CN-121996830-A - Article searching method, article searching device and electronic equipment

CN121996830ACN 121996830 ACN121996830 ACN 121996830ACN-121996830-A

Abstract

The embodiment of the disclosure discloses an article retrieval method, an article retrieval device and electronic equipment. The method comprises the steps of responding to a received article retrieval request, generating a request vector, generating retrieval attribute data according to retrieval conditions in the article retrieval request, retrieving an article vector set under a target leaf node matched with the request vector from a pre-constructed article vector tree according to the request vector, and retrieving a recall article set matched with the retrieval attribute data from articles indicated by the article vector set according to description attribute data of the articles so as to carry out screening processing on a service layer. This embodiment is related to information retrieval technology, and may integrate some filtering rules of the business layer into the vector recall phase of the engine layer in advance. Therefore, invalid commodity advertisements which do not meet the demands of users can be filtered in advance, so that the data quantity of invalid data required to be processed by a subsequent business layer is reduced, and the overall retrieval efficiency is improved.

Inventors

ZHU LIMING
HUANG PEI
QIN YINUO
LI SEN

Assignees

北京沃东天骏信息技术有限公司

Dates

Publication Date: 20260508
Application Date: 20241107

Claims (13)

1. An article retrieval method comprising: Generating a request vector in response to receiving an article retrieval request, and generating retrieval attribute data according to retrieval conditions in the article retrieval request, wherein the request vector and the retrieval attribute data characterize different data information; According to the request vector, retrieving an article vector set under a target leaf node matched with the request vector from a pre-constructed article vector tree, wherein the article vector tree is obtained by clustering article vectors of articles; And retrieving a recall item set matched with the retrieval attribute data from the items indicated by the item vector set according to the description attribute data of the items so as to carry out screening processing on a business layer.
2. The item retrieval method according to claim 1, wherein the generating retrieval attribute data according to the retrieval condition in the item retrieval request includes: Generating a Boolean expression according to the retrieval conditions in the article retrieval request; for each search condition in the Boolean expression, calculating a hash value of the search condition according to a hash algorithm of a bloom filter to obtain a bit number set of the search condition; And taking the bit array of each search condition in the Boolean expression as the search attribute data of the article search request.
3. The item retrieval method according to claim 2, wherein retrieving a recalled item set matching the retrieved attribute data from the items indicated by the item vector set according to the item description attribute data, comprises: For each item indicated by the item vector set, determining a bit array of descriptive attribute data of each item according to the bloom filter; Matching the bit number group of the description attribute data of each article with the bit number group of the retrieval attribute data; and determining the item indicated by the bit array of the description attribute data matched with the bit array of each retrieval condition in the retrieval attribute data as a recall item, and obtaining a recall item set.
4. The method of item retrieval as recited in claim 1, wherein said method further comprises: constructing a bloom filter object corresponding to the article to store and retrieve description attribute data of the article; and calculating the hash value of the description attribute data of the article, and storing the calculated hash value into a corresponding bloom filter to obtain a bit array of the description attribute data of the article.
5. The method of item retrieval as recited in claim 1, wherein said method further comprises: Carrying out standardized processing on attribute information of the article according to preset standard attributes; and generating description attribute data in an article preset format according to the standardized attributes and the corresponding attribute values.
6. The method of item retrieval as recited in claim 4, wherein said method further comprises: In response to detecting the item data change, acquiring data in the bloom filter which is latest stored with the description attribute data of the item, and acquiring a latest item vector tree to be stored in the memory; For the article indicated by the article vector under each leaf node in the latest article vector tree, determining the information of the bloom filter object corresponding to the article, and storing the information under the leaf node to which the article belongs.
7. The item retrieval method as recited in claim 6, wherein said obtaining data in a bloom filter that has recently stored descriptive attribute data of an item, and obtaining a latest item vector tree, in response to detecting an item data change, comprises: Responding to the detection of the version information change of the index file, and obtaining the index file of the latest version, wherein the index file is obtained by adopting a preset data description language to carry out serialization processing under the condition that the construction of the article vector tree and the bloom filter is completed; And performing deserialization processing on the acquired index file by adopting the preset data description language to acquire bloom filter data and an article vector tree in the file.
8. The item retrieval method as recited in claim 1, wherein said item vector tree is constructed by: Acquiring related information of an object on a platform, and generating an object vector, wherein the related information comprises at least one item selected from text title information in an object display page, identification information of a manufacturer and an object picture; Carrying out clustering analysis on the article vectors by adopting a clustering algorithm until the clustering ending condition is met, and obtaining an article vector tree; Wherein for a node in the item vector tree, an average value of item vectors of items belonging to the node is taken as a node vector of the node, and item identifications of items belonging to the leaf node are stored under each leaf node.
9. The article retrieval method according to one of claims 1 to 8, wherein the method further comprises: And screening the recall items in the recall item set according to the item retrieval request and a preset business rule, and taking the processing result as a retrieval result of the item retrieval request.
10. An article retrieval device comprising: A request data generation unit configured to generate a request vector in response to receiving an article retrieval request, and generate retrieval attribute data according to a retrieval condition in the article retrieval request, wherein the request vector characterizes different data information from the retrieval attribute data; A vector tree retrieving unit configured to retrieve, from a pre-constructed article vector tree, an article vector set under a target leaf node matched with the request vector according to the request vector, wherein the article vector tree is obtained by clustering article vectors of articles; And the attribute retrieval unit is configured to retrieve a recall item set matched with the retrieval attribute data from the items indicated by the item vector set according to the description attribute data of the items so as to be subjected to screening processing by a business layer.
11. An electronic device, comprising: One or more processors; a storage device having one or more programs stored thereon, The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the article retrieval method of any of claims 1-9.
12. A computer readable medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the article retrieval method according to any of claims 1-9.
13. A computer program product comprising a computer program which, when executed by a processor, implements the article retrieval method of any one of claims 1-9.

Description

Article searching method, article searching device and electronic equipment Technical Field The embodiment of the disclosure relates to the technical field of information retrieval, in particular to an article retrieval method, an article retrieval device and electronic equipment. Background In the architecture of an advertisement system, an advertisement retrieval platform often plays a critical role, and is generally positioned at the upstream of the advertisement system, so as to finish the preliminary matching of users to advertisement materials. And screening out hundred-level advertisement sets matched with the interests of the user from the massive advertisement material library in real time for further screening and sorting in downstream links. The key capabilities of the relevant advertisement retrieval platform are mainly focused on the following two core modules: 1. The engine layer recalls that the layer is responsible for selecting tens of thousands of advertisement sets associated with the interests of the user from a mass of advertisement materials by utilizing various recall strategies. 2. And (3) filtering and coarse ordering the service layer, namely filtering the recalled advertisement set by further service rules, and ensuring that the selected advertisement accords with the rule constraint under the specific service scene. However, the inventors have found that since the engine layer recall link fails to adequately account for business rule constraints in a variety of scenarios, its recalled advertisement sets are typically filtered in large amounts when the business layer performs business rule filtering, resulting in a reduction in the number of results ultimately returned. This also affects the retrieval efficiency of the advertisement retrieval platform. The above information disclosed in this background section is only for enhancement of understanding of the background of the inventive concept and, therefore, may contain information that does not form the prior art that is already known to those of ordinary skill in the art in this country. Disclosure of Invention The disclosure is in part intended to introduce concepts in a simplified form that are further described below in the detailed description. The disclosure is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Some embodiments of the present disclosure propose a gait feature determination method, a gait feature determination apparatus, an electronic device, a computer readable medium and a computer program product to solve one or more of the technical problems mentioned in the background section above. In a first aspect, some embodiments of the present disclosure provide an article retrieval method, which includes generating a request vector in response to receiving an article retrieval request, and generating retrieval attribute data according to a retrieval condition in the article retrieval request, wherein the request vector and the retrieval attribute data represent different data information, retrieving an article vector set under a target leaf node matched with the request vector from a pre-constructed article vector tree according to the request vector, wherein the article vector tree is obtained by clustering article vectors of articles, and retrieving a recall article set matched with the retrieval attribute data from articles indicated by the article vector set according to description attribute data of the articles, so as to perform screening processing on a service layer. In some embodiments, the generation of the search attribute data according to the search conditions in the article search request comprises the steps of generating a Boolean expression according to the search conditions in the article search request, calculating hash values of the search conditions according to a hash algorithm of a bloom filter for each search condition in the Boolean expression to obtain a bit array of the search condition, and taking the bit array of each search condition in the Boolean expression as the search attribute data of the article search request. In some embodiments, retrieving a recall item set matched with the retrieve attribute data from items indicated by the item vector set according to the description attribute data of the items includes determining, for each item indicated by the item vector set, a bit array of the description attribute data of each item according to a bloom filter, matching the bit array of the description attribute data of each item with the bit array of the retrieve attribute data, and determining, as the recall item, the item indicated by the bit array of the description attribute data matching the bit array of each retrieve condition in the retrieve attribute data, to obtain the recall item set. In some embodiments, the method further comprises constructing a bloom f