Text search result sorting method and device, electronic equipment, readable medium and program product

Abstract

The invention discloses a text search result sorting method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of data retrieval, and the text search result sorting method comprises the steps that when a search statement is received, pre-configured intervention parameters are read; according to the search statement and the pre-configured intervention parameters, determining a plurality of search results corresponding to the search statement, and a full matching value, a partial matching value and a release time characteristic value corresponding to each search result; according to the title of each search result and the content of the main body, determining the matching condition of each search result and the search statement; sorting the search results according to the matching conditions of the search results, the full matching values, the partial matching values and the release time characteristic values corresponding to the search results, and obtaining and returning the sorted search results. According to the technical scheme, the problem that search results of an existing text search scheme are poorer in sorting reasonability is solved.

Description

Technical Field

[0001] The present application relates to the field of data retrieval technology, and in particular to a method, device, electronic device, computer-readable storage medium, and computer program product for sorting text search results. Background Art

[0002] At present, the commonly used text search ranking method is generally for Internet documents, mainly by comprehensively scoring the relevance of the search keywords in the searched files, and then sorting the searched document set according to the score. Among them, the factors affecting the scoring include the search keyword frequency, the length of each document, and the coverage of the search keyword in all documents.

[0003] However, in this text search solution, since the weight of the search keywords is unclear, it is easy to cause the weight of the full hit parameter of the title and text to be too low. When the user searches, if the title and text do not match completely, the search results will be ranked too high if all other fields match completely. The ranking of the text search results is not reasonable, and the user needs to query again manually, which reduces the search efficiency.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of the present application is to provide a text search result sorting method, device, electronic device, computer-readable storage medium and computer program product, aiming to solve the technical problem that the search result sorting rationality of the current text search scheme is poor.

[0006] To achieve the above purpose, the present application proposes a text search result sorting method, which includes:

[0007] When receiving a search statement, reading preconfigured intervention parameters, wherein the preconfigured intervention parameters at least include a full match value, a partial match value, and a release time feature value;

[0008] Determine, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results;

[0009] Determining the matching status of each search result with the search statement according to the title and text content of each search result;

[0010] According to the matching conditions of the search results, the full matching values, partial matching values ​​and release time feature values ​​respectively corresponding to the search results, the search results are sorted to obtain and return the sorted search results.

[0011] In one embodiment, the step of determining a plurality of search results corresponding to the search statement according to the search statement and the preconfigured intervention parameter comprises:

[0012] Input each stock document into the preset entity recognition model to generate the corresponding entity dictionary;

[0013] Determining whether the search statement matches the entity dictionary;

[0014] If there is a hit, searching the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result in each of the existing documents;

[0015] If there is no hit, it is determined whether the search statement has all hits in the existing documents, and if all hits are found, the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result are returned.

[0016] In one embodiment, before the step of reading the preconfigured intervention parameters, the method further comprises:

[0017] Inputting the search statement into a preset entity recognition model to obtain at least one entity corresponding to the search statement;

[0018] An updated search statement is generated according to the association relationship between the entities, wherein the updated search statement is used for the search service to determine the corresponding search result.

[0019] In one embodiment, before the step of determining, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results, the method includes:

[0020] Filtering sensitive words in the search sentence, and correcting the filtered search sentence;

[0021] Segment the corrected search sentence to obtain at least one search term;

[0022] Removing stop words from each of the search terms according to a preset stop word list, and adding synonyms based on the search terms after the stop words are removed and a preset synonym table to obtain updated search terms;

[0023] The updated search term is modified into a preset standard grammatical form so that the search result corresponding to the updated search term is returned through the search service.

[0024] In one embodiment, the step of sorting the search results according to the matching conditions of the search results, the full match values, the partial match values ​​and the release time feature values ​​respectively corresponding to the search results, and obtaining and returning the sorted search results includes:

[0025] Determine multiple features corresponding to each search result and feature values ​​corresponding to each feature according to the matching conditions of each search result, the full matching value, the partial matching value and the release time feature value corresponding to each search result, wherein the matching conditions include at least one of title matching and text matching;

[0026] Calculate the feature score of each search result according to the feature values ​​and weights corresponding to each of the features, wherein the weight of the full match value is greater than the weight of the partial match value, and the weight of the title match is greater than the weight of the text match;

[0027] Based on the feature scores of the search results, the search results are sorted in descending order, and the sorted search results are obtained and returned.

[0028] In one embodiment, before the step of sorting the search results in descending order based on the feature scores of the search results, the method further includes:

[0029] Filter search results whose feature scores are below a preset score threshold;

[0030] Filter out search results whose scores of each of the features are lower than a preset score threshold;

[0031] Based on a preset custom score dictionary, the feature scores of search results of the type corresponding to the custom score dictionary are updated to obtain updated feature scores.

