CN113468420A - Method and system for recommending products - Google Patents
Method and system for recommending products Download PDFInfo
- Publication number
- CN113468420A CN113468420A CN202110726350.6A CN202110726350A CN113468420A CN 113468420 A CN113468420 A CN 113468420A CN 202110726350 A CN202110726350 A CN 202110726350A CN 113468420 A CN113468420 A CN 113468420A
- Authority
- CN
- China
- Prior art keywords
- product
- conversation
- features
- classification model
- candidate
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000013145 classification model Methods 0.000 claims abstract description 47
- 230000003068 static effect Effects 0.000 claims abstract description 27
- 238000012163 sequencing technique Methods 0.000 claims abstract description 19
- 230000008569 process Effects 0.000 claims description 12
- 238000009223 counseling Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000009193 crawling Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007667 floating Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 102100038367 Gremlin-1 Human genes 0.000 description 1
- 101001032872 Homo sapiens Gremlin-1 Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 210000001520 comb Anatomy 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000013486 operation strategy Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9538—Presentation of query results
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The application relates to a method and a system for recommending products, wherein the method for recommending the products comprises the following steps: obtaining conversation content, extracting conversation characteristics in the conversation content, obtaining client static characteristics in an intelligent conversation system, and splicing the conversation characteristics and the client static characteristics; inputting a splicing result into a classification model, and recalling a plurality of first candidate products by the classification model, wherein when the classification model is trained, the classification model is trained according to the conversation features and the product selection results in the historical conversation contents and according to the static features of customers; according to the method and the system for recommending the products, the first candidate products are sequenced according to the preset product recommending strategy, and the recommended products are determined according to the product sequencing.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method and a system for recommending a product.
Background
With the rapid development of artificial intelligence technology, intelligent dialogue systems have come into operation, and most of the current intelligent dialogue systems are used for answering customer questions, for example, in intelligent marketing type intelligent dialogue systems, under the condition that a customer consults an intelligent system for a specified product, the intelligent dialogue systems recall similar consultation sentences according to the consultation contents of the customer and the similarity, determine reply contents corresponding to the consultation sentences, and feed the reply contents back to the customer.
In addition to a simple question and answer function, in the related art, the intelligent dialogue system can actively recommend products to the client, but because the content of the client representation system of the intelligent dialogue system is thin, and the static characteristics of the client are only the surname, the gender and the mobile phone number of the client at many times, in the related art, the intelligent dialogue system presets a target product to be recommended and a recommendation process, and recommends the same target product to different clients, but the matching degree of the recommended product and the client is low, and the probability of successful product recommendation is low.
Aiming at the problem that the success rate of product recommendation is low when an intelligent dialogue system recommends products to customers in the related art, an effective solution is not provided yet.
Disclosure of Invention
The embodiment of the application provides a method and a system for recommending products, which at least solve the problem that when an intelligent dialogue system recommends products to a client in the related art, the success rate of recommending the products is low.
In a first aspect, an embodiment of the present application provides a method for recommending a product, which is applied to an intelligent dialog system, and the method includes:
obtaining conversation content, extracting conversation characteristics in the conversation content, obtaining client static characteristics in the intelligent conversation system, and splicing the conversation characteristics and the client static characteristics;
inputting a splicing result into a classification model, and recalling a plurality of first candidate products by the classification model, wherein when the classification model is trained, the classification model is trained according to conversation features and product selection results in historical conversation contents and according to the static features of customers;
and sequencing the plurality of first candidate products according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
In some embodiments, the dialog features include intention classification results, and the process of extracting the dialog features in the dialog content includes:
splitting the conversation content into a plurality of clauses, determining whether the clauses exist in a question and sentence pattern list, and if so, determining an intention classification result of the clauses according to the question and sentence pattern list;
if not, inputting the clauses into a classification model, and outputting the intention classification result of the clauses by the classification model.
In some embodiments, the intention classification result includes a product type and a product attribute consulted by the client, and after extracting the dialog feature in the dialog content, the method includes:
determining product names corresponding to the conversation features according to the conversation features and the association relation among prestored product types, product names and product attributes to obtain a plurality of second candidate products;
summarizing the first candidate product and the second candidate product to obtain a total candidate product set;
and sequencing the plurality of candidate product total sets according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
In some of these embodiments, the method comprises:
after the conversation content is obtained, extracting the information of the conversation content in a conversation turn, and filling the information into a preset slot position of a global conversation information table;
and before determining whether the clause exists in the list of the question sentence pattern, determining whether the clause structure is complete, and if not, completing the clause according to the information of the historical turns in the global information table.
In some embodiments, after the splitting of the dialog content into a plurality of clauses and before the determining whether the clause is present in a list of question and grammar sentences, the method comprises: and extracting the entity in the clause, determining the entity type of the entity, and replacing the entity in the clause with the entity type.
In some embodiments, after the extracting the entity in the clause, the method further comprises: and determining whether the description of the entity meets a preset standard, and if not, changing the description of the entity into a standard description in a sentence.
In a second aspect, an embodiment of the present application provides a system for product recommendation, which is applied to an intelligent dialog system, and the system includes:
the acquisition module is used for acquiring conversation content, extracting conversation features in the conversation content, acquiring client static features in the intelligent conversation system, and splicing the conversation features and the client static features;
the system comprises a recall module, a classification module and a display module, wherein the recall module is used for inputting splicing results into a classification model, and the classification model recalls a plurality of first candidate products, and when the classification model is trained, the classification model is trained according to conversation features and product selection results in historical conversation contents and according to the static features of customers;
and the determining module is used for sequencing the plurality of first candidate products according to a preset product recommendation strategy and determining recommended products according to the product sequencing.
In some embodiments, the dialog features include intention classification results, and in the obtaining module, the extracting the dialog features in the dialog content includes:
splitting the conversation content into a plurality of clauses, determining whether the clauses exist in a question and sentence pattern list, and if so, determining an intention classification result of the clauses according to the question and sentence pattern list;
if not, inputting the clauses into a classification model, and outputting the intention classification result of the clauses by the classification model.
In some of these embodiments, the intent classification results include product types and product attributes consulted by the client;
the recall module is further configured to, after extracting the conversation features in the conversation content, determine product names corresponding to the conversation features according to the conversation features and according to pre-stored association relations among product types, product names and product attributes, and obtain a plurality of second candidate products;
the determining module is further configured to sum the first candidate product and the second candidate product to obtain a total set of candidate products; and sequencing the plurality of candidate product total sets according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
In some embodiments, the obtaining module is further configured to:
after the conversation content is obtained, extracting the information of the conversation content in a conversation turn, and filling the information into a preset slot position of a global conversation information table;
and before determining whether the clause exists in the list of the question sentence pattern, determining whether the clause structure is complete, and if not, completing the clause according to the information of the historical turns in the global information table.
Compared with the related art, the product recommendation method provided by the embodiment of the application extracts the dialogue features in the dialogue content by acquiring the dialogue content, and acquires the client static features in the intelligent dialogue system, and splices the dialogue features and the client static features; inputting a splicing result into a classification model, and recalling a plurality of first candidate products by the classification model, wherein when the classification model is trained, the classification model is trained according to the conversation features and the product selection results in the historical conversation contents and according to the static features of customers; according to the method and the system, the plurality of first candidate products are ranked according to the preset product recommendation strategy, and the recommended products are determined according to the product ranking, so that the problem that when the intelligent dialogue system recommends products to clients in the related art, the success rate of product recommendation is low is solved, and the marketing efficiency of companies is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a schematic diagram of an application environment of a method for product recommendation according to an embodiment of the present application;
FIG. 2 is a flow chart of a method of product recommendation according to a first embodiment of the present application;
FIG. 3 is a flow chart of extracting dialog features in dialog content according to a second embodiment of the present application;
FIG. 4 is a flow chart of a method of product recommendation according to a third embodiment of the present application;
FIG. 5 is a flow chart of a method of product recommendation according to a fourth embodiment of the present application;
FIG. 6 is a flow chart of a method of product recommendation according to a fifth embodiment of the present application;
fig. 7 is a block diagram of a system for product recommendation according to a sixth embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments provided in the present application without any inventive step are within the scope of protection of the present application.
It is obvious that the drawings in the following description are only examples or embodiments of the present application, and that it is also possible for a person skilled in the art to apply the present application to other similar contexts on the basis of these drawings without inventive effort. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of ordinary skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms referred to herein shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar words throughout this application are not to be construed as limiting in number, and may refer to the singular or the plural. The present application is directed to the use of the terms "including," "comprising," "having," and any variations thereof, which are intended to cover non-exclusive inclusions; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or elements, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Reference to "connected," "coupled," and the like in this application is not intended to be limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as referred to herein means two or more. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. Reference herein to the terms "first," "second," "third," and the like, are merely to distinguish similar objects and do not denote a particular ordering for the objects.
The method for recommending a product provided by the present application can be applied to an application environment shown in fig. 1, fig. 1 is a schematic diagram of an application environment of the method for recommending a product according to an embodiment of the present application, as shown in fig. 1, a terminal 101 is deployed with an intelligent dialog system, a server 102 obtains dialog contents of the terminal 101 through a network and runs the method for recommending a product to obtain a product recommendation result, the intelligent dialog system obtains the product recommendation result to recommend a product to a client, the terminal 101 can be, but is not limited to, various dialog robots, personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, and the server 102 can be implemented by an independent server or a server cluster formed by a plurality of servers.
The present embodiment provides a method for recommending a product, which is applied to an intelligent dialog system, and fig. 2 is a flowchart of a method for recommending a product according to a first embodiment of the present application, as shown in fig. 2, the flowchart includes the following steps:
step S201, obtaining conversation content, extracting conversation features in the conversation content, obtaining client static features in an intelligent conversation system, splicing the conversation features and the client static features, wherein after the conversation content is obtained, a conversation text can be corrected firstly, then the conversation text is input into a pre-trained FastText word vector model, a vector corresponding to each token is returned, and the vector is used as token embedding of a subsequent model;
step S202, inputting a splicing result into a classification model, and recalling a plurality of first candidate products by the classification model, wherein when the classification model is trained, the classification model is trained according to a conversation characteristic and a product selection result in historical conversation content and according to a static characteristic of a client;
step S203, according to a preset product recommendation policy, ranking the plurality of first candidate products, and determining a recommended product according to the product ranking, where the recommendation policy relates to product attributes, customer attributes, marketing policies, operation policies, and the like, and for example, the recommendation policy is set as: and (4) recommending the product with the marketing strategy in the top priority, and ranking the product A at the top according to the recommending strategy under the condition that the product A has a unary starting action in the current period.
Through steps S201 to S203, compared to the problem that the success rate of product recommendation is low when the intelligent dialog system recommends a product to a client in the related art, the present embodiment fully utilizes the chance of direct contact to reach the client, explores new relevant features of the client from the dialog content of the dynamically changed client and the intelligent dialog system, and combines the new relevant features of the client with the static features of the client, so that the matching degree of the recommended product obtained in the present embodiment with the client is higher, secondly, because the client is in continuous change, compared with the static features of the client, the dialog process reflects the latest situation of the client, the real-time performance is stronger, so that the product recommended to the client is closer to the current situation of the client, thirdly, the present embodiment trains a classification model according to the dialog features and product selection results in the historical dialog content, and fully utilizes the historical data which have been recommended to other clients and successfully selected by other clients in the intelligent dialog system, product recommendation is performed on the basis of the historical data, the recommended products are more likely to be adopted by the customers, and therefore the success rate of product recommendation of the intelligent dialogue system is improved, the problem that the success rate of product recommendation is low when the intelligent dialogue system recommends the products to the customers in the related art is solved, and the marketing efficiency of companies is improved.
In some of the embodiments, the dialog feature includes an intention classification result, fig. 3 is a flowchart of extracting a dialog feature in dialog content according to the second embodiment of the present application, and as shown in fig. 3, the flowchart includes the following steps:
step S301, splitting the dialog content into a plurality of clauses, and determining whether the clause exists in the list of the question-and-law sentence pattern, if so, determining an intention classification result of the clause according to the list of the question-and-law sentence pattern, where the intention classification result may be a hierarchical structure, for example, the counseling category (first layer) includes counseling product information (second layer), and the counseling product information (second layer) includes counseling interest (third layer);
step S302, if not, inputting clauses into a classification model, and outputting the intention classification result of the clauses by the classification model, wherein the classification model can be an architecture combining LSTM and Attention, thereby fully taking key information and context sequence information into consideration.
Through steps S301 to S302, the embodiment combs out the question-sentence patterns with different intentions in advance, and realizes the precise matching of the conversation content, so that the real intentions behind the language of the client can be deeply mined in the conversation process with the client, and the matching degree of the product recommendation result obtained by the embodiment with the client is high.
In some of these embodiments, after splitting the dialog content into multiple clauses, before determining whether a clause exists in the list of question-sentences: the entity in the clause can be extracted, the entity type of the entity is determined, and the entity in the clause is replaced by the entity type; by replacing the entities in the clauses with entity types, the generalization capability of the model is improved, so that the model focuses more on the relationship among the entity types rather than a single entity.
Further, after the entity in the sentence is extracted, whether the description of the entity meets a preset standard needs to be determined, and if not, the description of the entity in the sentence is changed into a standard description; for example, unifying "build bank" and "build in" into "bank of chinese construction", unifying "tomorrow" into specific date of the next day, and the like, and for example, do you go for business in the next week if the conversation content acquired to the client is "do you go for business in the next week? "in case, the intelligent dialogue system unifies the next week into a specific time period, and finds out the transaction deadline date of the product in a pre-stored product knowledge graph database, and determines whether the time period is before the transaction deadline date; by carrying out normalization processing on the identified entity, the meaning of a sentence can be understood more accurately, the condition that a customer cannot answer the entity by changing the kind of expression is avoided, smooth communication between the customer and an intelligent dialogue system is guaranteed, and meanwhile, the probability of finding matched data in pre-stored data can be improved when product recommendation is carried out, so that a proper product is recommended for the customer.
Since the intention itself also implies the selection tendency of the client to the product, in some embodiments, the intention classification result includes the product type and product attribute consulted by the client, and these product-related intentions can be utilized to make candidate product recalls, fig. 4 is a flowchart of a method for recommending the product according to the third embodiment of the present application, and as shown in fig. 4, after extracting the dialog features in the dialog content, the flowchart includes the following steps:
step S401, determining product names corresponding to the conversation features according to the conversation features and the pre-stored association relationship among the product types, the product names and the product attributes, to obtain a plurality of second candidate products, optionally, the association relationship among the product types, the product names and the product attributes may be pre-stored in a knowledge graph, and the construction process of the knowledge graph includes: firstly, a knowledge graph data structure design is carried out, a data model of Neo4j can be adopted, a graph is constructed based on labels and attributes, for example, entities are divided into different types such as deposit, financing, loan and credit card, the different types of entities have different attributes correspondingly, and deposit products mainly comprise attributes such as product names, organization names, applicable groups, upper floating of standard rate of a more central bank, upper floating of amount of money, highest interest rate, highest deposit duration and upper line; secondly, crawling knowledge map data sources through official pages and other channels of a business party, and manually confirming data which can be put in storage after preliminary arrangement; thirdly, data warehousing is performed, the data storage can use an open source database HugeGraph, and the bottom layer supports distribution based on the TinkerPop3 framework of Apache; after the knowledge graph database is constructed, a Gremlin language is adopted at the bottom layer based on HugeGraph, Java Client is used for inquiring, and product names corresponding to conversation characteristics are determined according to the association relation among product types, product names and product attributes in the knowledge graph to obtain a plurality of second candidate products;
step S402, summarizing the first candidate product and the second candidate product to obtain a total candidate product set;
step S403, sorting the plurality of candidate product total sets according to a preset product recommendation strategy, and determining recommended products according to the product sorting.
Through steps S401 to S403, in this embodiment, in consideration of the fact that there is a case where there is little or no historical conversation content related to a part of products when training a classification model, in addition to obtaining a first candidate product according to a selection result of a conversation feature and historical conversation content, a second candidate product conforming to the conversation feature is obtained according to the obtained conversation feature and data in a pre-stored product knowledge graph, and the first candidate product and the second candidate product are summarized to obtain a total set of candidate products, so that the candidate products are enriched, and when a plurality of total sets of candidate products are ranked according to a preset product recommendation policy, a more appropriate recommended product can be determined.
Considering that a complete expression of the intention of a client sometimes involves multiple rounds of conversations, and many conversations are performed based on the context records of previous conversations, this results in part of the information being expressed by the omission of the client, which is sometimes referred to as pronouns and sometimes omitted as pronouns, for example, a robot introduces a loan product, the client asks "how much interest rate is", after the robot answers, the client says "the interest rate is high", the real intention is to want a product with a lower interest rate than the product, the first step is to let the intelligent dialog system really know the real intention of the client, and the first step is information completion, in some of these examples, fig. 5 is a flow chart of a method for product recommendation according to a fourth example of the present application, as shown in fig. 5, the flow chart includes the following steps:
step S501, after obtaining the conversation content, extracting the conversation content information in a conversation turn, and filling the information into a preset slot position of a global conversation information table;
step S502, before determining whether the clause exists in the list of the question sentence pattern, determining whether the clause structure is complete, if not, completing the clause according to the information of the historical turns in the global information table.
Through the steps 501 to 502, slot filling is performed on each round of client conversation in the global conversation information table, so that on one hand, the change of the global intention of the client can be conveniently tracked, the situation that the semantics of a sentence is unknown when the sentence is recognized due to the fact that the client omits information is avoided, on the other hand, the latest intention of the client can be conveniently known, optionally, a product can be used as an entry point in the whole conversation process, conversation is continuously performed around the intention before the intention of the client on the global information table is transferred, the content of a new round of conversation is continuously acquired, and the content information is updated to the global conversation information table, so that products which are more suitable for the client can be continuously recommended to the client according to the global information table.
In some embodiments, fig. 6 is a flowchart of a method for recommending a product according to a fifth embodiment of the present application, and as shown in fig. 6, in each round of dialog, dialog speech is obtained, a text is obtained after ASR, the dialog text is understood, the understanding process includes text error correction, embedding extraction, named entity recognition and intention recognition, a relevant product recall is performed in a knowledge graph database of the product according to intention classification features obtained by the intention recognition, the recall process includes Schema (Schema) design, data crawling, data storage and data query, finally, the recalled products are ranked, product attributes, client attributes, marketing strategies, operation strategies and the like are synthesized, and then the round of dialog information is updated in a global dialog information table for global intention tracking.
An embodiment of the present application further provides a product recommendation system, which is applied to an intelligent dialog system, fig. 7 is a block diagram of a product recommendation system according to a sixth embodiment of the present application, and as shown in fig. 7, the system includes:
the acquisition module 701 is used for acquiring the conversation content, extracting the conversation features in the conversation content, acquiring the client static features in the intelligent conversation system, and splicing the conversation features and the client static features;
a recall module 702, configured to input a splicing result into a classification model, and recall a plurality of first candidate products by the classification model, where, when training the classification model, the classification model is trained according to a conversation feature and a product selection result in historical conversation content, and according to a customer static feature;
the determining module 703 is configured to rank the plurality of first candidate products according to a preset product recommendation policy, and determine a recommended product according to the product rank.
In some embodiments, the dialog features include intention classification results, and in the obtaining module 701, the process of extracting the dialog features in the dialog content includes:
splitting the conversation content into a plurality of clauses, determining whether the clauses exist in a question and sentence pattern list, and if so, determining an intention classification result of the clauses according to the question and sentence pattern list;
if not, inputting the clauses into the classification model, and outputting the intention classification result of the clauses by the classification model.
In some of these embodiments, the intent classification results include the product type and product attributes consulted by the client;
the recall module 702 is further configured to, after extracting the conversation features in the conversation content, determine product names corresponding to the conversation features according to the conversation features and according to the pre-stored association relationship among the product types, the product names, and the product attributes, so as to obtain a plurality of second candidate products;
the determining module 703 is further configured to sum up the first candidate product and the second candidate product to obtain a total set of candidate products; and sequencing the plurality of candidate product total sets according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
In some embodiments, the obtaining module 701 is further configured to:
after the conversation content is obtained, in a conversation turn, extracting the conversation content information, and filling the information into a preset slot position of a global conversation information table;
and determining whether the clause structure is complete before determining whether the clause exists in the list of the question sentence pattern, and if not, completing the clause according to the information of the historical turns in the global information table.
It should be understood by those skilled in the art that various features of the above-described embodiments can be combined in any combination, and for the sake of brevity, all possible combinations of features in the above-described embodiments are not described in detail, but rather, all combinations of features which are not inconsistent with each other should be construed as being within the scope of the present disclosure.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for recommending products is applied to an intelligent dialogue system, and is characterized by comprising the following steps:
obtaining conversation content, extracting conversation characteristics in the conversation content, obtaining client static characteristics in the intelligent conversation system, and splicing the conversation characteristics and the client static characteristics;
inputting a splicing result into a classification model, and recalling a plurality of first candidate products by the classification model, wherein when the classification model is trained, the classification model is trained according to conversation features and product selection results in historical conversation contents and according to the static features of customers;
and sequencing the plurality of first candidate products according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
2. The method of claim 1, wherein the conversational features comprise intent classification results, and wherein extracting conversational features in the conversational content comprises:
splitting the conversation content into a plurality of clauses, determining whether the clauses exist in a question and sentence pattern list, and if so, determining an intention classification result of the clauses according to the question and sentence pattern list;
if not, inputting the clauses into a classification model, and outputting the intention classification result of the clauses by the classification model.
3. The method of claim 2, wherein the intention classification result includes a product type and a product attribute of a customer consultation, and after extracting the dialog feature in the dialog content, the method includes:
determining product names corresponding to the conversation features according to the conversation features and the association relation among prestored product types, product names and product attributes to obtain a plurality of second candidate products;
summarizing the first candidate product and the second candidate product to obtain a total candidate product set;
and sequencing the plurality of candidate product total sets according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
4. The method of claim 2, wherein the method comprises:
after the conversation content is obtained, extracting the information of the conversation content in a conversation turn, and filling the information into a preset slot position of a global conversation information table;
and before determining whether the clause exists in the list of the question sentence pattern, determining whether the clause structure is complete, and if not, completing the clause according to the information of the historical turns in the global information table.
5. The method of claim 2, wherein after the splitting of the dialog content into a plurality of clauses and before the determining whether the clause is present in a list of question sentences, the method comprises: and extracting the entity in the clause, determining the entity type of the entity, and replacing the entity in the clause with the entity type.
6. The method of claim 5, wherein after the extracting the entities in the clauses, the method further comprises: and determining whether the description of the entity meets a preset standard, and if not, changing the description of the entity into a standard description in a sentence.
7. A system for recommending products, which is applied to an intelligent dialogue system, and is characterized in that the system comprises:
the acquisition module is used for acquiring conversation content, extracting conversation features in the conversation content, acquiring client static features in the intelligent conversation system, and splicing the conversation features and the client static features;
the system comprises a recall module, a classification module and a display module, wherein the recall module is used for inputting splicing results into a classification model, and the classification model recalls a plurality of first candidate products, and when the classification model is trained, the classification model is trained according to conversation features and product selection results in historical conversation contents and according to the static features of customers;
and the determining module is used for sequencing the plurality of first candidate products according to a preset product recommendation strategy and determining recommended products according to the product sequencing.
8. The system according to claim 7, wherein the dialog features include intention classification results, and in the obtaining module, the process of extracting the dialog features in the dialog content includes:
splitting the conversation content into a plurality of clauses, determining whether the clauses exist in a question and sentence pattern list, and if so, determining an intention classification result of the clauses according to the question and sentence pattern list;
if not, inputting the clauses into a classification model, and outputting the intention classification result of the clauses by the classification model.
9. The system of claim 8, wherein the intent classification results include product types and product attributes consulted by the client;
the recall module is further configured to, after extracting the conversation features in the conversation content, determine product names corresponding to the conversation features according to the conversation features and according to pre-stored association relations among product types, product names and product attributes, and obtain a plurality of second candidate products;
the determining module is further configured to sum the first candidate product and the second candidate product to obtain a total set of candidate products; and sequencing the plurality of candidate product total sets according to a preset product recommendation strategy, and determining a recommended product according to the product sequencing.
10. The system of claim 8, wherein the acquisition module is further configured to:
after the conversation content is obtained, extracting the information of the conversation content in a conversation turn, and filling the information into a preset slot position of a global conversation information table;
and before determining whether the clause exists in the list of the question sentence pattern, determining whether the clause structure is complete, and if not, completing the clause according to the information of the historical turns in the global information table.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110726350.6A CN113468420B (en) | 2021-06-29 | 2021-06-29 | Product recommendation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110726350.6A CN113468420B (en) | 2021-06-29 | 2021-06-29 | Product recommendation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113468420A true CN113468420A (en) | 2021-10-01 |
| CN113468420B CN113468420B (en) | 2024-04-05 |
Family
ID=77873738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110726350.6A Active CN113468420B (en) | 2021-06-29 | 2021-06-29 | Product recommendation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113468420B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150178371A1 (en) * | 2013-12-23 | 2015-06-25 | 24/7 Customer, Inc. | Systems and methods for facilitating dialogue mining |
| US20180189857A1 (en) * | 2017-01-05 | 2018-07-05 | Microsoft Technology Licensing, Llc | Recommendation through conversational ai |
| US20180213284A1 (en) * | 2017-01-24 | 2018-07-26 | Adobe Systems Incorporated | Recommending content based on group collaboration |
| CN109146610A (en) * | 2018-07-16 | 2019-01-04 | 众安在线财产保险股份有限公司 | It is a kind of intelligently to insure recommended method, device and intelligence insurance robot device |
| US20190050494A1 (en) * | 2017-08-08 | 2019-02-14 | Accenture Global Solutions Limited | Intelligent humanoid interactive content recommender |
| CN109767318A (en) * | 2018-12-15 | 2019-05-17 | 深圳壹账通智能科技有限公司 | Loan product recommended method, device, equipment and storage medium |
| US20190243899A1 (en) * | 2018-02-07 | 2019-08-08 | Rulai, Inc. | Method and system for a chat box eco-system in a federated architecture |
| CN110738545A (en) * | 2019-08-30 | 2020-01-31 | 深圳壹账通智能科技有限公司 | Product recommendation method and device based on user intention identification, computer equipment and storage medium |
| CN112115246A (en) * | 2020-08-14 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Content recommendation method and device based on conversation, computer equipment and storage medium |
| CN112581203A (en) * | 2019-09-30 | 2021-03-30 | 微软技术许可有限责任公司 | Providing explanatory product recommendations in a session |
| CN112925892A (en) * | 2021-03-23 | 2021-06-08 | 苏州大学 | Conversation recommendation method and device, electronic equipment and storage medium |
-
2021
- 2021-06-29 CN CN202110726350.6A patent/CN113468420B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150178371A1 (en) * | 2013-12-23 | 2015-06-25 | 24/7 Customer, Inc. | Systems and methods for facilitating dialogue mining |
| US20180189857A1 (en) * | 2017-01-05 | 2018-07-05 | Microsoft Technology Licensing, Llc | Recommendation through conversational ai |
| US20180213284A1 (en) * | 2017-01-24 | 2018-07-26 | Adobe Systems Incorporated | Recommending content based on group collaboration |
| US20190050494A1 (en) * | 2017-08-08 | 2019-02-14 | Accenture Global Solutions Limited | Intelligent humanoid interactive content recommender |
| US20190243899A1 (en) * | 2018-02-07 | 2019-08-08 | Rulai, Inc. | Method and system for a chat box eco-system in a federated architecture |
| CN109146610A (en) * | 2018-07-16 | 2019-01-04 | 众安在线财产保险股份有限公司 | It is a kind of intelligently to insure recommended method, device and intelligence insurance robot device |
| CN109767318A (en) * | 2018-12-15 | 2019-05-17 | 深圳壹账通智能科技有限公司 | Loan product recommended method, device, equipment and storage medium |
| CN110738545A (en) * | 2019-08-30 | 2020-01-31 | 深圳壹账通智能科技有限公司 | Product recommendation method and device based on user intention identification, computer equipment and storage medium |
| CN112581203A (en) * | 2019-09-30 | 2021-03-30 | 微软技术许可有限责任公司 | Providing explanatory product recommendations in a session |
| WO2021066903A1 (en) * | 2019-09-30 | 2021-04-08 | Microsoft Technology Licensing, Llc | Providing explainable product recommendation in a session |
| CN112115246A (en) * | 2020-08-14 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Content recommendation method and device based on conversation, computer equipment and storage medium |
| CN112925892A (en) * | 2021-03-23 | 2021-06-08 | 苏州大学 | Conversation recommendation method and device, electronic equipment and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| 周纯伊: "对话式音乐推荐技术及系统实现", 中国优秀硕士学位论文全文数据库 * |
| 胡珉;冯俊兰;王燕蒙;闪云香;: "中国移动智能客服系统研究及实现", 电信工程技术与标准化, no. 10 * |
| 薛伟莲;王蕴慧;: "一种基于对话的电子商务推荐系统", 辽宁师范大学学报(自然科学版), no. 02 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113468420B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10853582B2 (en) | Conversational agent | |
| US11960694B2 (en) | Method of using a virtual assistant | |
| US20180082184A1 (en) | Context-aware chatbot system and method | |
| US20190272269A1 (en) | Method and system of classification in a natural language user interface | |
| US20190243886A1 (en) | Methods and systems for improving machine learning performance | |
| US9466294B1 (en) | Dialog management system | |
| WO2018033030A1 (en) | Natural language library generation method and device | |
| CN103970791B (en) | A kind of method, apparatus for recommending video from video library | |
| CN110782318A (en) | Marketing method and device based on audio interaction and storage medium | |
| CN111241237A (en) | Intelligent question and answer data processing method and device based on operation and maintenance service | |
| EP3822814A2 (en) | Human-machine interaction method and apparatus based on neural network | |
| CN112650842A (en) | Human-computer interaction based customer service robot intention recognition method and related equipment | |
| CN114399396A (en) | Insurance product recommendation method and device, computer equipment and storage medium | |
| CN111444321B (en) | Question answering method, device, electronic equipment and storage medium | |
| CN113051389B (en) | Knowledge pushing method and device | |
| CN114706945A (en) | Intention recognition method and device, electronic equipment and storage medium | |
| CN113468420B (en) | Product recommendation method and system | |
| CN113961811B (en) | Event map-based conversation recommendation method, device, equipment and medium | |
| CN111126073A (en) | Semantic retrieval method and device | |
| CN112989046B (en) | Real-time speech prejudging method, device, computer equipment and storage medium | |
| US20230063713A1 (en) | Sentence level dialogue summaries using unsupervised machine learning for keyword selection and scoring | |
| CN109408679A (en) | Method, apparatus, electronic equipment and the storage medium of intelligent management application program | |
| CN114020896A (en) | Intelligent question and answer method, system, electronic equipment and storage medium | |
| CN112330387A (en) | Virtual broker applied to house-watching software | |
| CN113505293B (en) | Information pushing method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |