CN110674384B - Component model matching method - Google Patents
Component model matching method Download PDFInfo
- Publication number
- CN110674384B CN110674384B CN201910926938.9A CN201910926938A CN110674384B CN 110674384 B CN110674384 B CN 110674384B CN 201910926938 A CN201910926938 A CN 201910926938A CN 110674384 B CN110674384 B CN 110674384B
- Authority
- CN
- China
- Prior art keywords
- model
- product
- retrieval
- search
- matching
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9532—Query formulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/242—Query formulation
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a component model matching method, which comprises the following steps: step 1: splitting and sorting the keywords input by the user according to a preset splitting rule; step 2: searching the split and sorted keywords in a data source by using an ES (ES); and step 3: and displaying the retrieval result of the ES according to the matching degree score, recording a result set with the score larger than the preset matching degree in the retrieval result as a search matching result set, and finishing the search. Aiming at the problems that the retrieval speed is limited by a memory and the retrieval result of an electric device is accurate in the prior art, the retrieval speed and the retrieval accuracy are improved by pre-arranging a data source, and a matching degree scoring rule is established and a screening application is established to optimize the retrieval result. In addition, the application also provides a replacement scheme, is initiated in the industry, especially designs the matching and replacing process aiming at the special attributes of the components, and has very good practical value.
Description
Technical Field
The invention relates to the technical field of computer software development, in particular to a component model searching, matching and replacing method.
Background
The elastic search is a distributed extensible real-time search and analysis engine, is a search engine established on the basis of a full-text search engine Apache Lucene (TM), and not only comprises a full-text search function, but also can perform the following work: distributed real-time file storage, and indexing each field so that it can be searched; a distributed search engine for real-time analysis; it can be extended to hundreds of servers, handling PB-level structured or unstructured data. At present, the common search engines in China, such as Baidu (in casio, cloud analysis, network alliance, prediction, library, direct number, wallet, wind control and other businesses, all apply ES, 30TB + data is imported into a single cluster every day, and 60TB + data, total data structure-log, Ali bus, Tencent and other companies all use the ES.
However, the search engine of the ES is a file system cache that depends heavily on the bottom layer, and if the memory allocated by the engine is insufficient and cannot accommodate all index segment file index data files, the retrieval speed is greatly affected; meanwhile, in the existing component model search in the electronic field, the search is generally performed completely based on the similarity algorithm, the search scheme can cause that important factors are ignored, the text content with a large number of characters is emphasized too much, the search accuracy is greatly influenced, a user can hardly find a desired product in a large string of search results, and the user may not be familiar with all brand products but can hardly search for a replaceable product due to the factors such as cost and the like.
Disclosure of Invention
The following presents a simplified summary of embodiments of the invention in order to provide a basic understanding of some aspects of the invention. It should be understood that the following summary is not an exhaustive overview of the invention. It is not intended to determine the key or critical elements of the present invention, nor is it intended to limit the scope of the present invention. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is discussed later.
According to one aspect of the application, in order to overcome the problem that the content really wanted to be searched is difficult to find due to the retrieval result obtained directly based on the similarity algorithm, a component model matching method is provided, and comprises the following steps:
step 1: splitting and sorting the keywords input by the user according to a preset splitting rule;
step 2: searching the split and sorted keywords in a data source by using an ES (distributed full-text search, ES for short);
and step 3: and displaying the search results of the ES according to the matching degree scores (the search results can be sorted from high to low in descending order), recording a result set with the score larger than the preset matching degree in the search results as a search matching result set, and finishing the search.
The search matching degree scoring rule is as follows:
step 31: according to the common search content, the attribute domain is summarized and sorted, the attribute domain at least comprises a model domain and a parameter domain, wherein the model domain is generated by processing and sorting the large models and the order marks of the products; the parameter domain is generated by processing and sorting key parameters such as contact form, leading-out terminal mode and the like;
step 32: setting a weight proportion according to a preset strategy aiming at the model domain and the parameter domain, wherein the weight proportion is preset as follows: model domain > parameter domain > other domains, the other domains being domains in the attribute domain other than the model domain and the parameter domain; the preset strategy is as follows: when the retrieved attribute is a model type, the weight ratio is set as: the model field > parameter field > other fields, the priority that will meet the condition is arranged to the forefront, at this moment, the model field presets the weight value to be 5-10; when the retrieved attribute is a parameter class, the weight ratio is set as: parameter domain > model domain > other domains, arrange the meeting condition to the front end, at this moment, the parameter domain presets the weight value to be 5-10; when the retrieved attributes are other domains, automatically calculating scores according to all the documents;
step 33: receiving a keyword of the user, and if the keyword is a suggested word selected by the user through the system, performing step 331; if the keyword is a user-defined keyword, execute step 332:
step 331: acquiring the suggested word and the dimensionality of the suggested word, directly and completely matching the suggested word, and accurately obtaining a retrieval result;
step 332: and judging whether the type of the keyword is a letter, a number or a Chinese character according to the regular expression:
if the type of the keyword is a letter or a number, preferentially retrieving a matched model domain according to the weight proportion of the model domain in the step 32, and obtaining a matching score according to the score calculated by the similarity algorithm and the weight proportion score; that is, the matching score is the score + the weight scale score, where the weight scale score is the (weight/scale base) score, and the weight and scale base are both custom values;
if the type of the keyword is Chinese character, preferentially searching the matched parameter domain according to the weight proportion of the parameter domain in the step 32, and obtaining a matching score according to the score calculated by the similarity algorithm and the weight proportion score;
then fuzzy search is carried out on other query sources (such as product content, specification parameters, attribute parameters and the like), and the score is calculated according to a similarity algorithm.
In step 32, the weights of the parameter domain and the model domain can be set by background control, and the specific gravity is specifically set according to the execution sequence of different attributes, so that the accuracy of the preferentially retrieved domain is higher and is used as a priority sequence, and the priority sequence is arranged in front; that is, the background can be set by user-defined mode, and the adjustment is carried out as required: for example, when the retrieved attribute is a model class, the model domain weight may be set to 5-10, and the scale base may be set to 100, that is, when the keyword is set to 10 in the whole term search weight in the model domain, the matching degree score is the final score + (10/100) × score, and when the keyword is set to 5 in the model domain, the matching degree score is the final score + (5/100) × score.
Setting a weight proportion for the model domain, the parameter domain and other domains (domains except the model domain and the parameter domain in the attribute domain) in step 31, setting a weight grading proportion according to a preset strategy (if the retrieved model is similar, the model is ranked to the forefront preferentially, if the retrieved parameter is similar, the model is ranked to the forefront meeting the conditions, if the retrieved parameter is similar, the model is ranked to the forefront, and if the retrieved model is not within the 2 classes, the grading is automatically calculated according to all documents): model field > parameter field > other fields.
The amount of the search matching result set obtained through the step 3 is still large, and for this reason, the method further comprises a step 4 of secondary screening: acquiring condition items selected by a user, acquiring condition item values of non-image condition items according to a similarity algorithm, acquiring condition item values of the image condition items through scanning according to an image similarity algorithm, performing retrieval matching on the condition item values in a search matching result set to obtain a secondary search matching result set corresponding to the condition item values, performing intersection (A & ltn & gt B & ltn & gt C) calculation on the secondary search matching result sets to obtain a final intersection part, namely a final retrieval matching result set, and finishing the search.
The condition items of the secondary screening in the step 4 are based on that the accuracy of a result set searched by the user may not exactly remember the specific content or the keyword content of the product is low, and the further guiding and the type selection are performed, so that the search result set can be further reduced to a greater extent, and the search accuracy of the user is improved; the condition items are screening items (including but not limited to pin numbers, contact forms, coil voltages, rated loads, leading-out terminal modes, pin position diagrams and the like) defined according to the characteristics of the components, and the screening items can be customized in the background and can be adjusted as required.
As a preferred scheme, the splitting rule in step 1 is as follows:
setting a preset rule, and presetting non-splitting numbers and splitting and recombining letters according to the characteristics of the coding of the components;
setting a self-defined splitting word, and summarizing key words (such as product models, product specifications, contact forms, leading-out terminal mode parameters and the like) according to the special attributes of the components;
setting a user-defined segmentation word, analyzing and searching a hot word according to big data, and summarizing common search keywords;
judging whether the keyword has a blank space or other special symbols:
if not, performing retrieval; the specific retrieval process comprises the following steps: the whole word matching retrieval, the whole word is split according to a user-defined split word and then matched and retrieved, the whole word is subjected to match retrieval according to the minimum particle (namely a single character or a single letter), and the whole word is split according to the minimum particle and then freely combined into different words and then subjected to match retrieval; the whole word is the whole keyword;
if the space or other special symbols are contained, the space or other special symbols are removed to form a new keyword, and then the search is executed. The specific process comprises the following steps: splitting the space or other special symbols into different keywords, matching the split keywords, performing free combination on the split keywords and then performing matching retrieval, removing the space or other special symbols to form a new keyword for matching retrieval, performing whole word matching retrieval, splitting the matching retrieval according to a user-defined splitting word (which can be preset), recombining the new keyword matching retrieval after splitting the user-defined splitting word and then performing matching retrieval, splitting the matching retrieval according to the minimum particle, and recombining the new word matching retrieval after splitting according to the minimum particle.
Meanwhile, in the splitting process of the keywords, a large number of fields which cannot be searched are split into a single storage space, so that vertical splitting of the sub-database and sub-table is realized, and the retrieval speed can be further improved.
In addition, aiming at the keywords input by the user, when deletion occurs, the keywords can be directly sent to the MQ message queue to inform the ES, and the suggested word is deleted; when updating or inserting occurs, besides updating or adding new suggested words, the keywords can be classified, and reordering can occur, so that the keywords can be quickly located during retrieval.
The method comprises the steps of splitting and sorting the splitting rules of the keywords input by a user to form various keywords, keywords and combinations thereof, then searching all the keywords, the keywords and the combinations thereof in a data source by using ES (ES), and obtaining a great number of search results.
Specifically, the pre-sorting of the data sources includes: the hot data which is frequently retrieved is classified, and the hot data can be stored in a filesystems cache, so that the retrieval speed can be greatly improved. When the retrieval is carried out each time, the cached hot data is accessed in advance, and the hot data enters the filesystems cache, so that the performance is accelerated by one time during the retrieval.
To the particularity of electron field components and parts, arrange in advance the data source, still include: the attributes describing the same component are uniformized in advance, so that the values with the same meaning have a uniform form; for example, if the models of the components are different from one manufacturer to another, the data sources may not be searched if the data sources are not sorted in advance. The attribute uniformity for describing the same component is to extend the model attribute corresponding to the same component in the data source to the component model set of all manufacturers, such as relays, and the model attribute may be [ JL-8GB/12, TD010-C6H6O-A, HRA- (S) -DC12V ].
To the particularity of electron field components and parts, arrange in advance the data source, still include: the method comprises the steps of projecting the numerical value attributes of the components into a specific small range in proportion in advance to eliminate the deviation of the retrieval result caused by the different sizes of the numerical value attributes.
The data source is pre-arranged, and the method further comprises the following steps: the data source is divided into 6 dimensions, namely the model, the specification and the model, the classification, the brand, the product parameters and the product content, and during retrieval, which dimension is preferentially retrieved is determined according to the keywords input by the user, so that the retrieval matching degree is improved. The method comprises the steps that a user inputs keywords, if the user finishes searching by selecting suggested words, the position of the suggested words and contents similar to the top and bottom of the suggested words are quickly located according to the dimensionality of the selected suggested words, and then product information is associated according to a star model; if the input keyword is directly used for searching, which dimension is preferentially searched is judged according to the input content, and if the input content is letters or numbers, the following steps are performed in sequence: searching for a model dimension, a specification model dimension, a product parameter dimension and a product content dimension; if the Chinese characters are input, searching is carried out according to the dimension of sequential classification, the dimension of brand, the dimension of product parameters and the dimension of product contents.
Wherein, the model dimension: the method is from the product model, when in storage, the data are stored in the document according to the sequence of the first letters, and the user inputs keywords so as to quickly position the query position.
Specification and model dimension: the product ordering marks are used for sorting according to the product models and the specification sequence during storage, so that similar specification models can be arranged together, and the retrieval speed is improved.
Classification dimension: the method is from the classification of the shopping mall, the similar classifications are arranged together according to the sequence of the classification names, and the retrieval speed is improved.
Brand dimension: the method is from the platform brands, and arranges similar brands together according to the sequence of the brand names, so that the retrieval speed is improved.
Product parameter dimension: the source specification parameters and the attribute parameters are sorted according to the name sequence of the attribute items, and the similar parameters are discharged together, so that the retrieval speed is increased.
Product content dimension: the content is sorted according to similarity, wherein the content is derived from the product characteristics and the product titles.
The retrieval interface is used for judging which dimension is preferentially retrieved according to the input content by inputting keywords by a user and pulling down guide suggested words, such as: if the input is the beginning of a letter or a number, preferentially searching the dimension of the model and the dimension of the model specification; if the Chinese character is input, the classification dimension, the brand dimension, the product parameter dimension and the product content dimension are preferentially searched.
Product information content dimension: a main picture of a source product, a product appearance picture, a product foot bitmap and the like.
Furthermore, on the basis of searching the data source of the suggested word, other contents needed by the result set are added, and product information content dimensions are increased more, so that all dimensions are connected through a star model.
The retrieval in the data source by using the ES specifically comprises the following steps: the user optionally submits accurate query conditions (brand, type, filter field, attribute, specification, etc.), and if no condition is selected, the keywords are sent to the ES for searching: completely matching the order mark through the keyword, and returning the order mark if the order mark is found; if not, adopting the keyword to search the large model (fuzzy search) and returning the large model; if the conditions (brand, type, filtering field, attribute, specification and the like) + the keywords are selected to send the ES for searching, the order mark is completely matched through the keywords and the searching conditions, if the order mark is found, the order mark is returned, and if the order mark is not found, the keywords and the searching conditions search the large model (fuzzy search) and return the large model.
Meanwhile, in practical application, the types of the components are different corresponding to the same component in different companies, so that the search result is more comprehensive and more accurate, and a user is helped to quickly search the matched product; the matching method of the present application further includes optional replacement steps, which may be used alone or in combination with the aforementioned search process, and specifically, the replacement steps include:
step 51: searching keywords input by a user in a data source by using an ES (distributed full text search, ES for short), and searching a model which is completely matched and confirmed to be input according to a cosine vector similarity calculation formula;
step 52: confirming the unique product according to the retrieval result, acquiring a pin bit map and necessary replacement parameters of the unique product, searching and comparing a model product information result set of the same pin bit map according to the pin bit map, and performing retrieval matching on the result set according to the necessary replacement parameters to obtain a model product information result set A; transmitting the confirmed specific model to model maintenance contrast on the ES to retrieve a matchable model information result set B; and after the result set A and the result set B are subjected to de-duplication, the result set A is a result set which can be matched. Further alternate operations may be performed by selecting an order label if a particular product is desired to be identified.
Wherein, the step 51 specifically includes: the method comprises the steps of obtaining a pin diagram of a keyword, namely taking a current pin diagram as a retrieval diagram, obtaining a pin diagram with high matching degree according to an SIFT algorithm, extracting and recording the pin diagram into a pin diagram set, and obtaining a matched product model information set according to the pin diagram set. Firstly, graying a bitmap, namely converting the bitmap into a grayscale map; then constructing a Dog scale space (namely an image pyramid and an image difference pyramid); searching (including position, angle and layer) and positioning key points, namely performing initial extreme value detection in a differential pyramid, fitting Taylor expansion of the detected extreme value points to perform sub-pixel positioning, excluding points with small threshold values, correcting the extreme value points, deleting unstable points, finding out feature points, confirming the main direction of the feature points, integrating feature point descriptors to form final description vectors, and performing the operation on all the feature points in the image to generate the feature point description vectors of the whole image; and then, a feature point set (namely an image feature point description vector) is made for each foot bitmap in the foot bitmap library, the feature vector on the original image is matched with the feature vector of each target image to obtain a matching value, and the foot bitmap with high matching degree (the edge contains the matching with higher matching degree and more matching points) is extracted and recorded into the foot bitmap set.
To further screen for accuracy, the following operations are also performed for a matchable result set: acquiring the number of pins of the retrieval model and other matchable product models, retrieving the number of pins of the retrieval model from a formula, rejecting the product model with the number of pins of the retrieval model being greater than the number of pins of the matchable product model, and acquiring a matchable product model information set; in order to further perform screening accuracy, the external dimensions (i.e. length, X width, X height) of the search model and other matchable product models are obtained and converted into a uniform unit, and then the difference between the external dimensions of the search model and other matchable product models is calculated, when the tolerance meets a preset interval (for example, -0.5< tolerance <0.5(mm)), the search model is considered to be within the matching range, and when the tolerance is outside the matching range, the search model is considered to be unmatchable.
When in use, the method can further comprise the step 53: if the product with the specific specification needs to be refined for replacement, the specification needs to be further selected, the product with the specific specification needing to be replaced is confirmed, necessary parameters for current replacement with the specific specification are obtained from product parameter dimensions, then each replaced parameter value is obtained from the product parameter dimensions (a product parameter dimension set which can be matched with the product model set on the basis is obtained according to a cosine vector similarity calculation formula) for searching and matching, a final matched product model set of each replaced parameter is obtained, and then intersection calculation is carried out on the matched product model sets of each replaced parameter through a formula (A & ltn & gt B & ltn & gt C), so that a final intersection part is obtained, namely the replaceable product specification.
Wherein, according to the search result, the unique product is confirmed, specifically: if the brand is a single brand and a large model, confirming the only product; if the product is of multiple brands, the brand needs to be supplemented, and the only product can be confirmed; if the content of the keyword part is accurately matched with the model and is a single brand, the unique product can be confirmed; if the partial content of the keyword is accurately matched with the models and the models of multiple brands, the brands need to be supplemented, and the only product can be confirmed; and if the model information of the keywords cannot be confirmed or the keywords do not have matchable information, prompting the user.
Compared with the prior art, the method has the following advantages: 1. aiming at the problems that the retrieval speed is limited by a memory and the retrieval result of an electric device is accurate in the prior art, the retrieval speed and the retrieval accuracy are improved by pre-arranging a data source, and a matching degree scoring rule is established and a screening application is established to optimize the retrieval result. 2. After the data sources are pre-arranged, the hot data which are frequently retrieved can be classified, and the hot data can be stored in a cache, so that the retrieval speed can be greatly improved. When the retrieval is carried out each time, the cached hot data is accessed in advance, and the hot data enters the filesystems cache, so that the performance is accelerated by one time during the retrieval. 3. Meanwhile, in the splitting process of the keywords, a large number of fields which cannot be searched are split into a single storage space, so that vertical splitting of the sub-database and sub-table is realized, and the retrieval speed can be further improved. 4. In addition, the application also provides a replacement scheme, which is initiated in the industry, particularly designs a matching and replacing process aiming at special attributes of components (names of various manufacturers of the same device are different, the pin numbers of products of different models of the same device are possibly different, and the like), and simultaneously further processes a matchable result set aiming at improving the screening accuracy, performs intersection calculation on the matchable product model sets of various replaced parameters to obtain a final intersection part, namely the replaceable product specification, and can realize more accurate replacement.
Drawings
The invention may be better understood by referring to the following description in conjunction with the accompanying drawings, in which like reference numerals are used throughout the figures to indicate like or similar parts. The accompanying drawings, which are incorporated in and form a part of this specification, illustrate preferred embodiments of the present invention and, together with the detailed description, serve to further explain the principles and advantages of the invention. In the drawings:
FIG. 1 is a parameter comparison diagram of the present invention.
Detailed Description
Embodiments of the present invention will be described below with reference to the accompanying drawings.
The invention relates to a component model searching, matching and replacing method, which comprises the following steps:
step 1: data source sorting: the method comprises the following steps of suggested word arrangement and product result set arrangement;
wherein, the suggested word arrangement comprises: product model, specification model, product brand, product classification, product title, product characteristics, product attributes, product specification;
the product result set arrangement comprises the following steps: product model, specification model, product brand, product category, product title, product characteristics, product attributes, product specification, product content (product drawing, profile drawing, footprint drawing, etc.);
the finishing process comprises the following steps: and carrying out the processes of product shelving, product updating and product deleting in real time to interact with the ES.
Putting the product on shelf: and transmitting a command to the ES to add the product.
Product off shelf/product delete: and transmitting a command to the ES to delete the product.
Product updating: if the product content changes, transmitting a command to the ES, deleting the old document and adding the changed content.
Step 2: real-time matching logic arrangement: the front end transmits the keywords to the rear end, the rear end judges the keywords and transmits the keywords to the ES for searching (the ES performs splitting search processing according to the set splitting rule and the user-defined splitting word) after taking the keywords, and the ES performs sorting display according to the matching degree score in a descending order after searching the result.
Wherein, the splitting rule is as follows:
setting a preset rule, and presetting non-splitting numbers and splitting and recombining letters according to the characteristics of the coding of the components;
setting a self-defined splitting word, and summarizing key words (such as product model, product specification, contact form, leading-out terminal mode parameters and the like) according to the particularity of the components;
setting a user-defined segmentation word, analyzing a search hot word according to big data analysis, and summarizing common search keywords;
and step 3: and logically sorting the search results:
the method comprises the steps that a user inputs keywords, and if search is completed by selecting suggested words, product content is accurately obtained according to the dimensionality to which the selected suggested words belong;
if the input keywords are directly used for searching, the dimension of the input keywords is used for judging which dimension is preferentially searched, and if the input keywords are letters or numbers, the input keywords are used for searching according to the sequential model dimension and the specification model dimension; if the Chinese characters are input, searching according to the parameter dimension of the sequential products; after the dimensions are searched, other dimensions are searched.
The scoring rule of the matching degree is described as follows by taking a relay as an example:
according to the characteristics of the relay, keywords (such as product models, product specifications and the like) are induced, splitting rules are set, the splitting matching rate is improved, and the matching degree is improved;
setting domain weights, setting weights of different domains according to the characteristics of the relay and a common searching mode, and providing the sequence of query sources (such as product small models and product large models);
and setting secondary screening options, and providing further screening for the user so as to narrow the range of the retrieval result.
And according to the rules, obtaining secondary retrieval results with high matching degree according to the screening options selected by the user and the similarity algorithm and the graph similarity algorithm, calculating each secondary retrieval result set, and displaying the retrieval result set with high matching degree finally.
C. (keyword) operation result description:
c1, enter key match (order mark): and (4) result set: a precise query (e.g., search HRM-1H-220V-N directly to retrieve the small model of all brands);
c2, entry keyword matching (large model): and (4) result set: a precise query (e.g., search HRM directly retrieves the model of all brands) + a fuzzy query (e.g., search H or search R or search M or search HR or search RM matches the results of related keywords of all brands;
sorting: the accurate large-scale horn is arranged at the forefront end; (other fuzzy queries) are ranked from high to low according to degree of match; search terms are highlighted in the list results: model, brand, identity.
C3, enter keyword match (not order mark and not large type)
A. And (4) result set:
accurately matching and inputting data with A retrieval value;
supporting fuzzy matching input A to retrieve data containing A;
supporting word segmentation such as user input ab C;
search for data containing A
Retrieving data containing B as B
Retrieving C-containing data
Sorting by similarity, such as including AB before including A;
searching according to the real-time matching range;
large-scale model: large products such as HRM, HF3FF and the like;
order marking: the product is small in size such as HF3FF/012-1HST and the like;
and (4) classification: platform setting classification (multi-stage classification) such as solid state relay, power relay, etc.;
brand name: search results show the brand's goods;
title: such as searching for product titles, and displaying the products containing special sale titles as a result;
parameter values: for example, searching 280 degrees and 12CM, and searching a product with a search result containing the two parameter values;
specification value: for example, the search result needs to include the commodities with the two specification values in an open-close type and a patch type;
supporting classified search such as searching signal relays and searching out products of the category;
brand search is supported, such as a user searches for products of a macro brand;
sorting: sorting according to the matching degree from high to low;
search terms are highlighted in the list results: model, brand, identity.
C4, operation result (search condition + keyword) shows: and filtering the search condition under the keyword result to finally obtain a result.
And 4, step 4: and product comparison, including product parameter comparison and product picture comparison (pin diagram, wiring diagram, appearance diagram and performance curve diagram).
Referring to fig. 1, the product parameter comparison includes product attribute parameter comparison, items with the same attribute parameter value, hidden items with different attribute parameter values but not displayed and with the same attribute item, highlighted items, and product image comparison (pin map, wiring diagram, outline diagram, performance graph) between the product and other products.
The replacement process of the electric device model includes:
step 1: data source sorting:
step 11: suggested word arrangement
Wherein, the suggested word arrangement comprises: product type, specification type, product brand;
the finishing process comprises the following steps: and carrying out the processes of product shelving, product updating and product deleting in real time to interact with the ES.
Putting the product on shelf: and transmitting a command to the ES to add the product.
Product off shelf/product delete: and transmitting a command to the ES to delete the product.
Product updating: if the product content changes, transmitting a command to the ES, deleting the old document and adding the changed content.
Step 12: product result set collation
The product result set arrangement comprises the following steps: product model, specification model, product brand, product characteristics, product content (product drawing, appearance drawing, footprint drawing, etc.);
the finishing process comprises the following steps: and carrying out the processes of product shelving, product updating and product deleting in real time to interact with the ES.
Putting the product on shelf: and transmitting a command to the ES to add the product.
Product off shelf/product delete: and transmitting a command to the ES to delete the product.
Product updating: if the product content changes, transmitting a command to the ES, deleting the old document and adding the changed content.
Model maintenance comparison:
A. newly adding: and transmitting a command to the ES to add new product information.
B. Updating: if the product content changes, transmitting a command to the ES, deleting the old document and adding the changed content.
C. And (3) deleting: and transmitting a command to the ES to delete the product information.
Step 2: and (3) real-time matching logic arrangement: the front end transmits the keywords to the rear end, the rear end judges the keywords and transmits the keywords to the ES for searching (the ES performs splitting search processing according to the set splitting rule and the user-defined splitting word) after taking the keywords, and the ES performs sorting display according to the matching degree score in a descending order after searching the result.
Wherein, the splitting rule is as follows:
setting a preset rule, and presetting non-splitting numbers and splitting and recombining letters according to the characteristics of the coding of the components;
setting a self-defined splitting word, and summarizing key words (such as product model, product specification, contact form, leading-out terminal mode parameters and the like) according to the particularity of the components;
setting a user-defined segmentation word, analyzing a search hot word according to big data analysis, and summarizing common search keywords;
and step 3: and logically sorting the matching result:
step 31: the front end transmits the keywords into the rear end, and the rear end transmits the keywords into the ES for retrieval to confirm the input model; the method specifically comprises the following steps: obtaining the pin map information of the key words, and searching other model information of the same pin map; retrieving and returning according to the keyword to model maintenance contrast;
step 32: unique product search validation is performed according to the following logic:
entering keywords to search and confirm unique products:
A. if the brand is a single large model, the unique product is confirmed.
B. If the product is multi-brand, the brand needs to be supplemented, and the only product can be confirmed.
C. If the content of the keyword portion exactly matches the model and is a single brand, the unique product can be confirmed.
D. If the partial content of the keyword is matched with the model and is multi-brand model, the brand needs to be complemented, and the unique product can be confirmed.
E. And if the model information of the keywords cannot be confirmed or the keywords do not have matchable information, prompting the user.
By the selected keyword: the unique product is directly validated.
And 3, acquiring a pin bitmap of the unique product, and searching and comparing a model product information result set A of the same pin bitmap according to the pin bitmap.
Special result set: and transmitting the confirmed specific model to model maintenance contrast on the ES to retrieve a matchable model information result set B.
And 5, after the duplication of the result set A and the result set B is removed, the result set is a result set which can be matched.
If a specific product is desired to be identified, further replacement operations may be performed by selecting an order label.
And 4, step 4: and logically sorting replacement results:
step 41: the front end transmits the keywords into the rear end, the rear end transmits the keywords into the ES for retrieval, and the input type is confirmed to be an order mark, and the method specifically comprises the following steps: obtaining the pin map information of the key words, and searching other model information of the same pin map; and retrieving and returning according to the keyword-model maintenance contrast.
Step 42: unique product search validation is performed according to the following logic:
entering keywords to search and confirm unique products:
A. if a single brand order is marked, the unique product is identified.
B. If the order mark of multiple brands is used, the brand needs to be supplemented, and the only product can be confirmed.
C. If it is the keyword portion content that exactly matches the order label and is a single brand, then the unique product can be identified.
D. If the partial content of the keyword exactly matches the order mark and the order mark of multiple brands needs to be supplemented with the brands, the only product can be confirmed.
E. If the order marking information of the keywords cannot be confirmed or the keywords have no matchable information, prompting the user.
By the selected keyword: the unique product is directly validated.
And 3, acquiring a pin bitmap of the unique product, and searching and comparing a model product information result set A of the same pin bitmap according to the pin bitmap.
Acquiring the standard specification of the unique product, screening from the standard specification items to a result set A, screening a result set B completely containing the standard specification of the unique product (the replacement item needs to completely contain the standard specification of the replaced item, namely the replacement item A has 5 standard specifications, the replaced item B has 4 standard specifications, then B can be replaced by A, but A can not be replaced by B)
Special result set: and transmitting the confirmed specific order mark to model maintenance comparison on the ES to retrieve a matchable model information result set C.
And 6, after the duplication of the result set B and the result set C is removed, the result set is an alternative result set.
It should be emphasized that the term "comprises/comprising" when used herein, is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.
In addition, the method of the present invention is not limited to be performed in the time sequence described in the specification, and may be performed in other time sequences, in parallel, or independently. Therefore, the order of execution of the methods described in this specification does not limit the technical scope of the present invention.
While the present invention has been disclosed above by the description of specific embodiments thereof, it should be understood that all of the embodiments and examples described above are illustrative and not restrictive. Various modifications, improvements and equivalents of the invention may be devised by those skilled in the art within the spirit and scope of the appended claims. Such modifications, improvements and equivalents are also intended to be included within the scope of the present invention.
Claims (1)
1. A component model matching method is characterized in that: the method comprises the following steps:
step 1: splitting and sorting the keywords input by the user according to a preset splitting rule;
step 2: searching the split and sorted keywords in a data source by using an ES (ES);
and step 3: displaying the retrieval result of the ES according to the matching degree score, recording a result set with the score larger than the preset matching degree in the retrieval result as a search matching result set, and finishing the search;
and 4, step 4: performing secondary screening on the search matching result set obtained in the step 3;
and 5: a replacing step, comprising:
step 51: searching the keywords input by the user in a data source by using an ES (identity extraction) and searching a model which is completely matched and confirmed to be input according to a similarity algorithm;
step 52: confirming the unique product according to the retrieval result, acquiring a pin bit map and necessary replacement parameters of the unique product, searching and comparing a model product information result set of the same pin bit map according to the pin bit map, and performing retrieval matching on the result set according to the necessary replacement parameters to obtain a model product information result set A; transmitting the confirmed specific model to model maintenance contrast on the ES to retrieve a matchable model information result set B; after the duplication of the result set A and the result set B is removed, the result set A and the result set B are matched;
in step 3, scoring the search result of the ES according to the matching degree specifically includes:
step 31: according to the common search content, the search content is summarized and sorted into different attribute domains; the attribute domain at least comprises a model domain and a parameter domain, wherein the model domain is formed by processing and arranging a large model of a product and an order mark; the parameter domain is formed by processing and sorting parameters of a contact form and a leading-out terminal form;
step 32: setting a weight proportion according to a preset strategy aiming at the model domain and the parameter domain, wherein the weight proportion is preset as follows: model domain > parameter domain > other domains, the other domains being domains in the attribute domain other than the model domain and the parameter domain; the preset strategy is as follows: when the retrieved attribute is a model type, the weight ratio is set as: model field > parameter field > other fields, the priority to meet the condition is ranked to the forefront; when the retrieved attribute is a parameter class, the weight ratio is set as: parameter domain > model domain > other domains, the ones that will satisfy the condition are arranged to the front end; when the retrieved attributes are other domains, automatically calculating scores according to all the documents;
step 33: receiving a keyword of the user, and if the keyword is a suggested word of the system, performing step 331; if the keyword is a user-defined keyword, step 332:
step 331: acquiring the suggested word and the dimensionality of the suggested word, directly and completely matching the suggested word, and accurately obtaining a retrieval result;
step 332: judging whether the type of the key word is letter, number or Chinese character according to the regular expression,
if the type of the keyword is a letter or a number, preferentially retrieving a matched model domain according to the weight proportion of the model domain in the step 32, and obtaining a matching score according to the score calculated by the similarity algorithm and the weight proportion score; wherein, the weight proportion score (weight/proportion base) is scored, and the weight and the proportion base are self-defined numerical values;
if the type of the keyword is Chinese character, preferentially searching the matched parameter domain according to the weight proportion of the parameter domain in the step 32, and obtaining a matching score according to the score calculated by the similarity algorithm and the weight proportion score;
then carrying out fuzzy search on other query sources, and calculating scores according to a similarity algorithm;
the step 4 comprises the following processes:
acquiring a condition item selected by a user, obtaining a condition item value according to a similarity algorithm by a non-image condition item, obtaining a condition item value according to an image similarity algorithm after scanning the condition item of the image, performing retrieval matching on the condition item value in a search matching result set to obtain a secondary search matching result set correspondingly matched with each condition item value, performing intersection calculation on each secondary search matching result set to obtain a final intersection part, namely a final retrieval matching result set, and finishing the search;
in the step 4, the condition items are screening items defined according to the characteristics of the components, and the screening items comprise pin numbers, contact forms, coil voltages, rated loads, leading-out terminal modes and pin position diagrams;
the splitting rule in the step 1 is as follows:
setting a preset rule, and presetting non-splitting numbers and splitting and recombining letters according to the characteristics of the coding of the components;
setting a self-defined segmentation word, and summarizing key words according to the attributes of the components;
setting a user-defined segmentation word, analyzing and searching a hot word according to big data, and summarizing common search keywords;
judging whether the keyword has a blank space or other special symbols:
if not, performing retrieval;
if the search result contains a space or other special symbols, removing the space or other special symbols to form a new keyword, and then executing the search;
step 2 also comprises the step of pre-sorting the data sources before retrieval;
the method for pre-sorting the data sources specifically comprises the following steps: the attributes describing the same component are uniformized in advance, so that the values with the same meaning have a uniform form; projecting the numerical value attributes of the components into a specific small range in proportion in advance to eliminate the deviation of the retrieval result caused by the different sizes of the numerical value attributes;
the step 51 specifically includes: obtaining a pin diagram of a keyword, namely taking a current pin diagram as a retrieval diagram, obtaining a pin diagram with high matching degree according to an SIFT algorithm, extracting and recording the pin diagram into a pin diagram set, and obtaining a matched product model information set according to the pin diagram set;
the step 51 of obtaining the matchable product model information set according to the pin map set specifically includes: obtaining the pin numbers of the retrieval model and other matchable product models, retrieving the pin numbers of the model and the other matchable product models according to a formula, rejecting the product models with the pin numbers of the retrieval model which are more than the pin numbers of the matchable product models, and obtaining matchable product model information sets;
in step 52, for the result set that can be matched, the following operations are also performed: and obtaining the external dimensions of the retrieval model and other matchable product models, converting the external dimensions into a unified unit, performing difference calculation by using the external dimensions of the retrieval model and other matchable product models, and when the tolerance meets a preset interval, determining that the external dimensions are in a matching range, and determining that the external dimensions are out of the range and cannot be matched.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910926938.9A CN110674384B (en) | 2019-09-27 | 2019-09-27 | Component model matching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910926938.9A CN110674384B (en) | 2019-09-27 | 2019-09-27 | Component model matching method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110674384A CN110674384A (en) | 2020-01-10 |
| CN110674384B true CN110674384B (en) | 2022-03-11 |
Family
ID=69079582
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910926938.9A Active CN110674384B (en) | 2019-09-27 | 2019-09-27 | Component model matching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110674384B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112309511B (en) * | 2020-09-18 | 2023-04-14 | 青岛檬豆网络科技有限公司 | Parameter decomposition and purchase prediction method for tantalum electrolytic capacitor |
| CN112199451B (en) * | 2020-09-30 | 2024-07-16 | 京东科技控股股份有限公司 | Commodity identification method, commodity identification device, computer equipment and storage medium |
| CN112580685B (en) * | 2020-11-19 | 2022-10-28 | 青岛檬豆网络科技有限公司 | MLCC (multilayer ceramic capacitor) capacitor parameter matching method |
| CN112580686B (en) * | 2020-11-19 | 2023-05-02 | 青岛檬豆网络科技有限公司 | KNN algorithm aluminum electrolytic capacitor purchase prediction method based on Mahalanobis distance |
| CN112579629A (en) * | 2020-12-10 | 2021-03-30 | 青岛檬豆网络科技有限公司 | Method for helping purchasers of electronic component enterprises to accurately find products |
| CN113626561A (en) * | 2021-08-16 | 2021-11-09 | 深圳市云采网络科技有限公司 | Component model identification method, device, medium and equipment |
| CN114090750A (en) * | 2021-11-10 | 2022-02-25 | 深圳前海硬之城信息技术有限公司 | Method and equipment for quickly and accurately matching component models and storage medium |
| CN114912002A (en) * | 2022-05-30 | 2022-08-16 | 深圳市云采网络科技有限公司 | Electronic component searching method and device, electronic equipment and storage medium |
| CN114969457A (en) * | 2022-06-17 | 2022-08-30 | 珠海格力电器股份有限公司 | Product retrieval method, product retrieval device, electronic equipment and readable medium |
| CN117609529B (en) * | 2023-10-20 | 2024-09-10 | 粤港澳大湾区(广东)国创中心 | Substitution search method and system for electronic components |
| CN117874309B (en) * | 2024-03-12 | 2024-05-24 | 北京全路通信信号研究设计院集团有限公司 | Train control data processing method and device, electronic equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9565275B2 (en) * | 2012-02-09 | 2017-02-07 | Rockwell Automation Technologies, Inc. | Transformation of industrial data into useful cloud information |
| CN103593478A (en) * | 2013-11-28 | 2014-02-19 | 广东工业大学 | Resource retrieval method and device of electronic component |
| CN108062411A (en) * | 2017-12-29 | 2018-05-22 | 深圳市智搜信息技术有限公司 | A kind of system and method for searching electronic component data message |
| CN109241017A (en) * | 2018-10-23 | 2019-01-18 | 电子科技大学 | A kind of creation in PCB component package library and querying method |
-
2019
- 2019-09-27 CN CN201910926938.9A patent/CN110674384B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110674384A (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110674384B (en) | Component model matching method | |
| US6212532B1 (en) | Text categorization toolkit | |
| CN107085583B (en) | Electronic document management method and device based on content | |
| KR100627195B1 (en) | System and method for searching electronic documents created with optical character recognition | |
| CN103106199A (en) | Text retrieval method and test retrieval device | |
| CN107665217A (en) | A kind of vocabulary processing method and system for searching service | |
| US20230147941A1 (en) | Method, apparatus and device used to search for content | |
| CN111428494A (en) | Intelligent error correction method, device and equipment for proper nouns and storage medium | |
| WO2014097670A1 (en) | Document classification device and program | |
| CN106909669B (en) | Method and device for detecting promotion information | |
| US20130031083A1 (en) | Determining keyword for a form page | |
| CN101138001A (en) | Learning processing method, learning processing device, and program | |
| CN110866408B (en) | Database creation device and search system | |
| CN112395881B (en) | Material label construction method and device, readable storage medium and electronic equipment | |
| CN106997390A (en) | A kind of equipment part or parts commodity transaction information search method | |
| CN103218420A (en) | Method and device for extracting page titles | |
| CN110019637B (en) | Sorting algorithm for standard document retrieval | |
| US20100063966A1 (en) | Method for fast de-duplication of a set of documents or a set of data contained in a file | |
| JP2013029891A (en) | Extraction program, extraction method and extraction apparatus | |
| CN111538839A (en) | Real-time text clustering method based on Jacobsard distance | |
| CN111814791A (en) | Method and device for identifying components in system diagram | |
| JP7126843B2 (en) | Learning target extraction device, learning target extraction method, and learning target extraction program | |
| CN114511027B (en) | Method for extracting English remote data through big data network | |
| WO2021241601A1 (en) | Information retrieval system | |
| CN117648616A (en) | Article classification method, apparatus, computer device 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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 361000 unit 201, No. 1708, Gangzhong Road, Xiamen area, China (Fujian) pilot Free Trade Zone, Xiamen, Fujian Applicant after: Xiamen Jingxin Electronics Co.,Ltd. Address before: No. 0211, unit 109, 62 Chengyi North Street, phase III, software park, Xiamen City, Fujian Province, 361000 Applicant before: XIAMEN JINGXIN ELECTRONIC Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20211015 Address after: 361000 unit 101, No. 1708, Gangzhong Road, Xiamen area, China (Fujian) pilot Free Trade Zone, Xiamen, Fujian Applicant after: Xiamen Ziyuan Information Technology Co.,Ltd. Address before: 361000 unit 201, No. 1708, Gangzhong Road, Xiamen area, China (Fujian) pilot Free Trade Zone, Xiamen, Fujian Applicant before: Xiamen Jingxin Electronics Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |