WO2020168839A1 - 物品召回方法、系统、电子设备及可读存储介质 - Google Patents

物品召回方法、系统、电子设备及可读存储介质 Download PDF

Info

Publication number
WO2020168839A1
WO2020168839A1 PCT/CN2020/070014 CN2020070014W WO2020168839A1 WO 2020168839 A1 WO2020168839 A1 WO 2020168839A1 CN 2020070014 W CN2020070014 W CN 2020070014W WO 2020168839 A1 WO2020168839 A1 WO 2020168839A1
Authority
WO
WIPO (PCT)
Prior art keywords
item
words
search
word
module
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.)
Ceased
Application number
PCT/CN2020/070014
Other languages
English (en)
French (fr)
Inventor
胡忆桐
高云
王娜
左丽丽
张亚红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Jingdong Century Trading Co Ltd
Beijing Jingdong Shangke Information Technology Co Ltd
Original Assignee
Beijing Jingdong Century Trading Co Ltd
Beijing Jingdong Shangke Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Jingdong Century Trading Co Ltd, Beijing Jingdong Shangke Information Technology Co Ltd filed Critical Beijing Jingdong Century Trading Co Ltd
Priority to US17/296,159 priority Critical patent/US11907659B2/en
Publication of WO2020168839A1 publication Critical patent/WO2020168839A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3347—Query execution using vector based model
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3325—Reformulation based on results of preceding query
    • G06F16/3326—Reformulation based on results of preceding query using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/01—Customer relationship services
    • G06Q30/014—Providing recall services for goods or products
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3438—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment monitoring of user actions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures

Definitions

  • the invention belongs to the field of big data processing, and in particular relates to an item recall method, system, electronic equipment and readable storage medium.
  • the technical problem to be solved by the present invention is to overcome the defect in the prior art that the item recall effect is poor due to the description difference, and provide an item recall method, system, electronic device and readable storage medium.
  • An item recall method includes:
  • Obtain user behavior data including clicked and/or purchased items and item information of each item;
  • the step of obtaining the retrieval term of each item in the target behavior data and the retrieval frequency of each retrieval term in the reverse association specifically includes:
  • the retrieval terms of each item in the target behavior data and the retrieval frequency of each retrieval term are acquired based on the relationship list.
  • the step of calculating the similarity between all search terms and the multiple article segmentation specifically includes:
  • the similarity between the search term set and the item segmentation set is calculated based on the Jaccard algorithm (an algorithm used to compare the similarity and difference between sample sets).
  • the item recall method before the step of recalling an item based on the search category word and the expansion word, the item recall method further includes:
  • Training the corpus based on word2vec (a model for generating word vectors) to obtain a word vector database, the word vector database storing a plurality of item feature words and a feature vector corresponding to each item feature word;
  • the item recall method before the step of recalling an item based on the search category word and the expansion word, the item recall method further includes:
  • Preset an item category list including multiple item categories and the attributes of all items under each item category;
  • the step of performing item recall based on the search category words and the expansion words specifically includes:
  • An electronic device includes a memory, a processor, and a computer program that is stored on the memory and can run on the processor.
  • the processor implements the above-mentioned item recall method when the computer program is executed.
  • a readable storage medium has a computer program stored thereon, and when the program is executed by a processor, the steps of the above item recall method are realized.
  • An item recall system which includes a behavior data acquisition module, an input module, a data extraction module, a search term acquisition module, a word segmentation module, a first calculation module, a first judgment module, an extended word generation module, and a recall module;
  • the behavioral data acquisition module is used to acquire behavioral data of the user, and the behavioral data includes items clicked and/or purchased and item information of each item;
  • the input module is used to input a search category word
  • the data extraction module is used to extract target behavior data containing the search category words from the behavior data
  • the retrieval term acquisition module is used to reversely correlate retrieval terms of each item in the target behavior data and retrieval frequency of each retrieval term;
  • the word segmentation module is used to segment the item information in the target behavior data to obtain multiple item segmentation
  • the first calculation module is used to calculate the similarity between all search terms and the multiple article segmentation
  • the first judgment module is used to judge whether the similarity is greater than a first preset threshold, and if so, call the extended word generation module;
  • the extended word generation module is used to extract the search words with a search frequency greater than a preset frequency among all the search words as the expansion words of the search category words;
  • the recall module is used for recalling items according to the search category words and the expansion words.
  • the search term acquisition module is used to generate a relationship list based on the behavior data, and the relationship list is used to characterize the corresponding relationship between the search term and the clicked and/or purchased item, and obtain all the information based on the relationship list. Describe the search terms of each item in the target behavior data and the search frequency of each search term.
  • the first calculation module includes a set generation unit and a calculation unit;
  • the set generating unit is used to de-duplicate all search words to generate a search word set, and is also used to de-duplicate the multiple article word segments to generate an article word set;
  • the calculation unit is configured to calculate the similarity between the search term set and the article word segmentation set based on the Jaccard algorithm.
  • the item recall system further includes a corpus building module, a word vector library generating module, and a second calculation module;
  • the corpus building module is used to construct a corpus, and the corpus includes article recommendation articles and basic information of the article;
  • the word vector database generation module is used to train the corpus based on word2vec to obtain a word vector database, and the word vector database stores multiple item feature words and feature vectors corresponding to each item feature word;
  • the second calculation module is configured to calculate the similarity between the search category word and each item feature word based on a similarity algorithm
  • the extended word generation module is also used to extract item feature words whose similarity with the search category words is greater than a second preset threshold to add to the extended words.
  • the item recall system further includes a second judgment module, a filtering module, and an item category list, and the item category list includes multiple item categories and attributes of all items under each item category;
  • the second judgment module is used to judge whether the extension word is included in the item category list, and if so, call the filtering module;
  • the filtering module is used to filter out the expansion words included in the category list
  • the recall module is used to perform item recall based on the filtered expanded words and the search category words.
  • the positive and progressive effect of the present invention is that when the present invention uses category words for item recall, it uses the acquired behavior data to extract the expanded words of item category words, which improves the coverage of recalled items while ensuring the accuracy of the recall. .
  • Fig. 1 is a flowchart of an item recall method according to Embodiment 1 of the present invention.
  • step 60 is a flowchart of step 60 in the item recall method of Embodiment 1 of the present invention.
  • Fig. 3 is a flowchart of an item recall method according to Embodiment 2 of the present invention.
  • FIG. 4 is a schematic structural diagram of an electronic device according to Embodiment 3 of the present invention.
  • FIG. 5 is a schematic diagram of modules of an item recall system according to Embodiment 5 of the present invention.
  • FIG. 6 is a schematic diagram of modules of an item recall system according to Embodiment 6 of the present invention.
  • An item recall method as shown in Figure 1, the item recall method includes:
  • Step 10 Obtain the user's behavior data; the behavior data includes the items clicked and/or purchased and the item information of each item;
  • Step 20 Input a search category word
  • Step 30 Extract target behavior data including search category words from the behavior data
  • Step 40 Obtain the retrieval term of each item in the target behavior data and the retrieval frequency of each retrieval term in reverse association;
  • Step 50 Perform word segmentation on the item information in the target behavior data to obtain multiple item segmentation
  • Step 60 Calculate the similarity between all search terms and multiple item segmentation
  • Step 70 Judge whether the similarity is greater than the first preset threshold, if yes, go to step 80; if not, it means that none of the search terms can be used as extension words of the search category words;
  • Step 80 Extract the search terms whose search frequency is greater than the preset frequency among all search terms as extension words of the search category words;
  • Step 90 Perform item recall based on the search category words and expansion words.
  • step 40 specifically includes:
  • a relationship list is generated according to the behavior data, and the relationship list is used to characterize the corresponding relationship between search terms and clicked and/or purchased items, and based on the relationship list, the search terms and search terms of each item in the target behavior data are obtained.
  • the search frequency of each search term is generated according to the behavior data, and the relationship list is used to characterize the corresponding relationship between search terms and clicked and/or purchased items, and based on the relationship list, the search terms and search terms of each item in the target behavior data are obtained.
  • the search frequency of each search term is generated according to the behavior data, and the relationship list is used to characterize the corresponding relationship between search terms and clicked and/or purchased items, and based on the relationship list, the search terms and search terms of each item in the target behavior data are obtained.
  • step 60 specifically includes:
  • Step 601 De-duplicate all search words to generate a search word set
  • Step 602 De-regenerate multiple item segmentation to generate an item segmentation set
  • Step 603 Calculate the similarity between the search term set and the item segmentation set based on the Jaccard algorithm.
  • the acquired behavior data is used to extract the expansion words of the item category words, which improves the coverage of the recalled items while ensuring the accuracy of the recall.
  • the item recall method of this embodiment is further improved on the basis of Embodiment 1. As shown in FIG. 3, before step 90, the item recall method further includes:
  • Step 81 Construct a corpus, which includes article recommendation articles and basic information about articles;
  • Step 82 Train the corpus based on word2vec to obtain a word vector library; the word vector library stores multiple item feature words and feature vectors corresponding to each item feature word;
  • Step 83 Calculate the similarity between the search category words and the feature words of each item based on the similarity algorithm
  • Step 84 Extract item feature words whose similarity with the search category words is greater than a second preset threshold and add them to extension words.
  • the extension words obtained based on the above method can easily be extended to adjacent categories or higher-level categories or other existing categories, for example, in the extension words of "beef” There is “chicken”, the expansion word of banana has “pineapple”, the expansion word of "computer desk” has “table”, if you need to recall the items corresponding to the existing category words, you only need to select the corresponding existing category Item recall can be carried out with the target words. In this application, these expanded words can be filtered through the existing category list. See Figure 3. Before step 90, the item recall method also includes:
  • Step 85 Preset an item category list, which includes multiple item categories and attributes of all items under each item category;
  • Step 86 Judge whether the extended word is included in the item category list, if yes, go to step 87;
  • Step 87 Filter out the extended words included in the category list
  • step 90 specifically includes:
  • article recommendation articles and article names of various platforms are used as corpus to train the word2vec model, and based on the trained model, feature words and corresponding word vectors are generated, and then based on similarity calculations, several words with the highest similarity are obtained.
  • the expansion word of the search category word it complements the expansion of the search category word using the behavior data process described above, and further increases the coverage of the recall.
  • An electronic device including a memory, a processor, and a computer program stored in the memory and capable of running on the processor.
  • the processor implements the computer program described in any one of Embodiments 1 or 2 when the processor executes the computer program. Item recall method.
  • FIG. 4 is a schematic structural diagram of an electronic device provided by this embodiment.
  • Figure 4 shows a block diagram of an exemplary electronic device 90 suitable for implementing embodiments of the present invention.
  • the electronic device 90 shown in FIG. 4 is only an example, and should not bring any limitation to the function and application scope of the embodiment of the present invention.
  • the electronic device 90 may be in the form of a general-purpose computing device, for example, it may be a server device.
  • the components of the electronic device 90 may include but are not limited to: at least one processor 91, at least one memory 92, and a bus 93 connecting different system components (including the memory 92 and the processor 91).
  • the bus 93 includes a data bus, an address bus, and a control bus.
  • the memory 92 may include a volatile memory, such as a random access memory (RAM) 921 and/or a cache memory 922, and may further include a read-only memory (ROM) 923.
  • RAM random access memory
  • ROM read-only memory
  • the memory 92 may also include a program tool 925 having a set (at least one) program module 924.
  • program module 924 includes but is not limited to: an operating system, one or more application programs, other program modules, and program data. In these examples Each or some combination of may include the realization of the network environment.
  • the processor 91 executes various functional applications and data processing by running a computer program stored in the memory 92.
  • the electronic device 90 may also communicate with one or more external devices 94 (such as keyboards, pointing devices, etc.). This communication can be performed through an input/output (I/O) interface 95.
  • the electronic device 90 may also communicate with one or more networks (for example, a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet) through the network adapter 96.
  • the network adapter 96 communicates with other modules of the electronic device 90 through the bus 93.
  • a computer-readable storage medium has a computer program stored thereon, and when the program is executed by a processor, the steps of the item recall method described in any one of Embodiments 1 or 2 are realized.
  • the readable storage medium may more specifically include but not limited to: portable disk, hard disk, random access memory, read only memory, erasable programmable read only memory, optical storage device, magnetic storage device or any of the above The right combination.
  • the present invention can also be implemented in the form of a program product, which includes program code.
  • program product runs on a terminal device
  • the program code is used to make the terminal device execute the implementation. Steps of the item recall method described in any one of Embodiments 1 or 2.
  • program code used to execute the present invention can be written in any combination of one or more programming languages, and the program code can be completely executed on the user equipment, partially executed on the user equipment, as an independent
  • the software package is executed, partly on the user’s device, partly on the remote device, or entirely on the remote device.
  • the item recall system includes a behavior data acquisition module 11, an input module 12, a data extraction module 13, a search term acquisition module 14, a word segmentation module 15, a first calculation module 16, and a first A judgment module 17, an expanded word generation module 18, and a recall module 19;
  • the behavior data acquisition module 11 is used to acquire behavior data of a user, the behavior data including items clicked and/or purchased and item information of each item;
  • the input module 12 is used to input a search category word
  • the data extraction module 13 is configured to extract target behavior data containing the search category words from the behavior data
  • the search term acquisition module 14 is used to reversely correlate the search terms of each item in the target behavior data and the search frequency of each search term; specifically, the search term acquisition module 14 is used to obtain the search terms according to the behavior data
  • a relationship list is generated, the relationship list is used to characterize the corresponding relationship between search terms and clicked and/or purchased items, and the search terms of each item in the target behavior data and the search terms of each search word are obtained based on the relationship list. Frequency of retrieval.
  • the word segmentation module 15 is used to segment the item information in the target behavior data to obtain multiple item segmentation;
  • the first calculation module 16 is used to calculate the similarity between all search terms and the multiple article segmentation
  • the first calculation module 16 includes a set generation unit 161 and a calculation unit 162;
  • the set generating unit 161 is used to de-duplicate all search words to generate a search term set, and is also used to de-duplicate the multiple article word segments to generate an article word set;
  • the calculation unit 162 is configured to calculate the similarity between the search term set and the article segmentation set based on the Jaccard algorithm.
  • the first judgment module 17 is used to judge whether the similarity is greater than a first preset threshold, and if so, call the extension word generation module 18;
  • the extended word generation module 18 is configured to extract the search words whose retrieval frequency is greater than the preset frequency among all the search words as the expansion words of the search category words;
  • the recall module 19 is used for recalling items according to the search category words and the expansion words.
  • the acquired behavior data is used to extract the expansion words of the item category words, which improves the coverage of the recalled items while ensuring the accuracy of the recall.
  • the item recall system of this embodiment is further improved on the basis of Embodiment 5.
  • the item recall system further includes a corpus construction module 21, a word vector library generation module 22, and a second calculation module 23;
  • the corpus construction module 21 is used to construct a corpus, and the corpus includes article recommendation articles and basic information of the article;
  • the word vector database generating module 22 is configured to train the corpus based on word2vec to obtain a word vector database, and the word vector database stores a plurality of item feature words and a feature vector corresponding to each item feature word;
  • the second calculation module 23 is configured to calculate the similarity between the search category word and each item feature word based on a similarity algorithm
  • the expanded word generating module 18 is also used to extract the item feature words whose similarity with the search category word is greater than a second preset threshold to add to the expanded word.
  • the extension words obtained based on the above method can easily be extended to adjacent categories or higher-level categories or other existing categories, for example, in the extension words of "beef” There is “chicken”, the expansion word of banana has “pineapple”, the expansion word of "computer desk” has “table”, if you need to recall the items corresponding to the existing category words, you only need to select the corresponding existing category Item recall can be carried out with the target words. In this application, these extended words can be filtered out through the existing category table. See Figure 6.
  • the item recall system also includes a second judgment module 25 and a filtering module 26. And an article category list 24, the article category list 24 includes multiple article categories and the attributes of all articles under each article category;
  • the second judgment module 25 is used to judge whether the extension word is included in the item category list 24, and if so, call the filtering module 26;
  • the filtering module 26 is used to filter out the expansion words included in the category list
  • the recall module 19 is used for recalling items according to the filtered expanded words and the search category words.
  • article recommendation articles and article names of various platforms are used as corpus to train the word2vec model, and based on the trained model, feature words and corresponding word vectors are generated, and then based on similarity calculations, several words with the highest similarity are obtained.
  • the expansion word of the search category word it complements the expansion of the search category word using the behavior data process described above, and further increases the coverage of the recall.

Landscapes

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

Abstract

一种物品召回方法、系统、电子设备及可读存储介质，所述物品召回方法包括：获取用户的行为数据，行为数据包括点击和/或购买的物品及每个物品的物品信息（10）；输入一检索类目词（20）；从行为数据中提取包含检索类目词的目标行为数据（30）；逆向关联获取目标行为数据中每个物品的检索词及每个检索词的检索频次（40）；对目标行为数据中物品信息进行分词，得到多个物品分词（50）；计算所有检索词与多个物品分词的相似度（60）；判断相似度是否大于第一预设阈值（70），若是，则提取所有检索词中检索频次大于预设频次的检索词作为检索类目词的扩展词（80）；根据检索类目词和扩展词进行物品召回（90）。该方法利用获取的行为数据提取得到物品类目词的扩展词，提高了召回物品的覆盖率。

Description

物品召回方法、系统、电子设备及可读存储介质
本申请要求申请日为2019/2/21的中国专利申请2019101303771的优先权。本申请引用上述中国专利申请的全文。
技术领域
本发明属于大数据处理领域，特别涉及一种物品召回方法、系统、电子设备及可读存储介质。
背景技术
在电商领域中，经常出现需要基于物品类目去做物品召回的类似场景，比如基于类目(例如“连衣裙”)去召回一批符合要求的物品的类似场景，为了后续筛选或者排序算法可以有更大的发挥空间，我们需要在不失准确的基础上，召回尽可能多的物品。现有技术会用到文字匹配的相关方法，但对于同一物品，可能存在描述上的差异，不一定会包含该类目词，导致无法被召回。
发明内容
本发明要解决的技术问题是为了克服现有技术中由于描述差异导致物品召回效果差的缺陷，提供一种物品召回方法、系统、电子设备及可读存储介质。
本发明是通过下述技术方案来解决上述技术问题：
一种物品召回方法，所述物品召回方法包括：
获取用户的行为数据，所述行为数据包括点击和/或购买的物品及每个物品的物品信息；
输入一检索类目词；
从所述行为数据中提取包含所述检索类目词的目标行为数据；
逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次；
对所述目标行为数据中物品信息进行分词，得到多个物品分词；
计算所有检索词与所述多个物品分词的相似度；
判断所述相似度是否大于第一预设阈值，若是，则提取所述所有检索词中检索频次大于预设频次的检索词作为所述检索类目词的扩展词；
根据所述检索类目词和所述扩展词进行物品召回。
较佳地，所述逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次的步骤具体包括：
根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应关系；
基于所述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
较佳地，所述计算所有检索词与所述多个物品分词的相似度的步骤具体包括：
对所有的检索词进行去重生成检索词集合；
对所述多个物品分词进行去重生成物品分词集合；
基于Jaccard算法(用于比较样本集之间的相似性与差异性的算法)计算所述检索词集合和所述物品分词集合的相似度。
较佳地，所述根据所述检索类目词和所述扩展词进行物品召回物品的步骤之前，所述物品召回方法还包括：
构建语料库，所述语料库包括物品推荐文章和物品基本信息；
基于word2vec(一种用于产生词向量的模型)对所述语料库进行训练，得到一词向量库，所述词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
基于相似度算法计算所述检索类目词与每个物品特征词的相似度；
提取与所述检索类目词的相似度大于第二预设阈值的物品特征词加入所述扩展词。
较佳地，所述根据所述检索类目词和所述扩展词进行物品召回物品的步骤之前，所述物品召回方法还包括：
预设一物品类目列表，所述物品类目列表包括多个物品类目及每个物品类目下所有物品的属性；
判断所述扩展词是否包含在所述物品类目列表中，若是，则滤除包含在所述类目列表的扩展词；
所述根据所述检索类目词和所述扩展词进行物品召回的步骤具体包括：
根据滤除后的扩展词和所述检索类目词进行物品召回。
一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现上述的物品召回方法。
一种可读存储介质，其上存储有计算机程序，所述程序被处理器执行时实现上述的物品召回方法的步骤。
一种物品召回系统，所述物品召回系统包括行为数据获取模块、输入模块、数据提 取模块、检索词获取模块、分词模块、第一计算模块、第一判断模块、扩展词生成模块和召回模块；
所述行为数据获取模块用于获取用户的行为数据，所述行为数据包括点击和/或购买的物品及每个物品的物品信息；
所述输入模块用于输入一检索类目词；
所述数据提取模块用于从所述行为数据中提取包含所述检索类目词的目标行为数据；
所述检索词获取模块用于逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次；
所述分词模块用于对所述目标行为数据中物品信息进行分词，得到多个物品分词；
所述第一计算模块用于计算所有检索词与所述多个物品分词的相似度；
所述第一判断模块用于判断所述相似度是否大于第一预设阈值，若是，则调用所述扩展词生成模块；
所述扩展词生成模块用于提取所述所有检索词中检索频次大于预设频次的检索词作为所述检索类目词的扩展词；
所述召回模块用于根据所述检索类目词和所述扩展词进行物品召回。
较佳地，所述检索词获取模块用于根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应关系，并基于所述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
较佳地，所述第一计算模块包括集合生成单元和计算单元；
所述集合生成单元用于对所有的检索词进行去重生成检索词集合，还用于对所述多个物品分词进行去重生成物品分词集合；
所述计算单元用于基于Jaccard算法计算所述检索词集合和所述物品分词集合的相似度。
较佳地，所述物品召回系统还包括语料库构建模块、词向量库生成模块和第二计算模块；
所述语料库构建模块用于构建语料库，所述语料库包括物品推荐文章和物品基本信息；
所述词向量库生成模块用于基于word2vec对所述语料库进行训练并得到一词向量库，所述词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
所述第二计算模块用于基于相似度算法计算所述检索类目词与每个物品特征词的相似度；
所述扩展词生成模块还用于提取与所述检索类目词的相似度大于第二预设阈值的物品特征词加入所述扩展词。
较佳地，所述物品召回系统还包括第二判断模块、滤除模块和一物品类目列表，所述物品类目列表包括多个物品类目及每个物品类目下所有物品的属性；
所述第二判断模块用于判断所述扩展词是否包含在所述物品类目列表中，若是，则调用所述滤除模块；
所述滤除模块用于滤除包含在所述类目列表的扩展词；
所述召回模块用于根据滤除后的扩展词和所述检索类目词进行物品召回。
本发明的积极进步效果在于：本发明在使用类目词进行物品召回时，利用获取的行为数据提取得到物品类目词的扩展词，在确保召回精准度的同时，提高了召回物品的覆盖率。
附图说明
图1为本发明实施例1的物品召回方法的流程图。
图2为本发明实施例1的物品召回方法中步骤60的流程图。
图3为本发明实施例2的物品召回方法的流程图。
图4为本发明实施例3的电子设备的结构示意图。
图5为本发明实施例5的物品召回系统的模块示意图。
图6为本发明实施例6的物品召回系统的模块示意图。
具体实施方式
下面通过实施例的方式进一步说明本发明，但并不因此将本发明限制在所述的实施例范围之中。
实施例1
一种物品召回方法，如图1所示，所述物品召回方法包括：
步骤10、获取用户的行为数据；行为数据包括点击和/或购买的物品及每个物品的物品信息；
步骤20、输入一检索类目词；
步骤30、从行为数据中提取包含检索类目词的目标行为数据；
步骤40、逆向关联获取目标行为数据中每个物品的检索词及每个检索词的检索频次；
步骤50、对目标行为数据中物品信息进行分词，得到多个物品分词；
步骤60、计算所有检索词与多个物品分词的相似度；
步骤70、判断相似度是否大于第一预设阈值，若是，则执行步骤80；若否，则表示检索词均不能作为所述检索类目词的扩展词；
步骤80、提取所有检索词中检索频次大于预设频次的检索词作为检索类目词的扩展词；
步骤90、根据检索类目词和扩展词进行物品召回。
其中，步骤40具体包括：
根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应关系，并基于所述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
参见图2，步骤60具体包括：
步骤601、对所有的检索词进行去重生成检索词集合；
步骤602、对多个物品分词进行去重生成物品分词集合；
步骤603、基于Jaccard算法计算检索词集合和物品分词集合的相似度。
本实施例中，在使用类目词进行物品召回时，利用获取的行为数据提取得到物品类目词的扩展词，在确保召回精准度的同时，提高了召回物品的覆盖率。
实施例2
本实施例的物品召回方法是在实施例1的基础上进一步改进，如图3所示，步骤90之前，所述物品召回方法还包括：
步骤81、构建语料库，语料库包括物品推荐文章和物品基本信息；
步骤82、基于word2vec对语料库进行训练，得到一词向量库；词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
步骤83、基于相似度算法计算检索类目词与每个物品特征词的相似度；
步骤84、提取与检索类目词的相似度大于第二预设阈值的物品特征词加入扩展词。
本实施例中，考虑到是对物品类目的扩展，基于上述方法得到的扩展词很容易扩展到邻近类目或者上级类目或者其他现有的类目，比如，“牛肉”的扩展词中有“鸡肉”，香蕉的扩展词中有“菠萝”，“电脑桌”的扩展词中有“桌子”，如果需要召回现有的类目词对应的物品，只需选取对应的现有的类目词进行物品召回即可，在本申请中，可以通过现有类目表去过滤来排除掉这些扩展词，参见图3，步骤90之前，所述物品召回方法还包括：
步骤85、预设一物品类目列表，物品类目列表包括多个物品类目及每个物品类目下 所有物品的属性；
步骤86、判断扩展词是否包含在物品类目列表中，若是，则执行步骤87；
步骤87、滤除包含在类目列表的扩展词；
进一步的，步骤90具体包括：
根据滤除后的扩展词和检索类目词进行物品召回。
本实施例中，利用各个平台的物品推荐文章、物品名称等作为语料，训练word2vec模型，并基于训练的模型产出特征词及对应的词向量，再基于相似度计算得到相似度最高的若干词作为检索类目词的扩展词，与上述利用行为数据进程检索类目词的扩展相辅相成，进一步增加召回的覆盖率。
实施例3
一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现实施例1或2中任意一个实施例所述的物品召回方法。
图4为本实施例提供的一种电子设备的结构示意图。图4示出了适于用来实现本发明实施方式的示例性电子设备90的框图。图4显示的电子设备90仅仅是一个示例，不应对本发明实施例的功能和使用范围带来任何限制。
如图4所示，电子设备90可以以通用计算设备的形式表现，例如其可以为服务器设备。电子设备90的组件可以包括但不限于：至少一个处理器91、至少一个存储器92、连接不同系统组件(包括存储器92和处理器91)的总线93。
总线93包括数据总线、地址总线和控制总线。
存储器92可以包括易失性存储器，例如随机存取存储器(RAM)921和/或高速缓存存储器922，还可以进一步包括只读存储器(ROM)923。
存储器92还可以包括具有一组(至少一个)程序模块924的程序工具925，这样的程序模块924包括但不限于：操作系统、一个或者多个应用程序、其它程序模块以及程序数据，这些示例中的每一个或某种组合中可能包括网络环境的实现。
处理器91通过运行存储在存储器92中的计算机程序，从而执行各种功能应用以及数据处理。
电子设备90也可以与一个或多个外部设备94(例如键盘、指向设备等)通信。这种通信可以通过输入/输出(I/O)接口95进行。并且，电子设备90还可以通过网络适配器96与一个或者多个网络(例如局域网(LAN)，广域网(WAN)和/或公共网络，例如因特网)通信。网络适配器96通过总线93与电子设备90的其它模块通信。应当明白，尽 管图中未示出，可以结合电子设备90使用其它硬件和/或软件模块，包括但不限于：微代码、设备驱动器、冗余处理器、外部磁盘驱动阵列、RAID(磁盘阵列)系统、磁带驱动器以及数据备份存储系统等。
应当注意，尽管在上文详细描述中提及了电子设备的若干单元/模块或子单元/模块，但是这种划分仅仅是示例性的并非强制性的。实际上，根据本申请的实施方式，上文描述的两个或更多单元/模块的特征和功能可以在一个单元/模块中具体化。反之，上文描述的一个单元/模块的特征和功能可以进一步划分为由多个单元/模块来具体化。
实施例4
一种计算机可读存储介质，其上存储有计算机程序，所述程序被处理器执行时实现实施例1或2中任意一个实施例所述的物品召回方法的步骤。
其中，可读存储介质可以采用的更具体可以包括但不限于：便携式盘、硬盘、随机存取存储器、只读存储器、可擦拭可编程只读存储器、光存储器件、磁存储器件或上述的任意合适的组合。
在可能的实施方式中，本发明还可以实现为一种程序产品的形式，其包括程序代码，当所述程序产品在终端设备上运行时，所述程序代码用于使所述终端设备执行实现实施例1或2中任意一个实施例所述的物品召回方法的步骤。
其中，可以以一种或多种程序设计语言的任意组合来编写用于执行本发明的程序代码，所述程序代码可以完全地在用户设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户设备上部分在远程设备上执行或完全在远程设备上执行。
实施例5
一种物品召回系统，如图5所示，所述物品召回系统包括行为数据获取模块11、输入模块12、数据提取模块13、检索词获取模块14、分词模块15、第一计算模块16、第一判断模块17、扩展词生成模块18和召回模块19；
所述行为数据获取模块11用于获取用户的行为数据，所述行为数据包括点击和/或购买的物品及每个物品的物品信息；
所述输入模块12用于输入一检索类目词；
所述数据提取模块13用于从所述行为数据中提取包含所述检索类目词的目标行为数据；
所述检索词获取模块14用于逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次；具体的，所述检索词获取模块14用于根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应关系，并基于所 述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
所述分词模块15用于对所述目标行为数据中物品信息进行分词，得到多个物品分词；
所述第一计算模块16用于计算所有检索词与所述多个物品分词的相似度；
其中，所述第一计算模块16包括集合生成单元161和计算单元162；
所述集合生成单元161用于对所有的检索词进行去重生成检索词集合，还用于对所述多个物品分词进行去重生成物品分词集合；
所述计算单元162用于基于Jaccard算法计算所述检索词集合和所述物品分词集合的相似度。
所述第一判断模块17用于判断所述相似度是否大于第一预设阈值，若是，则调用所述扩展词生成模块18；
所述扩展词生成模块18用于提取所述所有检索词中检索频次大于预设频次的检索词作为所述检索类目词的扩展词；
所述召回模块19用于根据所述检索类目词和所述扩展词进行物品召回。
本实施例中，在使用类目词进行物品召回时，利用获取的行为数据提取得到物品类目词的扩展词，在确保召回精准度的同时，提高了召回物品的覆盖率。
实施例6
本实施例的物品召回系统是在实施例5的基础上进一步改进，如图6所示，所述物品召回系统还包括语料库构建模块21、词向量库生成模块22和第二计算模块23；
所述语料库构建模块21用于构建语料库，所述语料库包括物品推荐文章和物品基本信息；
所述词向量库生成模块22用于基于word2vec对所述语料库进行训练并得到一词向量库，所述词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
所述第二计算模块23用于基于相似度算法计算所述检索类目词与每个物品特征词的相似度；
所述扩展词生成模块18还用于提取与所述检索类目词的相似度大于第二预设阈值的物品特征词加入所述扩展词。
本实施例中，考虑到是对物品类目的扩展，基于上述方法得到的扩展词很容易扩展到邻近类目或者上级类目或者其他现有的类目，比如，“牛肉”的扩展词中有“鸡肉”，香蕉的扩展词中有“菠萝”，“电脑桌”的扩展词中有“桌子”，如果需要召回现有的类目词对应的物品，只需选取对应的现有的类目词进行物品召回即可，在本申请中，可以通过现有类目表去过滤来排除掉这些扩展词，参见图6，所述物品召回系统还包括第二判断模 块25、滤除模块26和一物品类目列表24，所述物品类目列表24包括多个物品类目及每个物品类目下所有物品的属性；
所述第二判断模块25用于判断所述扩展词是否包含在所述物品类目列表24中，若是，则调用所述滤除模块26；
所述滤除模块26用于滤除包含在所述类目列表的扩展词；
所述召回模块19用于根据滤除后的扩展词和所述检索类目词进行物品召回。
本实施例中，利用各个平台的物品推荐文章、物品名称等作为语料，训练word2vec模型，并基于训练的模型产出特征词及对应的词向量，再基于相似度计算得到相似度最高的若干词作为检索类目词的扩展词，与上述利用行为数据进程检索类目词的扩展相辅相成，进一步增加召回的覆盖率。
虽然以上描述了本发明的具体实施方式，但是本领域的技术人员应当理解，这些仅是举例说明，在不背离本发明的原理和实质的前提下，可以对这些实施方式做出多种变更或修改。因此，本发明的保护范围由所附权利要求书限定。

Claims (12)

  1. 一种物品召回方法，其特征在于，所述物品召回方法包括：
    获取用户的行为数据，所述行为数据包括点击和/或购买的物品及每个物品的物品信息；
    输入一检索类目词；
    从所述行为数据中提取包含所述检索类目词的目标行为数据；
    逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次；
    对所述目标行为数据中物品信息进行分词，得到多个物品分词；
    计算所有检索词与所述多个物品分词的相似度；
    判断所述相似度是否大于第一预设阈值，若是，则提取所述所有检索词中检索频次大于预设频次的检索词作为所述检索类目词的扩展词；
    根据所述检索类目词和所述扩展词进行物品召回。
  2. 如权利要求1所述的物品召回方法，其特征在于，所述逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次的步骤具体包括：
    根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应关系；
    基于所述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
  3. 如权利要求1-2中至少一项所述的物品召回方法，其特征在于，所述计算所有检索词与所述多个物品分词的相似度的步骤具体包括：
    对所有的检索词进行去重生成检索词集合；
    对所述多个物品分词进行去重生成物品分词集合；
    基于Jaccard算法计算所述检索词集合和所述物品分词集合的相似度。
  4. 如权利要求1-3中至少一项所述的物品召回方法，其特征在于，所述根据所述检索类目词和所述扩展词进行物品召回物品的步骤之前，所述物品召回方法还包括：
    构建语料库，所述语料库包括物品推荐文章和物品基本信息；
    基于word2vec对所述语料库进行训练，得到一词向量库，所述词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
    基于相似度算法计算所述检索类目词与每个物品特征词的相似度；
    提取与所述检索类目词的相似度大于第二预设阈值的物品特征词加入所述扩展词。
  5. 如权利要求4所述的物品召回方法，其特征在于，所述根据所述检索类目词和所述扩展词进行物品召回物品的步骤之前，所述物品召回方法还包括：
    预设一物品类目列表，所述物品类目列表包括多个物品类目及每个物品类目下所有物品的属性；
    判断所述扩展词是否包含在所述物品类目列表中，若是，则滤除包含在所述类目列表的扩展词；
    所述根据所述检索类目词和所述扩展词进行物品召回的步骤具体包括：
    根据滤除后的扩展词和所述检索类目词进行物品召回。
  6. 一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至5任一项所述的物品召回方法。
  7. 一种可读存储介质，其上存储有计算机程序，其特征在于，所述程序被处理器执行时实现权利要求1至5任一项所述的物品召回方法的步骤。
  8. 一种物品召回系统，其特征在于，所述物品召回系统包括行为数据获取模块、输入模块、数据提取模块、检索词获取模块、分词模块、第一计算模块、第一判断模块、扩展词生成模块和召回模块；
    所述行为数据获取模块用于获取用户的行为数据，所述行为数据包括点击和/或购买的物品及每个物品的物品信息；
    所述输入模块用于输入一检索类目词；
    所述数据提取模块用于从所述行为数据中提取包含所述检索类目词的目标行为数据；
    所述检索词获取模块用于逆向关联获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次；
    所述分词模块用于对所述目标行为数据中物品信息进行分词，得到多个物品分词；
    所述第一计算模块用于计算所有检索词与所述多个物品分词的相似度；
    所述第一判断模块用于判断所述相似度是否大于第一预设阈值，若是，则调用所述扩展词生成模块；
    所述扩展词生成模块用于提取所述所有检索词中检索频次大于预设频次的检索词作为所述检索类目词的扩展词；
    所述召回模块用于根据所述检索类目词和所述扩展词进行物品召回。
  9. 如权利要求8所述的物品召回系统，其特征在于，所述检索词获取模块用于根据所述行为数据生成关系列表，所述关系列表用于表征检索词与点击和/或购买的物品的对应 关系，并基于所述关系列表获取所述目标行为数据中每个物品的检索词及每个检索词的检索频次。
  10. 如权利要求8-9中至少一项所述的物品召回系统，其特征在于，所述第一计算模块包括集合生成单元和计算单元；
    所述集合生成单元用于对所有的检索词进行去重生成检索词集合，还用于对所述多个物品分词进行去重生成物品分词集合；
    所述计算单元用于基于Jaccard算法计算所述检索词集合和所述物品分词集合的相似度。
  11. 如权利要求8-10中至少一项所述的物品召回系统，其特征在于，所述物品召回系统还包括语料库构建模块、词向量库生成模块和第二计算模块；
    所述语料库构建模块用于构建语料库，所述语料库包括物品推荐文章和物品基本信息；
    所述词向量库生成模块用于基于word2vec对所述语料库进行训练并得到一词向量库，所述词向量库存储有多个物品特征词及与每个物品特征词对应的特征向量；
    所述第二计算模块用于基于相似度算法计算所述检索类目词与每个物品特征词的相似度；
    所述扩展词生成模块还用于提取与所述检索类目词的相似度大于第二预设阈值的物品特征词加入所述扩展词。
  12. 如权利要求11所述的物品召回系统，其特征在于，所述物品召回系统还包括第二判断模块、滤除模块和一物品类目列表，所述物品类目列表包括多个物品类目及每个物品类目下所有物品的属性；
    所述第二判断模块用于判断所述扩展词是否包含在所述物品类目列表中，若是，则调用所述滤除模块；
    所述滤除模块用于滤除包含在所述类目列表的扩展词；
    所述召回模块用于根据滤除后的扩展词和所述检索类目词进行物品召回。
PCT/CN2020/070014 2019-02-21 2020-01-02 物品召回方法、系统、电子设备及可读存储介质 Ceased WO2020168839A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/296,159 US11907659B2 (en) 2019-02-21 2020-01-02 Item recall method and system, electronic device and readable storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910130377.1A CN111597297B (zh) 2019-02-21 2019-02-21 物品召回方法、系统、电子设备及可读存储介质
CN201910130377.1 2019-02-21

Publications (1)

Publication Number Publication Date
WO2020168839A1 true WO2020168839A1 (zh) 2020-08-27

Family

ID=72144286

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/070014 Ceased WO2020168839A1 (zh) 2019-02-21 2020-01-02 物品召回方法、系统、电子设备及可读存储介质

Country Status (3)

Country Link
US (1) US11907659B2 (zh)
CN (1) CN111597297B (zh)
WO (1) WO2020168839A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112329954A (zh) * 2020-11-04 2021-02-05 中国平安人寿保险股份有限公司 物品召回方法、装置、终端设备及存储介质
CN113763114A (zh) * 2021-03-04 2021-12-07 北京沃东天骏信息技术有限公司 物品信息搭配方法、装置及存储介质
CN114443989A (zh) * 2022-01-30 2022-05-06 北京百度网讯科技有限公司 排序方法、排序模型的训练方法、装置、电子设备及介质
US12174902B2 (en) 2022-01-30 2024-12-24 Beijing Baidu Netcom Science Technology Co., Ltd. Ranking of recall data

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112150251B (zh) * 2020-10-09 2024-07-16 北京明朝万达科技股份有限公司 物品名称治理方法和装置
CN113762535B (zh) * 2021-03-18 2024-12-06 北京沃东天骏信息技术有限公司 一种物品召回方法和装置
CN114881108A (zh) * 2022-03-30 2022-08-09 新华智云科技有限公司 一种商品推荐的召回方法、装置、设备及存储介质
CN115017385B (zh) * 2022-06-30 2025-06-13 北京沃东天骏信息技术有限公司 一种物品搜索方法、装置、设备和存储介质
CN115205085B (zh) * 2022-08-26 2025-09-26 广州华多网络科技有限公司 通信地址联想方法及其装置、设备、介质、产品
CN117056459B (zh) * 2023-08-07 2024-05-10 北京网聘信息技术有限公司 一种向量召回方法和装置
CN117235351B (zh) * 2023-08-25 2025-08-15 微梦创科网络科技（中国）有限公司 一种物料推荐的方法、装置、电子设备和计算机可读介质
CN117093604B (zh) * 2023-10-20 2024-02-02 中信证券股份有限公司 检索信息生成方法、装置、电子设备和计算机可读介质
CN119760097B (zh) * 2024-12-27 2025-11-14 北京中科闻歌科技股份有限公司 一种大模型rag召回策略智能规划方法、装置、介质及设备

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080162118A1 (en) * 2006-12-15 2008-07-03 International Business Machines Corporation Technique for Searching Out New Words That Should Be Registered in Dictionary For Speech Processing
CN104636334A (zh) * 2013-11-06 2015-05-20 阿里巴巴集团控股有限公司 一种关键词推荐方法和装置
CN106874492A (zh) * 2017-02-23 2017-06-20 北京京东尚科信息技术有限公司 搜索方法和装置
CN108664637A (zh) * 2018-05-15 2018-10-16 惠龙易通国际物流股份有限公司 一种检索方法及系统

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6098033A (en) * 1997-07-31 2000-08-01 Microsoft Corporation Determining similarity between words
US7343280B2 (en) * 2003-07-01 2008-03-11 Microsoft Corporation Processing noisy data and determining word similarity
CN100595759C (zh) * 2007-04-25 2010-03-24 北大方正集团有限公司 查询扩展方法和装置以及相关检索词库
US8768852B2 (en) * 2009-01-13 2014-07-01 Amazon Technologies, Inc. Determining phrases related to other phrases
US8655648B2 (en) * 2010-09-01 2014-02-18 Microsoft Corporation Identifying topically-related phrases in a browsing sequence
CN103577416B (zh) * 2012-07-20 2017-09-22 阿里巴巴集团控股有限公司 扩展查询方法及系统
CN103577432B (zh) * 2012-07-26 2017-07-14 阿里巴巴集团控股有限公司 一种商品信息搜索方法和系统
CN103729359B (zh) * 2012-10-12 2017-03-01 阿里巴巴集团控股有限公司 一种推荐搜索词的方法及系统
JP6093200B2 (ja) * 2013-02-05 2017-03-08 日本放送協会 情報検索装置及び情報検索プログラム
CN105183733A (zh) * 2014-06-05 2015-12-23 阿里巴巴集团控股有限公司 一种文本信息的匹配、业务对象的推送方法和装置
US20160070803A1 (en) * 2014-09-09 2016-03-10 Funky Flick, Inc. Conceptual product recommendation
CN108733706B (zh) * 2017-04-20 2022-12-20 腾讯科技（深圳）有限公司 热度信息的生成方法和装置
US10747792B2 (en) * 2017-11-30 2020-08-18 Bby Solutions, Inc. Streaming events analysis for search recall improvements
CN108304444B (zh) * 2017-11-30 2021-12-14 腾讯科技（深圳）有限公司 信息查询方法及装置
US11562011B2 (en) * 2018-10-29 2023-01-24 EMC IP Holding Company LLC Incorporating data into search engines using deep learning mechanisms

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080162118A1 (en) * 2006-12-15 2008-07-03 International Business Machines Corporation Technique for Searching Out New Words That Should Be Registered in Dictionary For Speech Processing
CN104636334A (zh) * 2013-11-06 2015-05-20 阿里巴巴集团控股有限公司 一种关键词推荐方法和装置
CN106874492A (zh) * 2017-02-23 2017-06-20 北京京东尚科信息技术有限公司 搜索方法和装置
CN108664637A (zh) * 2018-05-15 2018-10-16 惠龙易通国际物流股份有限公司 一种检索方法及系统

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112329954A (zh) * 2020-11-04 2021-02-05 中国平安人寿保险股份有限公司 物品召回方法、装置、终端设备及存储介质
CN112329954B (zh) * 2020-11-04 2024-06-11 中国平安人寿保险股份有限公司 物品召回方法、装置、终端设备及存储介质
CN113763114A (zh) * 2021-03-04 2021-12-07 北京沃东天骏信息技术有限公司 物品信息搭配方法、装置及存储介质
CN114443989A (zh) * 2022-01-30 2022-05-06 北京百度网讯科技有限公司 排序方法、排序模型的训练方法、装置、电子设备及介质
CN114443989B (zh) * 2022-01-30 2023-02-03 北京百度网讯科技有限公司 排序方法、排序模型的训练方法、装置、电子设备及介质
US12174902B2 (en) 2022-01-30 2024-12-24 Beijing Baidu Netcom Science Technology Co., Ltd. Ranking of recall data

Also Published As

Publication number Publication date
US11907659B2 (en) 2024-02-20
US20220019739A1 (en) 2022-01-20
CN111597297A (zh) 2020-08-28
CN111597297B (zh) 2025-08-22

Similar Documents

Publication Publication Date Title
WO2020168839A1 (zh) 物品召回方法、系统、电子设备及可读存储介质
US12346958B2 (en) Method and article of manufacture for automated generation of titles and descriptions for electronic commerce products
US8352496B2 (en) Entity name matching
CN102855259A (zh) 大规模数据聚类分析的并行化
US9990268B2 (en) System and method for detection of duplicate bug reports
CN113204667B (zh) 音频标注模型的训练与音频标注的方法、装置
CN112800197A (zh) 一种目标故障信息的确定方法和装置
CN110135769B (zh) 货品属性填充方法及装置、存储介质及电子终端
TWI780355B (zh) 維修對象的定損方法及裝置、電子設備
CN102844755A (zh) 提取命名实体的方法
CN110083756A (zh) 识别知识图数据结构中的冗余节点
CN109033220B (zh) 标注数据的自动选取方法、系统、设备和存储介质
US20210034704A1 (en) Identifying Ambiguity in Semantic Resources
US20140214844A1 (en) Multiple classification models in a pipeline
US9336249B2 (en) Decision tree with just-in-time nodal computations
CN116597443B (zh) 素材标签处理方法、装置、电子设备及介质
CN114328855B (zh) 文档查询方法、装置、电子设备和可读存储介质
CN117539976A (zh) 电力知识库更新方法、装置、设备、存储介质和程序产品
CN110543635A (zh) 基于深度学习的信息检测方法、装置及计算机存储介质
CN110874387A (zh) 移动设备识别符共现关系稀疏图的构建方法及装置
EP4594940A1 (en) Entity linking using a graph neural network
CN114357977B (zh) 一种反抄袭实现方法、系统、设备及存储介质
CN114239603B (zh) 业务需求匹配方法、装置、计算机设备和存储介质
CN115617652B (zh) 测试用例处理方法、装置、计算设备及计算机存储介质
CN121210572B (zh) 数据库预同步的实现方法、装置、设备、介质及产品

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20760328

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 21/01/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 20760328

Country of ref document: EP

Kind code of ref document: A1