[0032] In addition, the present application also proposes a text search result sorting device, the text search result sorting device comprising:

[0033] A parameter reading module, configured to read pre-configured intervention parameters when receiving a search statement, wherein the pre-configured intervention parameters at least include a full match value, a partial match value, and a release time feature value;

[0034] A statement search module, used to determine, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value and a release time feature value respectively corresponding to each of the search results;

[0035] A sentence matching module, used to determine the matching status of each search result with the search sentence according to the title and text content of each search result;

[0036] The result sorting module is used to sort the search results according to the matching conditions of the search results, the full match values, partial match values ​​and release time feature values ​​corresponding to the search results, and obtain and return the sorted search results.

[0037] In addition, to achieve the above objectives, the present application also proposes an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the text search result sorting method as described above.

[0038] In addition, to achieve the above objectives, the present application also proposes a storage medium, which is a computer-readable storage medium, and stores a computer program on the storage medium. When the computer program is executed by a processor, the steps of the text search result sorting method described above are implemented.

[0039] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the text search result sorting method as described above are implemented.

[0040] The present application proposes a method for sorting text search results. First, when a search statement is received, a pre-configured intervention parameter is read, wherein the pre-configured intervention parameter includes at least a full match value, a partial match value, and a release time characteristic value; according to the search statement and the pre-configured intervention parameter, multiple search results corresponding to the search statement and the full match value, partial match value, and release time characteristic value corresponding to each search result are determined; according to the title and the content of the body of each search result, the matching of each search result with the search statement is determined; according to the matching of each search result, the full match value, partial match value, and release time characteristic value corresponding to each search result are sorted, and the sorted search results are obtained and returned. The present application is equivalent to proposing a more comprehensive search sorting strategy than the traditional text search sorting scheme, which takes into account different factors affecting text sorting, including full match, partial match, and release time, and also sorts the search results by the matching of the title and body of the search results, so that the matching of the title and body, which has a greater impact on the text relevance, participates in the search result sorting, making the output sorted search results more reasonable and scientific, improving the efficiency of text search, and allowing users to more efficiently query the search results they need. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0043] Figure 1 This is a flow chart of a method for sorting text search results in an embodiment of the present application;

[0044] Figure 2 A schematic diagram of a feasible progressive search process in an embodiment of the present application;

[0045] Figure 3 A schematic diagram of a feasible prediction process of search intent in an embodiment of the present application;

[0046] Figure 4 This is a schematic diagram of a search sorting rule in an embodiment of the present application;

[0047] Figure 5 A schematic diagram of the full steps of text search and result sorting is provided for an embodiment of the present application;

[0048] Figure 6 A schematic diagram of a test interface for a feasible text search result sorting method in an embodiment of the present application;

[0049] Figure 7 This is a schematic diagram of the structure of a text search result sorting device in an embodiment of the present application;

[0050] Figure 8 Schematic diagram of the device structure of the hardware operating environment involved in the text search result sorting method in the embodiment of the present application.

[0051] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0053] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0054] Since the existing search ranking and recall schemes are all for Internet text information, they do not involve official documents with attachments in the intranet, and do not consider keyword matching of attachments. In addition, since the weight of search keywords is unclear and the situation of partial keyword matching is not considered, the weight of the title and text full hit parameters is too low. When the title and text are not fully matched during the user search, the search results are ranked too high when all other fields are matched, and even ranked before the search results with full title and text matching, requiring users to manually query again, reducing search efficiency. In addition, since synonyms are not taken seriously in the search, the search results with full synonym matches cannot be ranked first, which can easily lead to missed search results or inaccurate ranking.

[0055] In order to overcome the defects existing in the aforementioned scheme, the embodiment of the present application proposes a comprehensive search sorting and recycling strategy implementation method. Specifically, when sorting the search results, a complex analysis of the full or partial matching of the relevance of different text types (i.e., title, body, abstract, attachment, etc.) is considered, and the search results are sorted by calculating the combination of full hit features, word segmentation hit features, search term hit ratio features, search term hit ratio features in the field, and time feature weights. In addition, synonyms are added, and sorting elements such as sensitive words and stop words are filtered, so as to more accurately obtain the user's search target, and output a more reasonable search ranking, thereby improving the search accuracy and efficiency.

[0056] The present application embodiment provides a method for sorting text search results. Figure 1 This is a flow chart of the first embodiment of the text search result sorting method of the present application, the text search result sorting method comprises:

[0057] Step S10, when a search statement is received, read pre-configured intervention parameters, wherein the pre-configured intervention parameters at least include a full match value, a partial match value, and a release time feature value;

[0058] First, the text search result sorting method proposed in the embodiment of the present application can be applied to a text search platform. Specifically, a user can submit a search statement on the platform to search for the search results (in text form) that he wants to query on the platform. Specifically, a user can query various types of text through a search statement, such as looking for personnel information, application information, official documents, completed work, news, or announcements.

[0059] In addition, after the text search platform finishes processing the search statement entered by the user, it can call the pre-configured search interface to request the search service, transmit the search statement to the search service, and perform the steps of Gateway interface authentication and forwarding request. In addition, it is also necessary to read the Json (javascript object notation, a lightweight data exchange format) intervention parameters pre-configured on the platform, where the intervention parameters are used to indicate what types of parameters the search service needs to return during this search.

[0060] Exemplarily, the intervention Json is mainly used to configure the personalized features corresponding to each scenario. For example, personalized features: store the full hit match value, partial match value and feature value of the article release time. Among them, full match is equivalent to the keywords in the search statement existing and continuous in the search results, partial match is equivalent to the keywords in the search statement partially matching successfully or discontinuously in the search results, and the release time feature value refers to the release time of the text corresponding to the search result. It can be understood that, when other matching conditions are the same, the closer the release time is to the current time, the higher the corresponding arrangement order can be.

[0061] Step S20, determining multiple search results corresponding to the search statement and full match values, partial match values, and release time feature values ​​corresponding to each search result, according to the search statement and the pre-configured intervention parameters;

[0062] After receiving the search statement and pre-configured intervention parameters, an http post request (sending data to the server and requesting the server to process the data) can be initiated to the algorithm service, where the Json content is included in the post request.

[0063] Exemplarily, after obtaining the operation instructions of clicking search and pressing Enter, after reading the intervention Json configured by the operation platform, a request is directly initiated to the algorithm service. In addition, after the algorithm service receives the aforementioned search statement and pre-configured intervention parameters, it can call the corresponding interface corresponding to the search statement to implement text search. It should be noted that for different search intents, the interface that needs to be called is also different. This is because the databases stored for different types of text are also different. For example, personnel information, official documents, news, and announcements are stored in different databases. The search service can call the interfaces corresponding to various databases to search the text content stored therein and obtain the corresponding search results. At this time, there is no order between the search results.

[0064] Step S30, determining the matching status of each search result with the search statement according to the title and text content of each search result;

[0065] Step S40, sorting the search results according to the matching conditions of the search results, the full matching values, partial matching values ​​and publishing time feature values ​​corresponding to the search results, and obtaining and returning the sorted search results.

[0066] In order to reasonably sort the search results, the sorting can be performed by calculating the feature score of each search result, wherein the matching situation between each search result and the search statement can be reflected in a numerical situation, for example, the title match can be recorded as 1, and the non-match can be recorded as 0, and the text match is also recorded, and the full match value and the partial match value can be recorded in a similar manner. In addition, the above various features can be assigned different weights respectively, and the weight of the matching situation with a greater impact can be larger, for example, the weight of the matching situation of the title and the search statement can be greater than the weight corresponding to the matching situation of the text and the search statement. Exemplarily, the weight of each feature can be obtained by calculating the product of the feature value of each feature and the weight, and then the weight of each feature corresponding to each search result is summed to obtain the feature score of each search result, and the search results are sorted according to the size of the feature score. After the search results are sorted, the page of the search platform is updated, and the sorted search results are returned and displayed. For example, after the search is completed (or after searching and displaying part of the search results), the page is updated from the page including the input box and the search statement to the page displaying the search results.

[0067] In another feasible embodiment, in addition to the matching of the title and body of each search result with the search statement, the matching of the abstract, attachment, etc. of each search result can also be combined to sort the search results more comprehensively. Among them, since the matching of the abstract, attachment, etc. has a smaller impact on the relevance of the text, its corresponding weight will also be smaller than the weight of the title and body.

[0068] The text search result sorting method of the embodiment of the present application is mainly to realize a complex analysis that comprehensively considers the relevance of different text types (i.e., title, body, abstract, attachment, etc.) when sorting the search results, and sorts the search by calculating the combination of full hit features, word segmentation hit features, search term hit ratio features, search term hit ratio features in the field, and time feature weights, so that the final sorting is completed and the output search results can display more important and more user-friendly search results in a more forward position, allowing users to search and obtain the text content they want faster and more efficiently, thereby improving the accuracy and efficiency of text search.

[0069] Further, in a feasible embodiment, the step of determining multiple search results corresponding to the search statement and the full match value, partial match value and release time feature value respectively corresponding to each of the search results includes:

[0070] Step S21, input each stock document into a preset entity recognition model to generate a corresponding entity dictionary;

[0071] It should be noted that the embodiment of the present application also provides a gradual search method, referring to Figure 2 , where the main functions of the entity recognition model include: through the entities of the identified search content, it can be determined whether the entities in the search statement currently input by the user include the entities included in the search content itself. Specifically, the entity dictionary can store the entities identified in all existing documents as an entity dictionary (hash (hash value) or MD5 (password hash) form) for judgment use. Among them, different types of existing documents (such as official documents, news, address books, personnel information, etc.) correspond to different dictionaries. After determining the search intent based on the user's search statement, the corresponding entity dictionary is called to determine whether the judgment statement hits the entity dictionary.

[0072] Step S22, determining whether the search statement hits the entity dictionary;

[0073] Step S23: If a match is found, search the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result in each existing document;

[0074] Step S24, if no hit is found, determine whether the search statement has all hits in the existing documents, and if all hits are found, return the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result.

[0075] Reference Figure 2 In Redis (remote dictionary service), the highest priority judgment is made on the input search statement through the entity recognition model. The judgment process includes: when the entity in the search statement hits the entity in the entity dictionary, the search is performed directly; if there is no hit in the entity dictionary, a part-of-speech judgment based on HanLp (Han Language Processing, Chinese language processing package) is performed to determine whether the part-of-speech is one of nz / n / s (proper noun / noun / place name), and further judge whether it is a full hit judgment of a certain field through weight to determine whether to search.

[0076] It should be noted that the highest priority is to directly call the search when the input word in the search box is in the entity dictionary. If the return result is 0, keep the current interface unchanged. If the return result is not 0, directly return the current search results and update them to the browser page. In addition, the part of speech is used to judge the input word in the current search box, including: full hit refers to whether the entity in the current search box has continuous words that are fully hit with a certain article in the current scene (for example, searching for "travel expenses" does not count as a full hit if it only contains the keyword "travel", and must contain the three words "travel expenses", and there is no text separating the three words), and only when the input word in the search box is fully hit, will the data display be directly returned.

[0077] Specifically, in the process of part-of-speech judgment, the weight design corresponds to only multiple sets of feature weight parameters for each field, including: part_of_doc (document part), char_query_match_ratio (full match weight) and query_match_ratio (partial match weight), etc., and the weight values ​​are different. For example, if there are n fields, the validity of the weight judgment is 1 to n.

[0078] In another feasible embodiment, the entity recognition model needs to be pre-trained. During the training process, sequence labeling is performed based on the current data set. The steps include: using text classification technology to perform algorithm tuning on data in a specific field to generate a model, thereby realizing the prediction function of search intent through the model, such as Figure 3 As shown, it includes: first inputting query (search statement), processing empty characters, using intent recognition based on rule dictionary (rule_keyword_arr, rule_keyword_array), using intent recognition based on Bert (a language representation model) model (bert_predict_arr, model_prediction_array), and then fusing the prediction results of the two, sorting (sorting) to get final_intent_predict_arr (final_intent_prediction_array), and further performing entity recognition-NER (Named Entity Recognition, named entity recognition), for example, whether to judge intent = (official document or address book), if not, determine that final_combined_predict_arr (final_combined_prediction_array) only contains intent recognition results; if Hayes, further determine whether the entity type = (time, name, predefined word, synonym), if so, based on the transcription of rules and dictionaries, final_combined_predict_arr contains intent recognition, entity recognition and / or transcription; if not, determine whether the entity type = pinyin, if so, determine based on the transcription of the baichuan (Baichuan) large model.

[0079] In addition, in the entity recognition model, three Chinese transformers (converter) pre-trained models are used for token classification (natural language processing tasks), namely albert-base-chinese (a Chinese pre-trained language model), albert-tiny-chinese (a Chinese pre-trained language model), and bert-tiny-chinese (a Chinese pre-trained language model). At the same time, the bert-base-chinese (a Chinese pre-trained speech model) model is used to correctly use the Chinese tokenizer (word segmenter). When searching, the user's main intention is recognized through the input search sentence, and after the entity is extracted, the entity is searched to achieve the purpose of accurately locating the user's search intention.

[0080] In a feasible embodiment, before the step of reading the preconfigured intervention parameters, the method further includes:

[0081] Step A10, inputting the search statement into a preset entity recognition model to obtain at least one entity corresponding to the search statement;

[0082] Step A20: generating an updated search statement according to the association relationship between the entities, wherein the updated search statement is used for the search service to determine the corresponding search result.

[0083] Specifically, through the entity recognition model, the entity recognition corresponding to the search term is trained, and only the entity term is searched during the search to achieve the purpose of identifying the user's main intention. Entity recognition can be divided into two main aspects: one is to determine which parts of the text input by the user constitute the entity, that is, the boundary recognition of the entity; the other is to determine the type of these entities, such as names of people, places, etc. It can be understood that these two tasks are usually combined to achieve accurate entity recognition. After the entity is extracted, the relationship in the input text is matched through a series of defined patterns, and the relationship between entities is identified by mining the frequent patterns and association rules in the text. Finally, the entity recognition model can obtain the user's potential search intention.

[0084] For example, a user can enter the text "What is the mobile phone number of XXXX" in the browser, which is eventually converted into two entities "XXXX" and "Mobile Phone Number" through the entity recognition model. The two entities are subordinate to each other, and the corresponding mobile phone number field can be searched and displayed based on the entity name "XXXX". It is understandable that the updated search statement at this time can be "Search for the mobile phone number of XXXX".

[0085] In a feasible embodiment, before the step of determining, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results, the method further includes:

[0086] Step B10, filtering sensitive words in the search sentence, and correcting the filtered search sentence;

[0087] Step B20, segmenting the corrected search sentence to obtain at least one search word;

[0088] Step B30, removing stop words from each search term according to a preset stop word list, adding synonyms based on the search term after the stop words are removed and a preset synonym table, to obtain an updated search term;

[0089] Step B40, modifying the updated search term into a preset standard grammatical form, so that the search result corresponding to the updated search term is returned through the search service.

[0090] In order to improve the usability of search statements, it is necessary to perform sensitive word filtering, word segmentation, stop word removal, synonym addition and other steps so that the returned search results can fully and accurately reflect user needs.

[0091] Exemplarily, in the embodiment of the present application, the search terms in the inverted index table are searched and recalled mainly through TD-IDF (Term Frequency–Inverse Document Frequency) and BM25 (Best Matching 25, an information retrieval algorithm), including: after the corresponding search terms are segmented, synonyms are added, sensitive words and stop words are filtered to obtain the word combination to be recalled, and based on the relevance of the words in the inverted index to the document, the following steps are specifically included:

[0092] Sensitive word matching: If a sensitive word is included, stop searching and return directly;

[0093] Error correction words: can be combined with the preset built-in model and operation platform configuration for error correction;

[0094] Stop words: Remove stop words based on the results after word segmentation;

[0095] Synonyms: Based on the results of word segmentation, add synonyms by combining standardized synonyms and the pre-configured synonym table in the platform (including one or more synonyms corresponding to each word);

[0096] Word segmentation: word segmentation based on preset word segmentation and new words configured on the operation platform;

[0097] Specifically, before word segmentation, the error correction words are replaced, and after replacement, the relevance between the words and the document based on the inverted index is used. It should be noted that the above-mentioned "word segmentation" situations include: word segmentation during fuzzy matching search; in the case of synonyms, the search words will be segmented, and after the synonyms are added, the corresponding synonyms will also be recalled using a similar strategy.

[0098] Finally, rewrite it into the ES (Elasticsearch, full-text search) standard query syntax for the search service to search in the stock documents, return the corresponding search results, and obtain the ES recall results. Among them, the recall number is greater than the number of paging items, so as to be used for sorting search results.

[0099] In a feasible embodiment, the step of sorting the search results according to the matching conditions of the search results, the full match values, the partial match values ​​and the release time feature values ​​respectively corresponding to the search results, and obtaining and returning the sorted search results includes:

[0100] Step S41, determining a plurality of features corresponding to each search result and a feature value corresponding to each feature according to the matching conditions of each search result, the full matching value, the partial matching value and the publishing time feature value corresponding to each search result, wherein the matching conditions include at least one of a title match and a text match;

[0101] Step S42, calculating the feature score of each search result according to the feature value and the corresponding weight corresponding to each feature, wherein the weight of the full match value is greater than the weight of the partial match value, and the weight of the title match is greater than the weight of the text match;

[0102] Step S43, sorting the search results in descending order based on the feature scores of the search results, obtaining and returning the sorted search results.

[0103] In the embodiment of the present application, the recall results (i.e., search results) can be further sorted based on the built-in sorting model and the sorting formula specified by RankModelInfo (probability-based model information). The sorting method calculates the score of each feature based on RankModelInfo, and obtains the sum of the scores of each feature of each search result, i.e., the feature score, and finally sorts according to the feature score.

[0104] In step S41, it is necessary to match the title and text of each search result with the search statement, as well as the full match value, partial match value and release time characteristic value, etc. For example, if the title and / or text matches successfully, the corresponding characteristic value is 1, if the match fails, the characteristic value is 0, if it is a full match (complete hit), the characteristic value is 1, and if it is a partial match (partial hit), the characteristic value is 0. The release time characteristic value can be determined by the reciprocal of the number of days (or hours) from the current time, because the closer the release time is to the current time, the larger the release time characteristic value. It should be noted that the above method of determining the characteristic value is only an example and does not limit the embodiments of the present application.

[0105] In addition, the weight of each feature is predetermined and set according to the degree of influence on the sorting. For example, the weight of the full match value is greater than the weight of the partial match value, and the weight of the title match is greater than the weight of the text match. The product of the feature value of each feature and the corresponding weight is calculated to obtain each feature score. Finally, the sum of the feature scores of the multiple features corresponding to each search result is calculated, and the search results are sorted in order from large to small to obtain the sorted search results. The interface of the search platform used by the user to enter the search statement is then updated to display the sorted search results.

[0106] In addition, in another feasible embodiment, in order to improve the accuracy and stability of feature score calculation, corresponding fields, weights, thresholds, normalization, etc. can be configured for each feature to achieve comprehensive sorting of search results based on multiple sorting features.

[0107] For example, the normalization process can be implemented by a functional formula:

[0108]

[0109] Where a is the normalization coefficient used to adjust the score distribution after normalization, and the default value is 1; the x vector is the exponential vector of the constant.

[0110] In addition, the aforementioned RankModelInfo structure is: contains the ranking formulas for the original query and the extra query, each type of ranking formula is of dict (a built-in data type) type, and the dict name is the corresponding query field name. The specific content of each type of ranking formula includes "features" (features) and "aggregate_algo" (a function used to eliminate the restrictions of conditional formats), that is, aggregate_algo supports "weight_avg" (weighted average), that is, weighted sum of all features; "features" is a list of specific ranking features, where each feature is in dict (dictionary) format, including feature name and feature parameters.

[0111] In another feasible embodiment, the search result sorting based on the feature scores of each feature may also include that the official document sorting is mainly performed by calculating the full hit feature, the word segmentation hit feature, the search term hit ratio feature, the search term in the field hit ratio feature, the time feature weight, etc.

[0112] like Figure 4 As shown in the figure, the score of each feature comes from the relevance and time rules. The relevance rules include: full match takes precedence, among which the full match score of the title accounts for the highest proportion; the full match score of the text accounts for the second largest proportion; for example, the full match score of the drafter, drafting department, document number, and attachment title accounts for an equal proportion, which is lower than the score of the text; the weight of the attachment text accounts for the lowest proportion. Based on this logic, for example, when all fields are fully matched, the score accounts for the highest proportion and ranks top 1, as shown in Table 1 below (the shaded part indicates a successful match). The specific rules include: if the full match score accounts for the same proportion, they are arranged in reverse chronological order; if both data titles are fully matched and other fields are partially matched, they are sorted by time; the partial match weight is higher than the maximum time weight; if only one field is partially matched, they are sorted by relevance; the time rules include: strictly sorting by time according to the ES search terms.

[0113] Table 1:

[0114]

[0115] The steps of returning the search results to the user specifically include: first, the search service returns the sorted search results through the interface, and the AI ​​(Artificial Intelligence) assistant presents the search results in the form of cards. The user can also click on the search result card to view more, thereby entering the search details page to query more information.

[0116] In another feasible embodiment, before the step of sorting the search results in descending order based on the feature scores of the search results, the method further includes:

[0117] Step C10, screening search results whose feature scores are lower than a preset score threshold;

[0118] Step C20, filtering out search results whose feature scores are lower than a preset score threshold;

[0119] Step C30: based on a preset custom score dictionary, updating the feature score of the search result of the type corresponding to the custom score dictionary to obtain an updated feature score.

[0120] The text search result sorting method provided in the embodiment of the present application also provides a method for screening, filtering and custom updating the feature scores of the search results. Specifically, the embodiment of the present application can also be applied to support setting a sorting formula for additional query fields. The user can add additional query fields in the "extras" field, set the corresponding sorting formula, and calculate the sorting score of the additional query.

[0121] Specifically, in the process of filtering search results, it can be done through a custom filter field. For example, by default, the filter score in rankModelInfo is the same as the ranking score. If the score of a document is less than the set MinScore (minimum score), the document will not be returned; at the same time, the user can additionally specify that some features do not participate in the calculation of the filter score. Users can pass in configurations that do not need to participate in the calculation of the filter score in the "filter_ignore_features" (filter features) of extras. The configuration must be of dictionary type and have the same format as rankModelInfo. The configuration content contains all feature names that do not need to participate in the sorting (fields must be specified), so as to perform separate filter score calculations.

[0122] The text search result sorting method of the embodiment of the present application also supports user-defined score updates: rankModelInfo supports users to pass in a custom score dictionary in extras (feature name: user_defined_score_dict (user-defined score), format {id(string):score(float)} (identity, string, score, floating point number), the sorting model will add the document's ranking score to the corresponding score as the final score of the document, and the sorting model will filter documents according to the original ranking score, that is, documents with original scores less than "MinScore" will still be filtered. The above-mentioned custom score update can be applied in certain special scenarios. For example, users in Department A need to query official documents of the department. After entering the corresponding search statement, search results including other departments may be searched. At this time, the feature scores of the search results corresponding to the official documents of Department A in all search results can be increased through the custom score dictionary, so that these search results are ranked higher, so that users can query the files they need faster and more efficiently.

[0123] For ease of understanding, in a feasible embodiment, a feasible text search and result sorting full step process is as follows: Figure 5As shown in the figure, the user first submits the search term through the browser input box, and then the Gateway performs authentication; after the authentication is passed, the intervention Json is read through the search service, and then the preset standard request is spliced; the request is initiated through the algorithm & interface; through QP (Quadratic Programming) analysis, sensitive words in the search term are removed, and the error correction words replace the wrong words, and then the word segmentation is performed, the stop words are removed, and synonyms are added; the query received by ES is rewritten, and ES returns the ES recall result through the Index function, and then the feature score of each recall result is calculated respectively, and the ES recall result is sorted based on the feature score; if Minsore is configured, data truncation is performed according to Minsore combined with reRankSize, and finally the sorted search results are obtained; the sorted search results are returned to the browser through the search service through Gateway for users to browse.

[0124] In another feasible embodiment, the present application embodiment also proposes a verification test method for the sorting effect of the aforementioned text search result sorting method, and the specific steps include: keyword sampling, recording the results of periodic hot words, summarizing the search words with the highest frequency as the test execution keyword samples, and executing the text search test process. Specifically, the test execution method adopts the search page execution, and the search page execution can be adopted. If problems are found, the problem is analyzed through the interface access, such as Figure 6 As shown, directly enter the search term "meeting minutes" in the search bar on the search page, and view the search results on the official document page. For official document search interface queries, directly use the interface call tool to make background calls on the operation platform, obtain the results returned in json format, and perform search accuracy and sorting. Finally, the query effect is judged and executed according to the business requirements and rules. The overall judgment rules include: according to the search range selection, the score of each item in the full match within the range is accumulated, and the hit score result is calculated; different match scores, high hit scores are ranked before low scores; the same match score is based on the time of publication, and the new time is ranked before the old time; the end sorting is partially matched, and the search terms with more hits are ranked first. If all the conditions corresponding to the above judgment rules are met, it can be considered that this text search and result sorting have achieved the expected effect.

[0125] In the embodiments of the present application, there is a good sorting effect, especially for the retrieval of internal enterprise information (such as: personnel address books, work orders, OA (Office Automation) office modules, official documents, etc.). On the one hand, it meets the needs of users who want to conduct comprehensive search evaluation and output search according to multiple document types, and proposes a comprehensive search sorting and recovery strategy implementation method. The sorting considers the complex analysis of the complete or partial matching of the relevance of different text types (i.e., title, text, abstract, attachment, etc.), and sorts the search by calculating the combination of full hit features, word segmentation hit features, search word hit ratio features, search word hit ratio features in the field, and time feature weights, and adds synonyms, filters sensitive words and stop words and other sorting elements; on the other hand, it considers the application of artificial intelligence technology to more accurately identify the user's search purpose, adopts a more accurate way to obtain the user's search target, and outputs a more reasonable search sorting to improve the search accuracy and efficiency.

[0126] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the method for sorting text search results of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0127] The present application also provides a text search result sorting device, referring to Figure 7 , the text search result sorting device comprises:

[0128] A parameter reading module 10, configured to read pre-configured intervention parameters when receiving a search statement, wherein the pre-configured intervention parameters at least include a full match value, a partial match value, and a release time feature value;

[0129] A statement search module 20, configured to determine, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results;

[0130] A sentence matching module 30, used to determine the matching status of each search result with the search sentence according to the title and text content of each search result;

[0131] The result sorting module 40 is used to sort the search results according to the matching status of the title and text of each search result with the search statement, the full match value, partial match value and release time feature value corresponding to each search result, and obtain and return the sorted search results.

[0132] In one embodiment, the sentence search module 20 is further used to:

[0133] Input each stock document into the preset entity recognition model to generate the corresponding entity dictionary;

[0134] Determining whether the search statement matches the entity dictionary;

[0135] If there is a hit, searching the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result in each of the existing documents;

[0136] If there is no hit, it is determined whether the search statement has all hits in the existing documents, and if all hits are found, the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result are returned.

[0137] In one embodiment, the parameter reading module 10 is further used for:

[0138] Inputting the search statement into a preset entity recognition model to obtain at least one entity corresponding to the search statement;

[0139] An updated search statement is generated according to the association relationship between the entities, wherein the updated search statement is used for the search service to determine the corresponding search result.

[0140] In one embodiment, the sentence search module 20 is further used to:

[0141] Filtering sensitive words in the search sentence, and correcting the filtered search sentence;

[0142] Segment the corrected search sentence to obtain at least one search term;

[0143] Removing stop words from each of the search terms according to a preset stop word list, and adding synonyms based on the search terms after the stop words are removed and a preset synonym table to obtain updated search terms;

[0144] The updated search term is modified into a preset standard grammatical form so that the search result corresponding to the updated search term is returned through the search service.

[0145] In one embodiment, the result sorting module 40 is further configured to:

[0146] Determine multiple features corresponding to each search result and feature values ​​corresponding to each feature according to the matching conditions of each search result, the full matching value, the partial matching value and the release time feature value corresponding to each search result, wherein the matching conditions include at least one of title matching and text matching;

[0147] Calculate the feature score of each search result according to the feature values ​​and weights corresponding to each of the features, wherein the weight of the full match value is greater than the weight of the partial match value, and the weight of the title match is greater than the weight of the text match;

[0148] Based on the feature scores of the search results, the search results are sorted in descending order, and the sorted search results are obtained and returned.

[0149] In one embodiment, the result sorting module 40 is further configured to:

[0150] Filter search results whose feature scores are below a preset score threshold;

[0151] Filter out search results whose scores of each of the features are lower than a preset score threshold;

[0152] Based on a preset custom score dictionary, the feature scores of search results of the type corresponding to the custom score dictionary are updated to obtain updated feature scores.

[0153] The text search result sorting device provided by the present application adopts the text search result sorting method in the above embodiment, which can solve the technical problem that the search result sorting rationality of the current text search scheme is poor. Compared with the prior art, the beneficial effects of the text search result sorting device provided by the present application are the same as the beneficial effects of the text search result sorting method provided by the above embodiment, and the other technical features in the text search result sorting device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0154] The present application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the text search result sorting method in the above-mentioned embodiment 1.

[0155] Reference below Figure 8 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic devices in the embodiments of the present application may include but are not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0156] like Figure 8 As shown, the electronic device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the electronic device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows an electronic device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0157] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0158] The electronic device provided by the present application adopts the text search result sorting method in the above embodiment, which can solve the technical problem that the search result sorting rationality of the current text search scheme is poor. Compared with the prior art, the beneficial effects of the electronic device provided by the present application are the same as the beneficial effects of the text search result sorting method provided by the above embodiment, and the other technical features in the electronic device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0159] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0160] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0161] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the text search result sorting method in the above embodiment.

[0162] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0163] The computer-readable storage medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0164] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by an electronic device, the electronic device: when receiving a search statement, reads pre-configured intervention parameters, wherein the pre-configured intervention parameters at least include a full match value, a partial match value and a release time characteristic value; determines, according to the search statement and the pre-configured intervention parameters, multiple search results corresponding to the search statement and the full match value, partial match value and release time characteristic value corresponding to each of the search results; determines the matching status of each of the search results with the search statement according to the title and body content of each of the search results; sorts each of the search results according to the matching status of each of the search results and the full match value, partial match value and release time characteristic value corresponding to each of the search results, and obtains and returns the sorted search results.

[0165] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0166] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0167] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0168] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned text search result sorting method, and can solve the technical problem that the search result sorting rationality of the current text search scheme is poor. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the text search result sorting method provided by the above-mentioned embodiment, and will not be repeated here.

[0169] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned text search result sorting method when executed by a processor.

[0170] The computer program product provided by the present application can solve the technical problem that the search result sorting of the current text search solution is poorly rational. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the text search result sorting method provided by the above embodiment, which will not be repeated here.

[0171] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for sorting text search results, characterized in that: The text search result sorting method comprises: When receiving a search statement, reading preconfigured intervention parameters, wherein the preconfigured intervention parameters at least include a full match value, a partial match value, and a release time feature value; Determine, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results; Determining the matching status of each search result with the search statement according to the title and text content of each search result; According to the matching conditions of the search results, the full matching values, partial matching values ​​and release time feature values ​​respectively corresponding to the search results, the search results are sorted to obtain and return the sorted search results.

2. The text search result sorting method according to claim 1, characterized in that: The step of determining a plurality of search results corresponding to the search statement and a full match value, a partial match value and a release time feature value respectively corresponding to each of the search results comprises: Input each stock document into the preset entity recognition model to generate the corresponding entity dictionary; Determining whether the search statement matches the entity dictionary; If there is a hit, searching the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result in each of the existing documents; If there is no hit, it is determined whether the search statement is fully hit in the existing documents, and if all hits are found, the search results corresponding to the search statement and the full match value, partial match value and release time feature value corresponding to each search result are returned.

3. The text search result sorting method according to claim 1, characterized in that: Before the step of reading the preconfigured intervention parameters, the method further comprises: Inputting the search statement into a preset entity recognition model to obtain at least one entity corresponding to the search statement; An updated search statement is generated according to the association relationship between the entities, wherein the updated search statement is used for the search service to determine the corresponding search result.

4. The text search result sorting method according to claim 1, characterized in that: Before the step of determining, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value, and a release time feature value respectively corresponding to each of the search results, the method further includes: Filtering sensitive words in the search sentence, and correcting the filtered search sentence; Segment the corrected search sentence to obtain at least one search term; Removing stop words from each of the search terms according to a preset stop word list, and adding synonyms based on the search terms after the stop words are removed and a preset synonym table to obtain updated search terms; The updated search term is modified into a preset standard grammatical form so that the search result corresponding to the updated search term is returned through the search service.

5. The text search result sorting method according to claim 1, characterized in that: The step of sorting the search results according to the matching conditions of the search results, the full match values, the partial match values ​​and the release time feature values ​​respectively corresponding to the search results, and obtaining and returning the sorted search results comprises: Determine multiple features corresponding to each search result and feature values ​​corresponding to each feature according to the matching conditions of each search result, the full matching value, the partial matching value and the release time feature value corresponding to each search result, wherein the matching conditions include at least one of title matching and text matching; Calculate the feature score of each search result according to the feature values ​​and weights corresponding to each of the features, wherein the weight of the full match value is greater than the weight of the partial match value, and the weight of the title match is greater than the weight of the text match; Based on the feature scores of the search results, the search results are sorted in descending order, and the sorted search results are obtained and returned.

6. The text search result sorting method according to claim 5, characterized in that: Before the step of sorting the search results in descending order based on the feature scores of the search results, the method further includes: Filter search results whose feature scores are below a preset score threshold; Filter out search results whose scores of each of the features are lower than a preset score threshold; Based on a preset custom score dictionary, the feature scores of search results of the type corresponding to the custom score dictionary are updated to obtain updated feature scores.

7. A text search result sorting device, characterized in that: The text search result sorting device comprises: A parameter reading module, configured to read pre-configured intervention parameters when receiving a search statement, wherein the pre-configured intervention parameters at least include a full match value, a partial match value, and a release time feature value; A statement search module, used to determine, according to the search statement and the preconfigured intervention parameters, a plurality of search results corresponding to the search statement and a full match value, a partial match value and a release time feature value respectively corresponding to each of the search results; A sentence matching module, used to determine the matching status of each search result with the search sentence according to the title and text content of each search result; The result sorting module is used to sort the search results according to the matching conditions of the search results, the full matching values, partial matching values ​​and release time feature values ​​corresponding to the search results, and obtain and return the sorted search results.

8. An electronic device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the text search result ranking method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the text search result sorting method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method for ranking text search results according to any one of claims 1 to 6 are implemented.

Images