 2012-07-18 阿里巴巴集团控股有限公司 Sequencing method and device of search results
CN103514178A (en) * 2012-06-18 2014-01-15 阿里巴巴集团控股有限公司 Searching and sorting method and device based on click rate
CN104077306A (en) * 2013-03-28 2014-10-01 阿里巴巴集团控股有限公司 Search engine result sequencing method and search engine result sequencing system
CN104239440A (en) * 2014-09-01 2014-12-24 百度在线网络技术（北京）有限公司 Method and device for representing search results
CN105488068A (en) * 2014-09-19 2016-04-13 阿里巴巴集团控股有限公司 Methods and apparatuses for searching music and establishing index, and search result judgment method
CN104778201A (en) * 2015-01-23 2015-07-15 湖南科技大学 Multi-query result combination-based prior art retrieval method
CN104899322A (en) * 2015-06-18 2015-09-09 百度在线网络技术（北京）有限公司 Search engine and implementation method thereof
CN106446122A (en) * 2016-09-19 2017-02-22 华为技术有限公司 Information retrieval method and device and computation device
US20180181573A1 (en) * 2016-12-27 2018-06-28 Beijing Baidu Netcom Science And Technology Co., Ltd. Search method and device for asking type query based on deep question and answer
CN106933947A (en) * 2017-01-20 2017-07-07 北京三快在线科技有限公司 A kind of searching method and device, electronic equipment
CN107832432A (en) * 2017-11-15 2018-03-23 北京百度网讯科技有限公司 A kind of search result ordering method, device, server and storage medium
CN110162593A (en) * 2018-11-29 2019-08-23 腾讯科技（深圳）有限公司 A kind of processing of search result, similarity model training method and device

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114491318A (en) * 2021-12-16 2022-05-13 北京百度网讯科技有限公司 Method, device and equipment for determining target information and storage medium
CN114491318B (en) * 2021-12-16 2023-09-01 北京百度网讯科技有限公司 Determination method, device, equipment and storage medium of target information
CN114548100A (en) * 2022-03-01 2022-05-27 深圳市医未医疗科技有限公司 Clinical scientific research auxiliary method and system based on big data technology

Also Published As

Publication number Publication date
CN113536156B (en) 2024-05-28

Similar Documents

Publication Publication Date Title
US11397772B2 (en) Information search method, apparatus, and system
CN109815308B (en) Method and device for determining intention recognition model and method and device for searching intention recognition
WO2020108608A1 (en) Search result processing method, device, terminal, electronic device, and storage medium
TWI557664B (en) Product information publishing method and device
US8078617B1 (en) Model based ad targeting
US9589277B2 (en) Search service advertisement selection
JP2021166109A (en) Fusion sorting model training method and device, search sorting method and device, electronic device, storage medium, and program
CN101241512A (en) Search method for redefining enquiry word and device therefor
CN111949814A (en) Searching method, searching device, electronic equipment and storage medium
US20110184946A1 (en) Applying synonyms to unify text search with faceted browsing classification
CN112087371A (en) Instant messaging group searching method, device, equipment and storage medium
CN109241403A (en) Item recommendation method, device, machinery equipment and computer readable storage medium
CN105512180A (en) Search recommendation method and device
CN111831821A (en) Training sample generation method and device of text classification model and electronic equipment
CN111475725A (en) Method, apparatus, device, and computer-readable storage medium for searching for content
CN113536156B (en) Search result ordering method, model building method, device, equipment and medium
CN112182145A (en) Text similarity determination method, device, equipment and storage medium
US8538946B1 (en) Creating model or list to identify queries
CN112182348A (en) Semantic matching judgment method and device, electronic equipment and computer readable medium
CN115248839A (en) Knowledge system-based long text retrieval method and device
CN114090877A (en) Position information recommendation method and device, electronic equipment and storage medium
CN112052397B (en) User characteristic generation method and device, electronic equipment and storage medium
CN116861060A (en) Private domain electronic commerce data searching method, device, equipment and storage medium
CN116383340A (en) Information searching method, device, electronic equipment and storage medium
CN114491232B (en) Information query method and device, electronic 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