WO2023029513A1 - 基于人工智能的搜索意图识别方法、装置、设备及介质 - Google Patents

基于人工智能的搜索意图识别方法、装置、设备及介质 Download PDF

Info

Publication number
WO2023029513A1
WO2023029513A1 PCT/CN2022/087820 CN2022087820W WO2023029513A1 WO 2023029513 A1 WO2023029513 A1 WO 2023029513A1 CN 2022087820 W CN2022087820 W CN 2022087820W WO 2023029513 A1 WO2023029513 A1 WO 2023029513A1
Authority
WO
WIPO (PCT)
Prior art keywords
intent
query statement
keyword
tag
matched
Prior art date
Application number
PCT/CN2022/087820
Other languages
English (en)
French (fr)
Inventor
张华�
Original Assignee
康键信息技术(深圳)有限公司
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 康键信息技术(深圳)有限公司 filed Critical 康键信息技术(深圳)有限公司
Publication of WO2023029513A1 publication Critical patent/WO2023029513A1/zh

Links

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/332Query formulation
    • G06F16/3329Natural language query formulation or dialogue systems
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/334Query execution
    • G06F16/3344Query execution using natural language analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/334Query execution
    • G06F16/3346Query execution using probabilistic model
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35Clustering; Classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/205Parsing
    • G06F40/216Parsing using statistical methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • G06F40/289Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295Named entity recognition

Definitions

  • the present application relates to the technical field of artificial intelligence, and in particular to an artificial intelligence-based search intent recognition method, device, device, and storage medium.
  • Search intent recognition refers to analyzing the user's search terms to obtain the user's intent and needs, so as to recommend the products or content that the user needs most. Therefore, improving the search recognition intent can improve the accuracy of product or content recommendation.
  • the search sentences entered by users are not standardized and have different lengths.
  • the obtained word vector representation ability is poor, resulting in a low accuracy rate of search intent recognition.
  • the present application provides an artificial intelligence-based search intent recognition method, the method comprising:
  • a word segmentation operation is performed on the query statement, at least one keyword of the query statement after the word segmentation operation is determined, and the entity type of each keyword is extracted;
  • the query sentence is input into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • the present application also provides an artificial intelligence-based search intent recognition device, the artificial intelligence-based search intent recognition device includes:
  • the first matching module for matching the query sentence input by the user with a preset rule set, and judging whether the target intent of the query sentence is matched in the rule set;
  • Extraction module used to perform a word segmentation operation on the query statement when it is judged that the target intent of the query statement is not matched in the rule set, determine at least one keyword of the query statement after the word segmentation operation, and extract each keyword the entity type of the word;
  • the second matching module used to match the entity type of each keyword with the pre-established tag dictionary tree, and judge whether the tag intent of the entity type is matched in the tag dictionary tree, when it is judged in the tag dictionary tree
  • the tag intent of the entity type is matched, the tag intent of the successfully matched entity type is obtained, and the target intent of the query statement is obtained based on the tag intent of the successfully matched entity type;
  • Recognition module used to input the query sentence into a pre-established intent recognition model to obtain the target intent of the query sentence when it is judged that the tag intent of the entity type is not matched in the tag dictionary tree.
  • the present application also provides an electronic device, the electronic device comprising:
  • the memory stores a program executable by the at least one processor, the program is executed by the at least one processor, so that the at least one processor can perform artificial intelligence-based search intent recognition as described below method:
  • a word segmentation operation is performed on the query statement, at least one keyword of the query statement after the word segmentation operation is determined, and the entity type of each keyword is extracted;
  • the query sentence is input into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • the present application also provides a computer-readable storage medium, the computer-readable storage medium stores an artificial intelligence-based search intent recognition program, and when the artificial intelligence-based search intent recognition program is executed by a processor, the following AI-based search intent recognition method:
  • a word segmentation operation is performed on the query statement, at least one keyword of the query statement after the word segmentation operation is determined, and the entity type of each keyword is extracted;
  • the query sentence is input into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • FIG. 1 is a schematic flow chart of a preferred embodiment of the artificial intelligence-based search intent recognition method of the present application
  • FIG. 2 is a block diagram of a preferred embodiment of an artificial intelligence-based search intent recognition device in the present application
  • FIG. 3 is a schematic diagram of a preferred embodiment of the electronic device of the present application.
  • AI artificial intelligence
  • the embodiments of the present application may acquire and process relevant data based on artificial intelligence technology.
  • artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
  • Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics.
  • Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
  • the present application provides an artificial intelligence-based search intent recognition method.
  • FIG. 1 it is a schematic flow chart of an embodiment of an artificial intelligence-based search intent recognition method of the present application.
  • the method can be executed by an electronic device, and the electronic device can be realized by software and/or hardware.
  • AI-based search intent recognition methods include:
  • Step S10 Match the query sentence input by the user with a preset rule set, and judge whether the target intent of the query sentence is matched in the rule set.
  • the application scenario of this solution can be that when a user uses an APP with a search function (for example, an APP in the smart medical field), the user's search intention is identified to feed back the search results that the user most wants. For example, it is recognized that the user is The intention of "purchasing medicine", "registering” or "asking a doctor” is to feed back relevant information to users.
  • This solution uses smart medical APP as an example to illustrate this solution. It should be noted that the actual application scenarios of this solution are not limited to the above APPs, but also social APPs and e-commerce APPs.
  • the query sentence entered by the user in the interactive interface is obtained.
  • the content entered by the user can be
  • the query sentence can also be a query word, for example, the name of a certain drug, the name of a hospital, or the name of a certain disease.
  • the rule set includes a variety of commonly used rule statements, and the rule set can include "drug rule set", " "Hospital rule set”, etc.
  • Drug rule set can refer to the full name of various drugs, and its corresponding intention can be that the user wants to search for the purchase link of the drug or learn about the detailed information of the drug.
  • Hospital rule set can refer to the hospital The corresponding intention may be that the user wants to see a doctor in the hospital or learn about the departments of the hospital and other information.
  • the judging whether the preset rule set matches the target intent of the query statement includes:
  • the matching rule statement is used as the target intent of the query statement
  • the query statement matches any rule statement in the rule set successfully, that is, the query statement is the same as the rule statement in the rule set, the intent corresponding to the rule statement is used as the target intent of the query statement, and relevant searches are fed back to the user according to the intent result. If the query statement fails to match all the rule statements in the rule set, it is determined that the target intent of the query statement is not matched in the rule set.
  • the query sentence entered by the user is "Fukuan Tablets" and the drug rule sentence is included in the rule set, then it is recognized that the user intends to purchase the drug or know the detailed information of the drug, and the purchase link of the drug and the The detailed information of the drug is fed back to the user.
  • Step S20 When it is judged that the target intent of the query statement is not matched in the rule set, perform a word segmentation operation on the query statement, determine at least one keyword of the query statement after the word segmentation operation, and extract the key words of each keyword Entity type.
  • the keyword of the query sentence can be identified.
  • the target intent of the query statement Specifically, when it is judged that the target intent of the query statement is not matched in the rule set, the word segmentation operation is performed on the query statement, for example, the word segmentation of the query statement can be performed using forward matching splitting, reverse matching splitting, or the least segmentation algorithm. Determine at least one keyword of the query statement after the word segmentation operation, and extract the entity type of each keyword according to the named entity recognition algorithm.
  • NER Named Entity Recognition
  • the entity extraction algorithm of this solution can be a conditional random field model, which is a discriminant probability model, which is often used to label or analyze sequence data, such as natural language text.
  • said determining at least one keyword of the query statement after the word segmentation operation includes:
  • the keyword of the query statement can also be called the subject word.
  • the business subject is a limited set, mainly including diseases, symptoms, curative effects, treatment methods, organ parts, brands, product categories, product attributes, departments, A collection of nouns such as hospital names.
  • a dictionary tree can be established to match the keywords that appear in the query statement, and each participle of the query statement is matched with the keywords corresponding to each node of the dictionary tree. If the participle matches the keyword of any node of the keyword dictionary tree successfully, The participle is then used as the keyword of the query statement.
  • said respectively matching said respective word segmentation with keywords corresponding to each node of said keyword dictionary tree includes:
  • the each participle is converted into pinyin and matched with the pre-built pinyin dictionary tree.
  • the pinyin of any participle matches the When the keyword corresponding to any node of the pinyin dictionary tree is matched successfully, the keyword corresponding to the node that matches successfully is used as the keyword of the query statement;
  • the query sentence entered by the user may contain typos, resulting in incorrect intent recognition, when the word segmentation fails to match the keywords corresponding to the keyword dictionary tree nodes, error correction can be performed on the word segmentation of the query sentence, and the error correction is based on pinyin.
  • Priority because there are many errors with the same pronunciation and different characters, and there is no ambiguity.
  • a corresponding keyword pinyin dictionary tree is also established, which will convert the word segmentation that cannot match the keyword into pinyin, and then match it again in the pinyin dictionary tree. If the match is successful, the successful match will be
  • the keywords in the pinyin dictionary tree are used as the keywords of the query statement. For example, if the participle is "Amoxicillin", it will match "Amoxicillin”.
  • the difference value between the word segmentation and the keyword in the Pinyin dictionary tree is calculated using the edit distance, and the keyword with the smallest difference value is selected as the keyword of the query statement.
  • the search scenarios for chemical names of drugs Since many chemical names are rare words, users will search in different ways with the same sound. For example, “Metformin Capsules”, most users will input “Metformin Capsules”, which can be corrected by editing distance.
  • Edit distance refers to the quantitative measurement of the degree of difference between two strings. The measurement method is to see how many times of processing is required to change one string into another.
  • determining at least one keyword of the query statement after the word segmentation operation also includes:
  • Calculate the word frequency of each participle in the query statement calculate the IDF value and TF value of each participle based on the word frequency, multiply the IDF value of each participle with the TF value corresponding to each participle to obtain the TF-IDF value of each participle , selecting a preset number of words with preset parts of speech as keywords of the query statement based on the TF-IDF value of each word segment.
  • the TF-IDF algorithm can also be used to extract some colloquial keywords, such as: diarrhea (diarrhea). Calculate the IDF (inverse document frequency value), and then calculate the TF (term frequency) value of each word in the question text.
  • TF (the number of times the word appears in the text)/(the sum of the number of times each word appears in the text)
  • the TF-IDF value can evaluate the word The importance of the word in the text, the greater the TF-IDF value, the higher the priority as a keyword, and the greater the TF-IDF value, the higher the importance of the word to the sentence.
  • a bidirectional matching algorithm can also be used to perform a word segmentation operation on the query statement, and the specific word segmentation steps include:
  • the read participle is matched with the preset thesaurus to obtain the first matching result, which contains the first phrase of the first quantity and the single word of the second quantity in the first matching result;
  • the read participle is matched with a preset thesaurus to obtain a second matching result, which contains a third number of second phrases and a fourth number of words;
  • first number is equal to the third number and the second number is less than or equal to the fourth number, or if the first number is less than the third number, then match the first The result is used as the word segmentation result of the query statement; if the first quantity is equal to the second quantity and the third quantity is greater than the fourth quantity, or, if the first quantity is greater than the third quantity, Then use the second matching result as the word segmentation result of the query sentence.
  • Step S30 Match the entity type of each keyword with the pre-established tag dictionary tree, and judge whether the tag intent of the entity type is matched in the tag dictionary tree, and when it is judged that the entity type is matched in the tag dictionary tree
  • the tag intent of the entity type is obtained, the tag intent of the successfully matched entity type is obtained, and the target intent of the query statement is obtained based on the tag intent of the successfully matched entity type.
  • the entity type of each keyword is matched with the pre-established tag dictionary tree.
  • the tag intent corresponding to the vocabulary each entity type has a corresponding tag intent, for example, if the entity type is "brand", the corresponding tag intent is the intent of "purchase”.
  • the tag intent corresponding to each entity type can be obtained, and the tag with the largest number of tag intents in each entity type can be selected as the target intent of the query sentence.
  • the dictionary tree also known as the prefix tree, is an ordered tree used to store associative arrays, where the keys are usually strings, and the keys are not directly stored in the nodes, but are determined by the positions of the nodes in the tree. All descendants of a node have the same prefix, which is the string corresponding to this node, and the root node corresponds to the empty string. In general, not all nodes have corresponding values, only the keys corresponding to leaf nodes and some internal nodes have relevant values.
  • the target intent of the query statement is obtained based on the label intent of the successfully matched entity type, including:
  • the target intent of the query statement is selected from tag intents corresponding to each entity type.
  • the first type of users refers to new users, and the tag with the largest number of tag intents can be used as the target intent of the query statement.
  • Clicks select the intent from the number of label intents in each entity type, for example, in the user's historical search, the number of clicks on the purchase link is the largest, and purchase can be used as the target intent of the query statement.
  • the priority rules the priority order of the tag intents of each entity type in different types of user query sentences is determined, so that the target intent of the query sentences input by the user can be more accurately identified.
  • the target intent of the query statement is obtained based on the label intent of the successfully matched entity type, including:
  • the entity type of the preset type can refer to "doctor's name".
  • the label intent corresponding to "doctor's name” will be " Inquiry" as the target intent of the query statement, that is, to determine whether the user wants to specify a doctor for consultation, and directly return the search results of the doctor. It can quickly determine whether the user intends to search for a doctor to consult a doctor. According to the historical user click data analysis, in the case of searching for a person's name, 90% of the users click on the doctor with the name to inquire.
  • Step S40 When it is judged that no tag intent of the entity type is matched in the tag dictionary tree, input the query sentence into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • the query statement can be input into the pre-established intent recognition model to obtain the target intent of the query statement.
  • the intention recognition model can be obtained by training based on the doc2vec model, and the intention recognition model can be obtained by using the patient complaints of the historical online consultation and the corresponding pseudo-diagnosis results to generate a sample set to train the doc2vec model.
  • the dictionary tree cannot match the corresponding label intent, and the query sentence is input into the intent recognition model, and the output result is "children with diarrhea ".
  • the method also includes:
  • the similarity value between the two is greater than the preset threshold (for example, 90%), and
  • the intent of the historical query sentence can be used as the target intent of the query sentence, and the intent of the query sentence input by the user can be compared with the intent of the historical query sentence. verify;
  • FIG. 2 it is a schematic diagram of functional modules of an artificial intelligence-based search intent recognition device 100 of the present application.
  • the artificial intelligence-based search intent recognition apparatus 100 described in this application can be installed in electronic equipment.
  • the artificial intelligence-based search intent recognition device 100 may include a first matching module 110 , an extraction module 120 , a second matching module 130 and a recognition module 140 .
  • the module described in the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of the electronic device and can complete fixed functions, and are stored in the memory of the electronic device.
  • each module/unit is as follows:
  • the first matching module 110 for matching the query sentence input by the user with a preset rule set, and judging whether the target intent of the query sentence is matched in the rule set.
  • the extraction module 120 is used to perform a word segmentation operation on the query statement when it is judged that the target intent of the query statement is not matched in the rule set, determine at least one keyword of the query statement after the word segmentation operation, and extract each The entity type of the keyword.
  • the second matching module 130 is used to match the entity type of each keyword with a pre-established tag dictionary tree, and judge whether the tag intent of the entity type is matched in the tag dictionary tree.
  • the tag intent of the entity type is matched, the tag intent of the successfully matched entity type is obtained, and the target intent of the query statement is obtained based on the tag intent of the successfully matched entity type.
  • the recognition module 140 is configured to input the query sentence into a pre-established intent recognition model to obtain the target intent of the query sentence when it is judged that no tag intent of the entity type is matched in the tag dictionary tree.
  • said determining at least one keyword of the query statement after the word segmentation operation includes:
  • said respectively matching each of the word segmentations with keywords corresponding to each node of the keyword dictionary tree includes:
  • the each participle is converted into pinyin and matched with the pre-built pinyin dictionary tree.
  • the pinyin of any participle matches the When the keyword corresponding to any node of the pinyin dictionary tree is matched successfully, the keyword corresponding to the node that matches successfully is used as the keyword of the query statement;
  • the target intent of the query statement is obtained based on the label intent of the successfully matched entity type, including:
  • the target intent of the query statement is selected from tag intents corresponding to each entity type.
  • the target intent of the query statement is obtained based on the label intent of the successfully matched entity type, including:
  • the judging whether the preset rule set matches the target intent of the query statement includes:
  • the matching rule statement is used as the target intent of the query statement
  • the identification module 140 is also used for:
  • FIG. 3 it is a schematic diagram of a preferred embodiment of the electronic device 1 of the present application.
  • the electronic device 1 includes but not limited to: a memory 11 , a processor 12 , a display 13 and a network interface 14 .
  • the electronic device 1 is connected to the network through the network interface 14 to obtain raw data.
  • the network may be an enterprise intranet (Intranet), Internet (Internet), Global System of Mobile Communication (Global System of Mobile communication, GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth (Bluetooth) , Wi-Fi, call network and other wireless or wired networks.
  • the memory 11 includes at least one type of readable storage medium
  • the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static Random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc.
  • the computer readable storage Media can be either non-volatile or volatile.
  • the storage 11 may be an internal storage unit of the electronic device 1 , such as a hard disk or a memory of the electronic device 1 .
  • the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in hard disk equipped with the electronic device 1, a smart memory card (SmartMediaCard, SMC), Secure Digital (SecureDigital, SD) card, flash memory card (FlashCard), etc.
  • the storage 11 may also include both an internal storage unit of the electronic device 1 and an external storage device thereof.
  • the memory 11 is generally used to store the operating system and various application software installed in the electronic device 1 , such as the program code of the artificial intelligence-based search intent recognition program 10 .
  • the memory 11 can also be used to temporarily store various types of data that have been output or will be output.
  • the processor 12 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments.
  • the processor 12 is generally used to control the overall operation of the electronic device 1 , for example, perform data interaction or communication-related control and processing.
  • the processor 12 is configured to run the program codes stored in the memory 11 or process data, for example, run the program codes of the search intent recognition program 10 based on artificial intelligence.
  • the display 13 may be called a display screen or a display unit.
  • the display 13 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an organic light-emitting diode (Organic Light-Emitting Diode, OLED) touch device, and the like.
  • the display 13 is used for displaying the information processed in the electronic device 1 and for displaying a visualized working interface, such as displaying the results of statistical data.
  • the network interface 14 may optionally include a standard wired interface or wireless interface (such as a WI-FI interface), and the network interface 14 is generally used to establish a communication connection between the electronic device 1 and other electronic devices.
  • a standard wired interface or wireless interface such as a WI-FI interface
  • FIG. 3 only shows the electronic device 1 with components 11-14 and the artificial intelligence-based search intent recognition program 10, but it should be understood that it is not required to implement all the components shown, and more or more components may be implemented instead. few components.
  • the electronic device 1 may further include a user interface, and the user interface may include a display (Display), an input unit such as a keyboard (Keyboard), and an optional user interface may also include a standard wired interface and a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an organic light-emitting diode (Organic Light-Emitting Diode, OLED) touch device, and the like.
  • the display may also be appropriately called a display screen or a display unit, and is used for displaying information processed in the electronic device 1 and for displaying a visualized user interface.
  • the electronic device 1 may also include a radio frequency (Radio Frequency, RF) circuit, a sensor, an audio circuit, etc., which will not be repeated here.
  • RF Radio Frequency
  • a word segmentation operation is performed on the query statement, at least one keyword of the query statement after the word segmentation operation is determined, and the entity type of each keyword is extracted;
  • the query sentence is input into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • the storage device may be the memory 11 of the electronic device 1 , or other storage devices connected in communication with the electronic device 1 .
  • the embodiment of the present application also provides a computer-readable storage medium, and the computer-readable storage medium may be non-volatile or volatile.
  • the computer-readable storage medium can be hard disk, multimedia card, SD card, flash memory card, SMC, read-only memory (ROM), erasable programmable read-only memory (EPROM), portable compact disk read-only memory (CD- ROM), USB memory, etc., or any combination of several.
  • the computer-readable storage medium includes a storage data area and a storage program area, the storage data area stores data created according to the use of blockchain nodes, and the storage program area stores an artificial intelligence-based search intent recognition program 10, the When the search intent recognition program 10 based on artificial intelligence is executed by the processor, the following operations are realized:
  • a word segmentation operation is performed on the query statement, at least one keyword of the query statement after the word segmentation operation is determined, and the entity type of each keyword is extracted;
  • the query sentence is input into a pre-established intent recognition model to obtain the target intent of the query sentence.
  • all the above-mentioned data can also be stored in a block chain node.
  • entity types and query statements, etc. these data can be stored in blockchain nodes.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with each other using cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation.
  • the technical solution of the present application can be embodied in the form of a software product in essence or the part that contributes to the prior art, and the computer software product is stored in a storage medium as described above (such as ROM/RAM , magnetic disk, optical disk), including several instructions to enable a terminal device (which may be a mobile phone, computer, electronic device, or network device, etc.) to execute the methods described in various embodiments of the present application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Human Computer Interaction (AREA)
  • Pathology (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请涉及人工智能及智慧医疗技术领域,提供了一种基于人工智能的搜索意图识别方法、装置、设备及存储介质。所述方法包括:通过规则集与查询语句进行匹配,若规则集中未匹配到查询语句的意图时对查询语句执行分词,确定分词操作后的查询语句关键词,并提取各关键词的实体类型,将各关键词的实体类型与标签词典树进行匹配,得到各实体类型对应的标签意图,基于各实体类型的标签意图得到查询语句的目标意图,当各关键词的实体类型与标签词典树均未匹配出标签意图时,将查询语句输入意图识别模型得到查询语句的目标意图。本申请更准确地识别医疗领域中查询语句的搜索意图。本申请还涉及区块链技术领域,上述实体类型可以存储于一区块链的节点中。

Description

基于人工智能的搜索意图识别方法、装置、设备及介质
本申请要求于2021年08月30日提交中国专利局、申请号为202111002384.7,发明名称为“基于人工智能的搜索意图识别方法、装置、设备及介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域,尤其涉及一种基于人工智能的搜索意图识别方法、装置、设备及存储介质。
背景技术
搜索意图识别是指对用户的搜索词条进行分析,获得用户意图与需求,从而向用户推荐用户最需要的产品或内容。因此提高搜索识别意图能够提高产品或内容推荐的准确率。
发明人意识到,目前的搜索意图识别方案大多是采用词向量对搜索词条进行语义表示,词向量基于上下文的含义得到,而在智能医疗领域,用户输入的搜索语句具有不规范且长度不一的情况,获得的词向量表示能力较差,导致搜索意图识别准确率低。
技术解决方案
本申请提供一种基于人工智能的搜索意图识别方法,该方法包括:
将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
本申请还提供一种基于人工智能的搜索意图识别装置,该基于人工智能的搜索意图识别装置包括:
第一匹配模块:用于将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
提取模块:用于当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
第二匹配模块:用于将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
识别模块:用于当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
本申请还提供一种电子设备,所述电子设备包括:
至少一个处理器;以及,
与所述至少一个处理器通信连接的存储器;其中,
所述存储器存储有可被所述至少一个处理器执行的程序,所述程序被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下所述的基于人工智能的搜索意图识别方法:
将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
本申请还提供一种计算机可读存储介质,所述计算机可读存储介质存储有基于人工智能的搜索意图识别程序,所述基于人工智能的搜索意图识别程序被处理器执行时,实现如下所述基于人工智能的搜索意图识别方法:
将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
附图说明
图1为本申请基于人工智能的搜索意图识别方法较佳实施例的流程图示意图;
图2为本申请基于人工智能的搜索意图识别装置较佳实施例的模块示意图;
图3为本申请电子设备较佳实施例的示意图;
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅用以解释本申请,并不用于限定本申请。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
本申请实施例可以基于人工智能技术对相关的数据进行获取和处理。其中,人工智能(ArtificialIntelligence,AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能,感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。
人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互系统、机电一体化等技术。人工智能软件技术主要包括计算机视觉技术、机器人技术、生物识别技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
本申请提供一种基于人工智能的搜索意图识别方法。参照图1所示,为本申请基于人工智能的搜索意图识别方法的实施例的方法流程示意图。该方法可以由一个电子设备执行,该电子设备可以由软件和/或硬件实现。基于人工智能的搜索意图识别方法包括:
步骤S10:将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图。
本方案的应用场景可以是用户在使用具有搜索功能的APP(例如,智能医疗领域的APP)时,对用户的搜索意图进行识别,以反馈用户最想要的搜索结果,例如,识别出用户是想“购药”、“挂号”或者“问诊”的意图,以反馈相关的信息给用户。本方案以智能医疗APP为例对本方案进行说明,需要说明的是,本方案的实际应用场景并不仅限于上述APP,还可以是社交类APP、电商类APP。
在本实施例中,当侦测到用户在安装有智能医疗APP的终端输入查询语句并发起搜索请求时,获取用户在交互界面中输入的查询语句,可以理解的是,用户输入的内容可以是查询语句,也可以是查询词,例如,某种药品的名称、医院的名称或者某种疾病的名称等。之后,将用户输入的查询语句与预设的规则集进行匹配,判断是否匹配到查询语句的目标意图,其中,规则集包括多种常用的规则语句,规则集可以包括“药品规则集”、“医院规则集”等,“药品规则集”可以是指的各种药品的全称,其对应的意图可以是用户想搜索该药品的购买链接或了解药品详细信息、“医院规则集”可以是指医院的全称,其对应的意图可以是用户想在该医院就诊或了解该医院的科室等信息。
在一个实施例中,所述判断在预设的规则集中是否匹配到所述查询语句的目标意图,包括:
当所述查询语句与所述规则集中任一规则语句匹配成功时,将匹配成功的规则语句作为所述查询语句的目标意图;
当所述查询语句与所述规则集中所有规则语句均匹配失败时,判断在所述规则集中未匹配到所述查询语句的目标意图。
当查询语句与规则集中任一规则语句匹配成功时,即查询语句与规则集中的规则语句完成相同,将该规则语句对应的意图作为该查询语句的目标意图,根据该意图向用户反馈相关的搜索结果。若查询语句与规则集中所有规则语句均匹配失败时,则判断在规则集中未匹配到查询语句的目标意图。
例如,用户输入的查询语句是“腹可安片”,且规则集中收录了该药品规则语句,则识别其意图是想购买该药品或了解该药品的详细信息,将该药品的购买链接及该药品的详细信息反馈至用户。
步骤S20:当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型。
在本实施例中,由于收录的规则集不可能覆盖用户所有可能输入的查询语句,因此当利用规则集不能匹配出用户输入的查询语句的目标意图时,可以根据查询语句的关键词来识别该查询语句的目标意图。具体地,当判断在规则集未匹配到查询语句的目标意图时,对查询语句执行分词操作,例如,可以利用正向匹配拆分、逆向匹配拆分或最少切分算法对查询语句进行分词。确定分词操作后的查询语句的至少一个关键词,并根据命名实体识别算法提取各关键词的实体类型。
命名实体识别(NER)是在自然语言处理中的一个经典问题,其应用也较广泛,例如从一句话中识别出人名、地名,从搜索中识别出产品的名字,识别药物名称等等。本方案的实体提取算法可以是条件随机场模型,它是一种判别式概率模型,常用于标注或分析序列资料,如自然语言文字。
在一个实施例中,所述确定分词操作后的查询语句的至少一个关键词,包括:
遍历预先构建的关键词词典树各节点对应的关键词,将所述查询语句执行分词操作后得到的各个分词,分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,当所述各个分词与所述关键词词典树任一节点对应的关键词匹配成功时,则将匹配成功的分词作为所述查询语句的关键词。
查询语句的关键词也可以称为主体词,在医疗健康领域内,业务主体是一个有限集合,主要包括疾病、症状、疗效、治疗方式、器官部位、品牌、商品类目、商品属性、科室、医院名称等一系列名词集合。可以建立词典树来匹配查询语句中出现的关键词,将查询语句的各个分词分别与词典树各节点对应的关键词进行匹配,若分词与关键词词典树的任一节点的关键词匹配成功,则将该分词作为查询语句的关键词。
进一步地,所述分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,包括:
当所述各个分词与所述关键词词典树所有节点对应的关键词均匹配失败时,将所述各个分词转换为拼音并与预先构建的拼音字典树进行匹配,当任一分词的拼音与所述拼音词典树任一节点对应的关键词匹配成功时,将匹配成功的节点对应的关键词作为所述查询语句的关键词;
当所述各个分词的拼音与所述拼音词典树所有节点对应的关键词均匹配失败时,基于编辑距离计算所述各个分词与所述拼音词典树所有节点对应的关键词的差异值,选取差异值最小的关键词作为所述查询语句的关键词。
由于用户输入的查询语句可能包括错别字导致意图识别有误,因此,当分词与关键词词典树节点对应的关键词均匹配失败时,可以对查询语句的分词进行纠错处理,纠错以拼音为优先,因为音同字不同的错误情况较多,并且没有歧义。在建立关键词词典树的同时,还建立对应的关键词拼音词典树,会将匹配不到关键词的分词转换成拼音,在拼音词典树中再匹配一遍,若匹配成功,则将匹配成功的拼音词典树中的关键词作为查询语句的关键词。例如,若分词为“啊莫西淋”,会匹配到“阿莫西林”。
如果拼音词典树中仍未匹配到查询语句分词对应的关键词,则使用编辑距离计算分词与拼音词典树中关键词的差异值,选取差异值最小的关键词作为查询语句的关键词。医疗领域会有很多药品化学名的搜索场景,由于化学名很多是生僻字,用户会用形同音不同的方式去搜索。例如,“二甲双胍胶囊”,大多数用户会输入“二甲双瓜胶囊”,可以通过编辑距离进行纠错。编辑距离是指对二个字符串的差异程度的量化量测,量测方式是看至少需要多少次的处理才能将一个字符串变成另一个字符串。
在一个实施例中,确定分词操作后的查询语句的至少一个关键词,还包括:
计算各个分词在所述查询语句中的词频,基于所述词频计算出各个分词的IDF值及TF值,将各分词的IDF值与各个分词对应的TF值相乘得到各个分词的TF-IDF值,基于各个分词的TF-IDF值选取预设数量的预设词性的词所述查询语句的关键词。
由于一些口语化的疾病名称可能不在词典树中,因此还可以用TF-IDF算法来抽取一些口语化的关键词,例如:拉肚子(腹泻)。计算出IDF(逆文档频率值),然后再计算出问诊文本中每个词的TF(词频)值。其中,TF=(词语在文本中出现次数)/(各词语在文本中出现次数的总和),将IDF值与TF值相乘,得到该词的TF-IDF值,TF-IDF值可以评估字词对于文本中的重要程度,TF-IDF值越大表示作为关键词的优先级越高,若TF-IDF值越大,该词对语句的重要性越高。
在一个实施例中,也可以利用双向匹配算法对查询语句执行分词操作,具体分词步骤包括:
根据正向最大匹配法将读取到的分词与预设词库进行匹配,得到第一匹配结果,所述第一匹配结果中包含有第一数量的第一词组和第二数量的单字;
根据逆向最大匹配法将读取到的分词与预设词库进行匹配,得到第二匹配结果,所述第二匹配结果中包含有第三数量的第二词组和第四数量的单字;
若所述第一数量与所述第三数量相等且所述第二数量小于或者等于所述第四数量,或者,若所述第一数量小于所述第三数量,则将所述第一匹配结果作为该查询语句的分词结果;若所述第一数量与所述第二数量相等且所述第三数量大于所述第四数量,或者,若所述第一数量大于所述第三数量,则将所述第二匹配结果作为该查询语句的分词结果。
通过该分词方法来分析切分文本内容中前后组合内容的粘性,由于通常情况下词组能代表核心观点信息的概率更大,即通过词组更能表达出核心观点信息,因此,通过正反向同时进行分词匹配找出单字数量更少,词组数量更多的分词匹配结果,以作为切分的语句的分词结果,可提高分词的准确性。
步骤S30:将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图。
在本实施例中,利用命名实体识别算法提取出查询语句中各关键词的实体类型之后,将各关键词的实体类型与预先建立的标签词典树进行匹配,标签词典树建立了医疗领域中常用词汇对应的标签意图,每个实体类型有对应的标签意图,例如,实体类型为“品牌”,则对应的标签意图是“购买”的意图。将查询语句关键词的实体类型与标签词典树进行匹配,可以得到各实体类型对应的标签意图,可以选取各实体类型中标签意图数最多的标签作为查询语句的目标意图。
例如,查询语句“汤臣倍健青少年液体钙软胶囊”,“汤臣倍健”的实体类型为“品牌”、“青少年”的实体类型为“人群”、“液体钙”的实体类型为“品类”、“软胶囊”实体类型为“药品名”,与标签词典树匹配的购买商品意图的有“品牌”、“液体钙”、“软件囊”3个词,匹配的问诊意图的有“人群”1个词,由此判断用户输入的查询语句是“购买”的意图。
词典树又称前缀树,是一种有序树,用于保存关联数组,其中的键通常是字符串,键不是直接保存在节点中,而是由节点在树中的位置决定。一个节点的所有子孙都有相同的前缀,也就是这个节点对应的字符串,而根节点对应空字符串。一般情况下,不是所有的节点都有对应的值,只有叶子节点和部分内部节点所对应的键才有相关的值。
在一个实施例中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
当所述用户为第一类型用户时,从各实体类型对应的标签意图中选取数量最大的标签意图作为所述查询语句的目标意图;
当所述用户为第二类型用户时,基于该用户的历史搜索对应的反馈信息,从各实体类型对应的标签意图中选取出所述查询语句的目标意图。
第一类型的用户是指新用户,可以将标签意图数量最多的标签作为查询语句的目标意图,第二类型用户是指老用户,可以根据该用户对历史搜索的反馈信息(即用户历史搜索的点击情况),从各实体类型中标签意图数选取出意图,例如,用户的历史搜索中,点击购买链接的次数是最多的,可以将购买作为该查询语句的目标意图。根据优先级规则,确定不同类型用户查询语句中各实体类型的标签意图的优先级顺序,可以更准确识别出用户输入的查询语句的目标意图。
在一个实施例中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
判断所述各关键词的实体类型中是否存在预设类型的实体类型,当判断存在预设类型的实体类型时,将所述预设类型的实体类型对应的标签意图作为所述查询语句的目标意图。
预设类型的实体类型可以是指“医生姓名”,例如,当关键词存在某个医生的真实姓名,该关键词的实体类型为“医生姓名”,则将“医生姓名”对应的标签意图“问诊”作为查询语句的目标意图,即判断用户是想指定医生问诊,直接返回该医生的搜索结果。可以快速确定用户是否是搜索医生问诊的意图,根据历史的用户点击数据分析,在搜索人名的情况下,90%的用户是点击该名字的医生进行问诊。
步骤S40:当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
在本实施例中,若查询语句关键词的实体类型与标签词典树均未匹配出标签意图时,说明用户输入的查询语句可能偏口语化或者偏方言化,标签词典树未能匹配出意图,因此,可以将查询语句输入预先建立的意图识别模型,得到查询语句的目标意图。其中,意图识别模型可以是根据doc2vec模型进行训练得到的,可以利用历史在线问诊的患者主诉和对应的拟诊结果生成样本集训练doc2vec模型得到意图识别模型。
例如,用户输入的查询语句为“小朋友拉粑粑像蛋花汤一样怎么办”之类的,词典树无法匹配出对应的标签意图,将该查询语句输入意图识别模型,输出的结果为“小儿腹泻”。
在一个实施例中,所述方法还包括:
计算所述查询语句与预设存储路径中各历史查询语句的相似度值,若存在相似度值大于预设阈值的目标历史查询语句,且所述目标历史查询语句的意图与所述意图识别模型识别所述查询语句得到的目标意图相同时,将所述目标历史查询语句的意图作为所述查询语句的目标意图。
将用户输入的查询语句与线上的历史查询语句进行相似度计算,根据相似度值来验证该查询语句的目标意图,若两者的相似度值大于预设阈值(例如,90%),且该历史查询语句的意图与意图识别模型识别查询语句得到的目标意图相同时,将该历史查询语句的意图作为查询语句的目标意图,可以利用历史查询语句的意图对用户输入的查询语句的意图进行验证;
若存在相似度值大于预设阈值的历史查询语句,但该历史查询语句的意图与意图识别模型识别查询语句得到的目标意图不相同,则将意图识别模型识别出的意图作为查询语句的目标意图。
参照图2所示,为本申请基于人工智能的搜索意图识别装置100的功能模块示意图。
本申请所述基于人工智能的搜索意图识别装置100可以安装于电子设备中。根据实现的功能,所述基于人工智能的搜索意图识别装置100可以包括第一匹配模块110、提取模块120、第二匹配模块130及识别模块140。本发所述模块也可以称之为单元,是指一种能够被电子设备处理器所执行,并且能够完成固定功能的一系列计算机程序段,其存储在电子设备的存储器中。
在本实施例中,关于各模块/单元的功能如下:
第一匹配模块110:用于将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图。
提取模块120,用于当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型。
第二匹配模块130,用于将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图。
识别模块140,用于当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
在一个实施例中,所述确定分词操作后的查询语句的至少一个关键词,包括:
遍历预先构建的关键词词典树各节点对应的关键词,将所述查询语句执行分词操作后得到的各个分词,分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,当所述各个分词与所述关键词词典树任一节点对应的关键词匹配成功时,则将匹配成功的分词作为所述查询语句的关键词。
在一个实施例中,所述分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,包括:
当所述各个分词与所述关键词词典树所有节点对应的关键词均匹配失败时,将所述各个分词转换为拼音并与预先构建的拼音字典树进行匹配,当任一分词的拼音与所述拼音词典树任一节点对应的关键词匹配成功时,将匹配成功的节点对应的关键词作为所述查询语句的关键词;
当所述各个分词的拼音与所述拼音词典树所有节点对应的关键词均匹配失败时,基于编辑距离计算所述各个分词与所述拼音词典树所有节点对应的关键词的差异值,选取差异值最小的关键词作为所述查询语句的关键词。
在一个实施例中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
当所述用户为第一类型用户时,从各实体类型对应的标签意图中选取数量最大的标签意图作为所述查询语句的目标意图;
当所述用户为第二类型用户时,基于该用户的历史搜索对应的反馈信息,从各实体类型对应的标签意图中选取出所述查询语句的目标意图。
在一个实施例中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
判断所述各关键词的实体类型中是否存在预设类型的实体类型,当判断存在预设类型的实体类型时,将所述预设类型的实体类型对应的标签意图作为所述查询语句的目标意图。
在一个实施例中,所述判断在预设的规则集中是否匹配到所述查询语句的目标意图,包括:
当所述查询语句与所述规则集中任一规则语句匹配成功时,将匹配成功的规则语句作为所述查询语句的目标意图;
当所述查询语句与所述规则集中所有规则语句均匹配失败时,判断在所述规则集中未匹配到所述查询语句的目标意图。
在一个实施例中,所述识别模块140还用于:
计算所述查询语句与预设存储路径中各历史查询语句的相似度值,若存在相似度值大于预设阈值的目标历史查询语句,且所述目标历史查询语句的意图与所述意图识别模型识别所述查询语句得到的目标意图相同时,将所述目标历史查询语句的意图作为所述查询语句的目标意图。
参照图3所示,为本申请电子设备1较佳实施例的示意图。
该电子设备1包括但不限于:存储器11、处理器12、显示器13及网络接口14。所述电子设备1通过网络接口14连接网络,获取原始数据。其中,所述网络可以是企业内部网(Intranet)、互联网(Internet)、全球移动通讯系统(GlobalSystemofMobilecommunication,GSM)、宽带码分多址(WidebandCodeDivisionMultipleAccess,WCDMA)、4G网络、5G网络、蓝牙(Bluetooth)、Wi-Fi、通话网络等无线或有线网络。
其中,存储器11至少包括一种类型的可读存储介质,所述可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等,所述计算机可读存储介质可以是非易失性的,也可以是易失性的。在一些实施例中,所述存储器11可以是所述电子设备1的内部存储单元,例如该电子设备1的硬盘或内存。在另一些实施例中,所述存储器11也可以是所述电子设备1的外部存储设备,例如该电子设备1配备的插接式硬盘,智能存储卡(SmartMediaCard, SMC),安全数字(SecureDigital, SD)卡,闪存卡(FlashCard)等。当然,所述存储器11还可以既包括所述电子设备1的内部存储单元也包括其外部存储设备。本实施例中,存储器11通常用于存储安装于所述电子设备1的操作系统和各类应用软件,例如基于人工智能的搜索意图识别程序10的程序代码等。此外,存储器11还可以用于暂时地存储已经输出或者将要输出的各类数据。
处理器12在一些实施例中可以是中央处理器(CentralProcessingUnit,CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器12通常用于控制所述电子设备1的总体操作,例如执行数据交互或者通信相关的控制和处理等。本实施例中,所述处理器12用于运行所述存储器11中存储的程序代码或者处理数据,例如运行基于人工智能的搜索意图识别程序10的程序代码等。
显示器13可以称为显示屏或显示单元。在一些实施例中显示器13可以是LED显示器、液晶显示器、触控式液晶显示器以及有机发光二极管(OrganicLight-EmittingDiode,OLED)触摸器等。显示器13用于显示在电子设备1中处理的信息以及用于显示可视化的工作界面,例如显示数据统计的结果。
网络接口14可选地可以包括标准的有线接口、无线接口(如WI-FI接口),该网络接口14通常用于在所述电子设备1与其它电子设备之间建立通信连接。
图3仅示出了具有组件11-14以及基于人工智能的搜索意图识别程序10的电子设备1,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。
可选地,所述电子设备1还可以包括用户接口,用户接口可以包括显示器(Display)、输入单元比如键盘(Keyboard),可选的用户接口还可以包括标准的有线接口、无线接口。可选地,在一些实施例中,显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及有机发光二极管(OrganicLight-EmittingDiode,OLED)触摸器等。其中,显示器也可以适当的称为显示屏或显示单元,用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
该电子设备1还可以包括射频(RadioFrequency,RF)电路、传感器和音频电路等等,在此不再赘述。
在上述实施例中,处理器12执行存储器11中存储的基于人工智能的搜索意图识别程序10时可以实现如下步骤:
将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
所述存储设备可以为电子设备1的存储器11,也可以为与电子设备1通讯连接的其它存储设备。
关于上述步骤的详细介绍,请参照上述图2关于基于人工智能的搜索意图识别装置100实施例的功能模块图以及图1关于基于人工智能的搜索意图识别方法实施例的流程图的说明。
此外,本申请实施例还提出一种计算机可读存储介质,所述计算机可读存储介质可以是非易失性的,也可以是易失性的。该计算机可读存储介质可以是硬盘、多媒体卡、SD卡、闪存卡、SMC、只读存储器(ROM)、可擦除可编程只读存储器(EPROM)、便携式紧致盘只读存储器(CD-ROM)、USB存储器等等中的任意一种或者几种的任意组合。所述计算机可读存储介质中包括存储数据区和存储程序区,存储数据区存储根据区块链节点的使用所创建的数据,存储程序区存储有基于人工智能的搜索意图识别程序10,所述基于人工智能的搜索意图识别程序10被处理器执行时实现如下操作:
将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
本申请之计算机可读存储介质的具体实施方式与上述基于人工智能的搜索意图识别方法的具体实施方式大致相同,在此不再赘述。
在另一个实施例中,本申请所提供的基于人工智能的搜索意图识别方法,为进一步保证上述所有出现的数据的私密和安全性,上述所有数据还可以存储于一区块链的节点中。例如实体类型及查询语句等,这些数据均可存储在区块链节点中。
需要说明的是,本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
需要说明的是,上述本申请实施例序号仅仅为了描述,不代表实施例的优劣。并且本文中的术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、装置、物品或者方法所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端设备(可以是手机,计算机,电子装置,或者网络设备等)执行本申请各个实施例所述的方法。
以上仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种基于人工智能的搜索意图识别方法,应用于电子设备,其中,所述方法包括:
    将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
    当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
    将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
    当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
  2. 如权利要求1所述的基于人工智能的搜索意图识别方法,其中,所述确定分词操作后的查询语句的至少一个关键词,包括:
    遍历预先构建的关键词词典树各节点对应的关键词,将所述查询语句执行分词操作后得到的各个分词,分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,当所述各个分词与所述关键词词典树任一节点对应的关键词匹配成功时,则将匹配成功的分词作为所述查询语句的关键词。
  3. 如权利要求2所述的基于人工智能的搜索意图识别方法,其中,所述分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,包括:
    当所述各个分词与所述关键词词典树所有节点对应的关键词均匹配失败时,将所述各个分词转换为拼音并与预先构建的拼音字典树进行匹配,当任一分词的拼音与所述拼音词典树任一节点对应的关键词匹配成功时,将匹配成功的节点对应的关键词作为所述查询语句的关键词;
    当所述各个分词的拼音与所述拼音词典树所有节点对应的关键词均匹配失败时,基于编辑距离计算所述各个分词与所述拼音词典树所有节点对应的关键词的差异值,选取差异值最小的关键词作为所述查询语句的关键词。
  4. 如权利要求1所述的基于人工智能的搜索意图识别方法,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    当所述用户为第一类型用户时,从各实体类型对应的标签意图中选取数量最大的标签意图作为所述查询语句的目标意图;
    当所述用户为第二类型用户时,基于该用户的历史搜索对应的反馈信息,从各实体类型对应的标签意图中选取出所述查询语句的目标意图。
  5. 如权利要求1所述的基于人工智能的搜索意图识别方法,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    判断所述各关键词的实体类型中是否存在预设类型的实体类型,当判断存在预设类型的实体类型时,将所述预设类型的实体类型对应的标签意图作为所述查询语句的目标意图。
  6. 如权利要求1所述的基于人工智能的搜索意图识别方法,其中,所述判断在预设的规则集中是否匹配到所述查询语句的目标意图,包括:
    当所述查询语句与所述规则集中任一规则语句匹配成功时,将匹配成功的规则语句作为所述查询语句的目标意图;
    当所述查询语句与所述规则集中所有规则语句均匹配失败时,判断在所述规则集中未匹配到所述查询语句的目标意图。
  7. 如权利要求1至6中任意一项所述的基于人工智能的搜索意图识别方法,其中,在将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图之后,所述方法还包括:
    计算所述查询语句与预设存储路径中各历史查询语句的相似度值,若存在相似度值大于预设阈值的目标历史查询语句,且所述目标历史查询语句的意图与所述意图识别模型识别所述查询语句得到的目标意图相同时,将所述目标历史查询语句的意图作为所述查询语句的目标意图。
  8. 一种基于人工智能的搜索意图识别装置,其中,所述装置包括:
    第一匹配模块:用于将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
    提取模块:用于当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
    第二匹配模块:用于将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
    识别模块:用于当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
  9. 一种电子设备,其中,所述电子设备包括:
    至少一个处理器;以及,
    与所述至少一个处理器通信连接的存储器;其中,
    所述存储器存储有可被所述至少一个处理器执行的程序,所述程序被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下所述的基于人工智能的搜索意图识别方法:
    将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
    当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
    将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
    当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
  10. 如权利要求9所述的电子设备,其中,所述确定分词操作后的查询语句的至少一个关键词,包括:
    遍历预先构建的关键词词典树各节点对应的关键词,将所述查询语句执行分词操作后得到的各个分词,分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,当所述各个分词与所述关键词词典树任一节点对应的关键词匹配成功时,则将匹配成功的分词作为所述查询语句的关键词。
  11. 如权利要求10所述的电子设备,其中,所述分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,包括:
    当所述各个分词与所述关键词词典树所有节点对应的关键词均匹配失败时,将所述各个分词转换为拼音并与预先构建的拼音字典树进行匹配,当任一分词的拼音与所述拼音词典树任一节点对应的关键词匹配成功时,将匹配成功的节点对应的关键词作为所述查询语句的关键词;
    当所述各个分词的拼音与所述拼音词典树所有节点对应的关键词均匹配失败时,基于编辑距离计算所述各个分词与所述拼音词典树所有节点对应的关键词的差异值,选取差异值最小的关键词作为所述查询语句的关键词。
  12. 如权利要求9所述的电子设备,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    当所述用户为第一类型用户时,从各实体类型对应的标签意图中选取数量最大的标签意图作为所述查询语句的目标意图;
    当所述用户为第二类型用户时,基于该用户的历史搜索对应的反馈信息,从各实体类型对应的标签意图中选取出所述查询语句的目标意图。
  13. 如权利要求9所述的电子设备,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    判断所述各关键词的实体类型中是否存在预设类型的实体类型,当判断存在预设类型的实体类型时,将所述预设类型的实体类型对应的标签意图作为所述查询语句的目标意图。
  14. 如权利要求9所述的电子设备,其中,所述判断在预设的规则集中是否匹配到所述查询语句的目标意图,包括:
    当所述查询语句与所述规则集中任一规则语句匹配成功时,将匹配成功的规则语句作为所述查询语句的目标意图;
    当所述查询语句与所述规则集中所有规则语句均匹配失败时,判断在所述规则集中未匹配到所述查询语句的目标意图。
  15. 一种计算机可读存储介质,其中,所述计算机可读存储介质存储有基于人工智能的搜索意图识别程序,所述基于人工智能的搜索意图识别程序被处理器执行时,实现如下所述基于人工智能的搜索意图识别方法的步骤:
    将用户输入的查询语句与预设的规则集进行匹配,判断在所述规则集中是否匹配到所述查询语句的目标意图;
    当判断在所述规则集中未匹配到所述查询语句的目标意图时,对所述查询语句执行分词操作,确定分词操作后的查询语句的至少一个关键词,并提取各关键词的实体类型;
    将各关键词的实体类型与预先建立的标签词典树进行匹配,判断在所述标签词典树中是否匹配到实体类型的标签意图,当判断在所述标签词典树中匹配到实体类型的标签意图时,获取匹配成功的实体类型的标签意图,基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图;
    当判断在所述标签词典树中未匹配到实体类型的标签意图时,将所述查询语句输入预先建立的意图识别模型,得到所述查询语句的目标意图。
  16. 如权利要求15所述的计算机可读存储介质,其中,所述确定分词操作后的查询语句的至少一个关键词,包括:
    遍历预先构建的关键词词典树各节点对应的关键词,将所述查询语句执行分词操作后得到的各个分词,分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,当所述各个分词与所述关键词词典树任一节点对应的关键词匹配成功时,则将匹配成功的分词作为所述查询语句的关键词。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述分别将所述各个分词与所述关键词词典树各节点对应的关键词进行匹配,包括:
    当所述各个分词与所述关键词词典树所有节点对应的关键词均匹配失败时,将所述各个分词转换为拼音并与预先构建的拼音字典树进行匹配,当任一分词的拼音与所述拼音词典树任一节点对应的关键词匹配成功时,将匹配成功的节点对应的关键词作为所述查询语句的关键词;
    当所述各个分词的拼音与所述拼音词典树所有节点对应的关键词均匹配失败时,基于编辑距离计算所述各个分词与所述拼音词典树所有节点对应的关键词的差异值,选取差异值最小的关键词作为所述查询语句的关键词。
  18. 如权利要求15所述的计算机可读存储介质,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    当所述用户为第一类型用户时,从各实体类型对应的标签意图中选取数量最大的标签意图作为所述查询语句的目标意图;
    当所述用户为第二类型用户时,基于该用户的历史搜索对应的反馈信息,从各实体类型对应的标签意图中选取出所述查询语句的目标意图。
  19. 如权利要求15所述的计算机可读存储介质,其中,所述基于匹配成功的实体类型的标签意图得到所述查询语句的目标意图,包括:
    判断所述各关键词的实体类型中是否存在预设类型的实体类型,当判断存在预设类型的实体类型时,将所述预设类型的实体类型对应的标签意图作为所述查询语句的目标意图。
  20. 如权利要求15所述的计算机可读存储介质,其中,所述判断在预设的规则集中是否匹配到所述查询语句的目标意图,包括:
    当所述查询语句与所述规则集中任一规则语句匹配成功时,将匹配成功的规则语句作为所述查询语句的目标意图;
    当所述查询语句与所述规则集中所有规则语句均匹配失败时,判断在所述规则集中未匹配到所述查询语句的目标意图。
PCT/CN2022/087820 2021-08-30 2022-04-20 基于人工智能的搜索意图识别方法、装置、设备及介质 WO2023029513A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111002384.7A CN113707300B (zh) 2021-08-30 2021-08-30 基于人工智能的搜索意图识别方法、装置、设备及介质
CN202111002384.7 2021-08-30

Publications (1)

Publication Number Publication Date
WO2023029513A1 true WO2023029513A1 (zh) 2023-03-09

Family

ID=78656595

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/087820 WO2023029513A1 (zh) 2021-08-30 2022-04-20 基于人工智能的搜索意图识别方法、装置、设备及介质

Country Status (2)

Country Link
CN (1) CN113707300B (zh)
WO (1) WO2023029513A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115954108A (zh) * 2023-03-10 2023-04-11 成都数之联科技股份有限公司 一种数据分析异步处理方法及系统及装置及介质
CN116340523A (zh) * 2023-05-30 2023-06-27 北京中关村科金技术有限公司 会话意图识别方法及装置、计算机设备、存储介质、软件
CN116628129A (zh) * 2023-07-21 2023-08-22 南京爱福路汽车科技有限公司 一种汽车配件搜索方法及系统

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113707300B (zh) * 2021-08-30 2024-06-25 康键信息技术(深圳)有限公司 基于人工智能的搜索意图识别方法、装置、设备及介质
CN114595696A (zh) * 2022-03-03 2022-06-07 Oppo广东移动通信有限公司 实体消歧方法、实体消歧装置、存储介质与电子设备
CN115982423A (zh) * 2023-03-18 2023-04-18 北京新发地农产品网络配送中心有限责任公司 商品搜索匹配方法、装置、设备及计算机可读存储介质
CN117725185B (zh) * 2024-02-06 2024-05-07 河北神玥软件科技股份有限公司 智能对话生成方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109376847A (zh) * 2018-08-31 2019-02-22 深圳壹账通智能科技有限公司 用户意图识别方法、装置、终端及计算机可读存储介质
CN109815314A (zh) * 2019-01-04 2019-05-28 平安科技(深圳)有限公司 一种意图识别方法、识别设备及计算机可读存储介质
CN110543633A (zh) * 2019-08-29 2019-12-06 腾讯科技(深圳)有限公司 语句意图识别方法、装置
CN113707300A (zh) * 2021-08-30 2021-11-26 康键信息技术(深圳)有限公司 基于人工智能的搜索意图识别方法、装置、设备及介质

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9002835B2 (en) * 2013-08-15 2015-04-07 Google Inc. Query response using media consumption history
CN105095187A (zh) * 2015-08-07 2015-11-25 广州神马移动信息科技有限公司 一种搜索意图识别方法及装置
CN105677725A (zh) * 2015-12-30 2016-06-15 南京途牛科技有限公司 一种用于旅游垂直搜索引擎的前置解析方法
CN109815492A (zh) * 2019-01-04 2019-05-28 平安科技(深圳)有限公司 一种基于识别模型的意图识别方法、识别设备及介质
CN110647987A (zh) * 2019-08-22 2020-01-03 腾讯科技(深圳)有限公司 一种应用程序中数据的处理方法、装置、电子设备及存储介质
CN110909548B (zh) * 2019-10-10 2024-03-12 平安科技(深圳)有限公司 中文命名实体识别方法、装置及计算机可读存储介质
CN111209374B (zh) * 2020-01-07 2023-06-06 平安科技(深圳)有限公司 数据查询展示方法、装置、计算机系统及可读存储介质
CN111368049B (zh) * 2020-02-26 2024-04-26 京东方科技集团股份有限公司 信息获取方法、装置、电子设备及计算机可读存储介质
CN111428027A (zh) * 2020-02-26 2020-07-17 中国平安财产保险股份有限公司 查询意图确定方法及相关装置
CN112463944B (zh) * 2020-12-22 2023-10-24 安徽商信政通信息技术股份有限公司 一种基于多模型融合的检索式智能问答方法及装置
CN113157890B (zh) * 2021-04-25 2024-06-11 深圳壹账通智能科技有限公司 智能问答方法、装置、电子设备及可读存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109376847A (zh) * 2018-08-31 2019-02-22 深圳壹账通智能科技有限公司 用户意图识别方法、装置、终端及计算机可读存储介质
CN109815314A (zh) * 2019-01-04 2019-05-28 平安科技(深圳)有限公司 一种意图识别方法、识别设备及计算机可读存储介质
CN110543633A (zh) * 2019-08-29 2019-12-06 腾讯科技(深圳)有限公司 语句意图识别方法、装置
CN113707300A (zh) * 2021-08-30 2021-11-26 康键信息技术(深圳)有限公司 基于人工智能的搜索意图识别方法、装置、设备及介质

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115954108A (zh) * 2023-03-10 2023-04-11 成都数之联科技股份有限公司 一种数据分析异步处理方法及系统及装置及介质
CN115954108B (zh) * 2023-03-10 2023-05-16 成都数之联科技股份有限公司 一种数据分析异步处理方法及系统及装置及介质
CN116340523A (zh) * 2023-05-30 2023-06-27 北京中关村科金技术有限公司 会话意图识别方法及装置、计算机设备、存储介质、软件
CN116340523B (zh) * 2023-05-30 2023-09-26 北京中关村科金技术有限公司 会话意图识别方法及装置、计算机设备、存储介质、软件
CN116628129A (zh) * 2023-07-21 2023-08-22 南京爱福路汽车科技有限公司 一种汽车配件搜索方法及系统
CN116628129B (zh) * 2023-07-21 2024-02-27 南京爱福路汽车科技有限公司 一种汽车配件搜索方法及系统

Also Published As

Publication number Publication date
CN113707300A (zh) 2021-11-26
CN113707300B (zh) 2024-06-25

Similar Documents

Publication Publication Date Title
WO2023029513A1 (zh) 基于人工智能的搜索意图识别方法、装置、设备及介质
WO2021000676A1 (zh) 问答方法、问答装置、计算机设备及存储介质
Alzahrani et al. Understanding plagiarism linguistic patterns, textual features, and detection methods
US9621601B2 (en) User collaboration for answer generation in question and answer system
WO2020253725A1 (zh) 一种药品推荐方法、电子设备和计算机可读存储介质
Yang et al. Joint relational embeddings for knowledge-based question answering
US9146987B2 (en) Clustering based question set generation for training and testing of a question and answer system
JP5936698B2 (ja) 単語意味関係抽出装置
CA2779349C (en) Predictive analysis by example
JP2020123318A (ja) テキスト相関度を確定するための方法、装置、電子機器、コンピュータ可読記憶媒体及びコンピュータプログラム
WO2021135469A1 (zh) 基于机器学习的信息抽取方法、装置、计算机设备及介质
US11250035B2 (en) Knowledge graph generating apparatus, method, and non-transitory computer readable storage medium thereof
CN112015900B (zh) 医学属性知识图谱构建方法、装置、设备及介质
US20220284174A1 (en) Correcting content generated by deep learning
WO2021213160A1 (zh) 基于图神经网络的医疗查询方法、装置、计算机设备及存储介质
CN111401066A (zh) 基于人工智能的词分类模型训练方法、词处理方法及装置
WO2021151270A1 (zh) 图像结构化数据提取方法、装置、设备及存储介质
CN112287069B (zh) 基于语音语义的信息检索方法、装置及计算机设备
CN109522397B (zh) 信息处理方法及装置
WO2022105493A1 (zh) 基于语义识别的数据查询方法、装置、设备及存储介质
US11080615B2 (en) Generating chains of entity mentions
Rodrigues et al. Advanced applications of natural language processing for performing information extraction
WO2023029510A1 (zh) 基于人工智能的远程问诊方法、装置、设备及介质
CN111986759A (zh) 电子病历的解析方法、系统、计算机设备与可读存储介质
CN114036902A (zh) 医学实体预标注方法、装置、存储介质及电子设备

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: 22862659

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE