CN106776701B - Problem determination method and device for item recommendation - Google Patents
Problem determination method and device for item recommendation Download PDFInfo
- Publication number
- CN106776701B CN106776701B CN201611000826.3A CN201611000826A CN106776701B CN 106776701 B CN106776701 B CN 106776701B CN 201611000826 A CN201611000826 A CN 201611000826A CN 106776701 B CN106776701 B CN 106776701B
- Authority
- CN
- China
- Prior art keywords
- item
- article
- semantic features
- node
- semantic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
-
- 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
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0203—Market surveys; Market polls
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0282—Rating or review of business operators or products
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure relates to a problem determination method and a device for item recommendation, wherein the method comprises the following steps: determining an article indicated by the answer to the previous guide question by using a pre-established decision tree according to the answer to the previous guide question; and acquiring the current guide question according to the article indicated by the answer of the previous guide question. The problem determination method for item recommendation provided by the disclosure can solve the problem that when the guidance problem is generated, the relevance between the upper and lower-level problems is low, so that the problem is relatively high in jumping performance, the relevance between the guidance problems can be improved, and the user experience of the answering user is improved.
Description
Technical Field
The present disclosure relates to the field of electronic commerce, and in particular, to a method and an apparatus for determining a problem for item recommendation.
Background
Currently, personalized customization of customers has become a technical hotspot, and for example, items in which the users are interested are recommended to the users. In order to make the recommended items meet the user's expectations, it is often necessary to fully understand the user's preferences before making recommendations, and it is a more efficient way to obtain the user's preferences through questionnaires. Questionnaires are a survey method for a user to answer a number of preset questions, and the user's preference can be obtained based on the answer to each question answered by the user, so that recommendation can be performed according to the user's preference. These problems may be referred to as boot problems. In the prior art, a decision tree algorithm is usually used for determining a guidance question, and the previous question and the next question obtained by the determination only consider how to select an article with the best discrimination to obtain the maximum information amount of the user interest, but do not consider the relevance of the upper layer question and the lower layer question.
Disclosure of Invention
The disclosure provides a problem determination method and device for item recommendation, which are used for solving the problem that when a guidance problem is generated, the relevance between upper and lower-level problems is low, so that the problem is large in leap.
According to a first aspect of embodiments of the present disclosure, there is provided a problem determination method for item recommendation, the method including:
determining an article indicated by the answer of the previous guide question by utilizing a pre-established decision tree according to the answer of the previous guide question, wherein the decision tree is established according to the scoring data of the articles in the total article set and the relation between semantic features;
and acquiring the current guide question according to the article indicated by the answer of the last guide question.
Optionally, the method for establishing the decision tree includes:
obtaining scoring data of the items in the total item set, wherein the scoring data comprises a score of each item in the total item set;
determining a root node of the decision tree according to the scoring data of the items in the total item set, and taking the root node as an upper node;
extracting semantic features of the upper node and semantic features of each item in N item sets, wherein the N item sets are determined according to the scoring data of each item in the total item set, and each item set comprises one or more items;
determining a lower layer node corresponding to each item set in the N item sets according to the semantic features of the upper layer nodes and the semantic features of each item in the N item sets to obtain N lower layer nodes corresponding to the N item sets;
taking the N lower-layer nodes as the upper-layer nodes respectively;
repeating the steps from the step of extracting the semantic features of the upper-layer nodes and the semantic features of each article in the N article sets to the step of taking the N lower-layer nodes as the upper-layer nodes respectively until nodes of an Mth layer are determined, wherein M is the total number of layers of the decision tree;
and establishing the decision tree according to the nodes from the first layer to the Mth layer.
Optionally, the determining a root node of the decision tree according to the score data of the items in the total item set, and taking the root node as an upper node includes:
classifying the score of each item in the total item set according to the score of each item in the total item set;
calculating a score variance for each class of scores for each item in the total set of items;
obtaining the sum of the score variances of each article in the total article set according to the score variances of all the categories of each article in the total article set;
and acquiring the article corresponding to the minimum value of the sum of the scoring variances in the total article set as the root node of the decision tree.
Optionally, the determining, according to the semantic features of the upper node and the semantic features of each item in the N item sets, a lower node corresponding to each item set in the N item sets to obtain N lower nodes corresponding to the N item sets includes:
acquiring semantic features of the upper node and semantic features of each item in the N item sets;
determining semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets respectively;
obtaining a discrimination error value of each article in the N article sets according to the semantic correlation between the semantic features of the upper node and the semantic features of each article in the N article sets;
acquiring an article corresponding to the minimum value of the discrimination error values in each article set according to the discrimination error value of each article in the N article sets;
and determining the article corresponding to the minimum value of the discrimination error values in each article set as a lower-layer node corresponding to each article set.
Optionally, the obtaining a discrimination error value of each article in the N article sets according to the semantic relevance between the semantic feature of the upper node and the semantic feature of each article in the N article sets includes: determining a discrimination error value for each item in the N item sets according to the score variance of each item in the N item sets by using a discrimination error value formula, the discrimination error value formula comprising:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of a first item set, j represents the upper node, the first item set is any one of the N item sets, Errt(i) A discrimination error value, e, representing the ith item of the first set of items2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the top node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness, W, of semantic features representing the upper node to semantic features of an ith item of the first set of itemsiA normalized weight value representing a semantic relatedness.
Optionally, the normalized weight value of the semantic relevance between the semantic features of the upper node and the semantic features of the ith item of the first item set is determined according to all the semantic features of the ith item of the first item set.
Optionally, the method further includes:
receiving an answer to the current guide question;
and generating recommendation information according to the answer of the current guide question.
Optionally, the generating recommendation information according to the answer to the current guidance question further includes:
when the answer of the current guide question is a positive answer, selecting an article similar to the node corresponding to the current guide question to generate the recommendation information; or
And when the answer of the current guide question is a negative answer or uncertain, acquiring the recommendation information by adopting a collaborative filtering algorithm.
Optionally, the method further includes:
and determining an interest map of the user according to the answer of the current guide question and the answers of one or more guide questions answered before the current guide question, wherein the interest map is used for adjusting the item sequence in the recommendation information.
According to a second aspect of embodiments of the present disclosure, there is provided an issue determination apparatus for item recommendation, the apparatus comprising:
the article determination module is used for determining an article indicated by the answer of the previous guide question by utilizing a pre-established decision tree according to the answer of the previous guide question, wherein the decision tree is established according to the scoring data of the articles in the total article set and the relation between semantic features;
and the guide question acquisition module is used for acquiring the current guide question according to the article indicated by the answer of the previous guide question.
Optionally, the apparatus for establishing a decision tree includes:
the score acquisition module is used for acquiring score data of the items in the total item set, wherein the score data comprises a score of each item in the total item set;
a root node determining module, configured to determine a root node of the decision tree according to the score data of the items in the total item set, and use the root node as an upper node;
a semantic feature extraction module, configured to extract semantic features of the upper node and semantic features of each item in N item sets, where the N item sets are determined according to score data of each item in the total item set, and each item set includes one or more items;
a lower layer node determining module, configured to determine, according to the semantic features of the upper layer node and the semantic features of each item in the N item sets, a lower layer node corresponding to each item set in the N item sets, to obtain N lower layer nodes corresponding to the N item sets;
the iteration module is used for respectively taking the N lower-layer nodes as the upper-layer nodes;
a loop module, configured to repeat the step of extracting the semantic features of the upper node and the semantic features of each item in the N item sets to the step of using the N lower nodes as the upper nodes, respectively, until determining a node on an M-th layer, where M is the total number of layers of the decision tree;
and the decision tree establishing module is used for establishing the decision tree according to the nodes from the first layer to the Mth layer.
Optionally, the root node determining module includes:
the classification submodule is used for classifying the score of each article in the total article set according to the score of each article in the total article set;
a first calculation submodule for calculating a score variance of each type of score for each item in the total set of items;
the second calculation submodule is used for obtaining the sum of the score variances of each article in the total article set according to the score variances of all the categories of each article in the total article set;
and the root node determining submodule is used for acquiring the article corresponding to the minimum value of the sum of the scoring variances in the total article set and taking the article as the root node of the decision tree.
Optionally, the lower node determining module includes:
a semantic feature obtaining submodule, configured to obtain a semantic feature of the upper node and a semantic feature of each item in the N item sets;
a semantic relevance determining submodule for determining semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets respectively;
a discrimination error value determining submodule, configured to obtain a discrimination error value of each item in the N item sets according to semantic relevance between the semantic features of the upper node and the semantic features of each item in the N item sets;
the minimum value obtaining submodule is used for obtaining an article corresponding to the minimum value of the distinguishing degree error value in each article set according to the distinguishing degree error value of each article in the N article sets;
and the lower-layer node determining submodule is used for determining the article corresponding to the minimum value of the discrimination error value in each article set as the lower-layer node corresponding to each article set.
Optionally, the discrimination error value determining submodule includes: determining a discrimination error value for each item in the N item sets according to the score variance of each item in the N item sets by using a discrimination error value formula, the discrimination error value formula comprising:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of a first item set, j represents the upper node, the first item set is any one of the N item sets, Errt(i) A discrimination error value, e, representing the ith item of the first set of items2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the top node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness, W, of semantic features representing the upper node to semantic features of an ith item of the first set of itemsiA normalized weight value representing a semantic relatedness.
Optionally, the normalized weight value of the semantic relevance between the semantic features of the upper node and the semantic features of the ith item of the first item set is determined according to all the semantic features of the ith item of the first item set.
Optionally, the apparatus further comprises:
a first receiving module, configured to receive an answer to the current guidance question;
and the first recommending module is used for generating recommending information according to the answer of the current guide question.
Optionally, the first recommending module is further configured to:
when the answer of the current guide question is a positive answer, selecting an article similar to the node corresponding to the current guide question to generate the recommendation information; or
And when the answer of the current guide question is a negative answer or uncertain, acquiring the recommendation information by adopting a collaborative filtering algorithm.
Optionally, the apparatus further includes an optimization module, where the optimization module:
the method comprises the steps of determining an interest map of a user according to the answer of the current guide question and the answers of one or more guide questions answered before the current guide question, wherein the interest map is used for adjusting the item sequence in the recommendation information.
The technical scheme provided by the embodiment of the disclosure can have the following beneficial effects:
determining an article indicated by an answer to a previous guide question by using a pre-established decision tree according to the answer to the previous guide question; and acquiring the current guide question according to the article indicated by the answer of the previous guide question. Therefore, the problem determination method for item recommendation provided by the disclosure can solve the problem that when the guidance problem is generated, the relevance between the upper-level problem and the lower-level problem is low, so that the problem is relatively high in jumping performance, the relevance between the guidance problems can be improved, and the user experience of the answering user can be improved.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure without limiting the disclosure. In the drawings:
FIG. 1 is a flow diagram illustrating a problem determination method for item recommendations according to an exemplary embodiment;
FIG. 2 is a flow diagram illustrating a method of building a decision tree for determining a guidance question in accordance with an exemplary embodiment;
FIG. 3 is a flow chart illustrating a method of lower level node determination according to the embodiment shown in FIG. 2;
FIG. 4 is a flow diagram illustrating another problem determination methodology for item recommendation in accordance with an exemplary embodiment;
FIG. 5 is a block diagram illustrating an issue determination device for item recommendations in accordance with an exemplary embodiment;
FIG. 6 is a block diagram of an apparatus for building a decision tree according to the embodiment shown in FIG. 5;
FIG. 7 is a block diagram illustrating a root node determination module according to the embodiment shown in FIG. 6;
FIG. 8 is a block diagram of an underlying node determination module shown in accordance with the embodiment shown in FIG. 6;
FIG. 9 is a block diagram illustrating yet another problem determination apparatus for item recommendation in accordance with an exemplary embodiment;
FIG. 10 is a block diagram illustrating yet another problem determination apparatus for item recommendation in accordance with an exemplary embodiment.
Detailed Description
The following detailed description of specific embodiments of the present disclosure is provided in connection with the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present disclosure, are given by way of illustration and explanation only, not limitation.
Fig. 1 is a flowchart illustrating a problem determination method for item recommendation according to an exemplary embodiment, and as shown in fig. 1, the problem determination method for item recommendation may include the following steps:
in step 110, the item indicated by the answer to the previous guiding question is determined by using a pre-established decision tree according to the answer to the previous guiding question, wherein the decision tree is established according to the scoring data of the items in the total set of items and the relation between semantic features.
The establishment of the decision tree is the basis for obtaining the next guidance problem and recommending the article, and the cold start problem possibly existing in the problem guidance-based recommendation process is solved through the establishment of the decision tree, so that the technical defect that the recommendation of a new user cannot be realized due to the absence of data basis is overcome. The following embodiments are used to describe the manner of establishing the decision tree of the present disclosure, and the present disclosure utilizes the correlation of semantic features between items to determine the item in the decision tree as each node, so a decision tree with semantic correlation between upper and lower nodes is established, and therefore, there is no large leap between the upper and lower nodes in the decision tree, and the correlation between guide questions can be improved, so as to improve the user experience of the answering user.
In addition, the article in the embodiments of the present disclosure may be understood as a name, or an information set including a name and attribute information corresponding to the name, where the name is used to refer to its corresponding actual article, the actual article may be a physical article, such as various products, or a virtual article, such as a movie, a song, or the like, or may be a human or an animal, or the name may be used to refer to a class of articles in reality.
In step 120, the current guide question is obtained according to the item indicated by the answer to the previous guide question.
For example, the decision tree described in step 110 is a basis for determining the guiding question, and after obtaining the answer to the previous guiding question, the layer to which the answer points and the corresponding node in the layer may be determined through the decision tree according to the answer, so that the item as the node is the item indicated by the answer to the previous guiding question, and thus the current guiding question can be determined according to the item (for example, the guiding question corresponding to the item may be determined according to the preset relationship between the guiding question and the item).
For example, suppose the last guidance question is "do you like a youth movie? And if the received answer to the previous guidance question is a favorite, it is determined that the node corresponding to the current layer of the decision tree is the movie "hour era" according to the answer, and the obtained current guidance question may be a guidance question expanded based on the movie "hour era", for example: "do you like" the small era "this movie? "or" do you feel "the small times" how nice to look? "etc., the specific question may be preset based on the movie. Therefore, based on the decision tree, because the upper and lower nodes have semantic relevance, the relevance between the current guide question and the previous guide question is obtained, and similarly, the relevance between the current guide question and the next guide question is also obtained, so that the experience of answering users can be improved, and the interest of participating in the answer of the guide question is brought to the users.
In summary, the problem determination method for item recommendation provided by the embodiment of the present disclosure determines an item indicated by an answer to a previous guidance problem by using a pre-established decision tree according to the answer to the previous guidance problem; and acquiring the current guide question according to the article indicated by the answer of the previous guide question. Therefore, the problem determination method for item recommendation provided by the disclosure can solve the problem that when the guidance problem is generated, the relevance between the upper-level problem and the lower-level problem is low, so that the problem is relatively high in jumping performance, the relevance between the guidance problems can be improved, and the user experience of the answering user can be improved.
Fig. 2 is a flow chart illustrating a method for establishing a decision tree for determining a guidance question according to an exemplary embodiment, and as shown in fig. 2, the method for establishing a decision tree described in step 110 may include the following steps:
in step 111, scoring data of the items in the total set of items is obtained, where the scoring data includes a score of each item in the total set of items.
In step 112, a root node of the decision tree is determined according to the scoring data of the items in the total item set, and the root node is used as an upper node.
For example, the total item set may be understood as a set of item samples for building a decision tree, and the selection of items in the total item set may be based on some popular items or common items of a user, or other types of items, and may be specifically selected according to actual needs. The scoring data of the items in the total set of items includes a score of each item in the total set of items, and the score of each item in the total set of items may be pre-scored by the user.
Determining a root node of the decision tree from the scoring data for the items in the total set of items may include: first, the score of each item in the total item set is classified according to the score of each item in the total item set, and the classification may be a fuzzy large-class classification, for example, taking an item as a movie, the score of a user may be classified into three main classes: "like", "dislike", "unknown", the classification can be according to the following rules: the score is 7-10 points (full score is 10 points) is like, the score is 3-6 points is dislike, and the rest is unknown; the variance of the score for each class of score for each item in the total set of items is then calculated, and the mean is calculated for all scores under each class's class, and the variance of the scores under that class is calculated, e.g., using e2(tl) denotes the variance of the score under the category "like" with e2(th) score variance under the category "dislike", denoted by e2(tu) indicates that the score variance under the classification is 'unknown', then the score variance of each article in the total article set is obtained according to the score variance of each classification of each article in the total article set, and finally the article corresponding to the minimum value of the score variance sums in the total article set is obtained and used as the root node of the decision tree. For example, after calculating the score variance of the three categories of "like", "dislike", and "unknown" scores for a movie, the score variances of the three categories are added, i.e., Errt(i)＝e2(tl)+e2(th)+e2(tu), the electricity can be obtainedSum of the variance scores of the shadows Errt(i) Thus comparing the sum of the scores for all items, the Errt(i) The ith item corresponding to the minimum value of (1) is used as a root node of the decision tree to be established, after the root node of the decision tree is obtained, the root node can be used as an upper node, and a lower node is expanded to complete the complete establishment of the decision tree, namely step 113 is performed.
In step 113, semantic features of the upper node and semantic features of each item in N item sets are extracted, the N item sets being determined according to the score of each item in the total item set, each item set including one or more items.
For example, after the upper node is determined, the lower node to be expanded by the decision tree is no longer single, that is, besides the root node, there may be a plurality of leaf nodes of each subsequent layer, and therefore, when the lower node corresponding to the upper node is screened, there may also be a plurality of item sets for determining the lower node. Illustratively, a plurality of item sets may be determined from the scoring data described in step 101. For example, a plurality of item sets with smaller score variance values, for example, N item sets with smaller score variance values, may be obtained by setting a threshold value using the score variance of each category of each item obtained in step 112, and then step 114 may be performed on the N item sets to screen out the items in each item set as the lower-level nodes.
As mentioned above, the article in the present disclosure may be understood as a name, or an information set including a name and attribute information corresponding to the name, where the name is used to represent the actual article referred to by the name, the actual article may be a physical article, such as various products, or a virtual article, such as a movie, a song, etc., or may be a human or an animal, or the name may represent a kind of article in the actual article referred to by the name. Therefore, after the N item sets are determined, semantic features of each item in each set can be extracted according to attribute information corresponding to the name of the item; or, if the article is a name and does not include corresponding attribute information, the corresponding attribute information may be searched according to the name (which may be searched from a pre-established database or may be searched through the internet), so that the semantic features of the article may be extracted.
In step 114, a lower level node corresponding to each item set of the N item sets is determined according to the semantic features of the upper level node and the semantic features of each item of the N item sets, so as to obtain N lower level nodes corresponding to the N item sets.
In step 115, the N lower nodes are respectively used as upper nodes.
Illustratively, a lower level node corresponding to one item set is determined by the method using semantic features in the previous step, and since N item sets of the lower level node determined in step 113 are total, the method in step 114 needs to select the corresponding lower level node for each item set in the N item sets in turn, so that one item that can be used as a lower level node is selected from each item set in the N item sets, and the corresponding N lower level nodes are obtained, so as to complete the node construction of the decision tree at the current level. Then, using the current layer as a new upper layer in an iterative manner, that is, using N lower-layer nodes of the current layer as upper-layer nodes of the next layer, and performing steps 113 to 115 again, and so on, determining leaf nodes in each subsequent layer of the decision tree by using the above method, so as to continue to perform subsequent leaf node expansion of the decision tree until determining nodes of the mth layer, where M is the total number of layers of the decision tree, thereby obtaining all nodes required by the decision tree.
In step 116, a decision tree is built based on the nodes of the first through M-th levels.
In summary, starting from the root node determined in step 112, the following steps 113 to 115 perform layer-by-layer expansion on other layer nodes of the decision tree based on the speech features of the article on the basis of the root node, until all the nodes required by the decision tree are obtained, so that each layer node of the decision tree and the upper and lower layer nodes thereof are semantically associated with each other, and therefore, the guiding problem determined by using the decision tree has an association with the previous guiding problem or the next guiding problem, thereby being capable of improving the answering experience of the answering user.
Fig. 3 is a flowchart of a method for determining a lower node according to the embodiment shown in fig. 2, and as shown in fig. 3, the determining a lower node corresponding to each item set of N item sets according to the semantic features of the upper node and the semantic features of each item in the N item sets in step 114, and obtaining N lower nodes corresponding to the N item sets may include the following steps:
in step 1141, semantic features of upper nodes and semantic features of each item in the N item sets are obtained.
For example, taking an article as a movie as an example, semantic features of the movie may mainly include four types: the director, actors, genre and name, for example, take the article as the "small era" of the movie, and the semantic features for the "small era" of the movie are extracted as follows: the director is Guojining, actors have old winter, Yang power and the like, the types belong to youth, the name is hour generation, and semantic feature extraction is carried out on each item in the N item sets in such a way for the following judgment.
In step 1142, semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets is determined.
Illustratively, according to the semantic features of the upper node determined in step 1141 and the semantic features of all the articles in the N article sets, the semantic relevance between the semantic features is obtained, and illustratively, the semantic feature may be defined as TF (i, j) for determining the semantic relevance, and the semantic relevance is defined as dis [ TF (i, j) ]. Then, the score of each item in each item set may be combined, and the determination in step 1143 is performed to obtain a differentiation degree error value of each item in each item set.
In step 1143, a differentiation degree error value of each item in the N item sets is obtained according to the semantic correlation between the semantic features of the upper node and the semantic features of each item in the N item sets.
Illustratively, a discrimination error value for each item in the N item sets may be determined from the score variance of each item in the N item sets by utilizing the following discrimination error value formula, which includes:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of the first item set, j represents the upper node, the first item set is any one of N item sets, Errt(i) A discrimination error value, e, for the ith item representing the first set of items2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the upper node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness of the semantic features of the upper node to the semantic features of the ith item of the first item set, WiA normalized weight value representing a semantic relatedness.
In the process of confirming the leaf nodes, the method is different from the method for determining the root nodes, and the semantic relevance between the upper and lower nodes is considered at this time, so that taking the first item set as an example, the scores of the items in the first item set are classified in sequence, the variance calculation is performed on the score under each classification, and the sum of the variance values of each item in the first item set in all the categories is obtained. The discrimination error value of the ith item of the first item set is the product of the sum of the scores and variances of the ith item of the first item set minus the semantic relevance of the semantic features of the upper node and the semantic features of the ith item of the first item set and the normalized weight value of the semantic relevance. That is to say, under the condition that the normalized weight value is not changed, the higher the semantic correlation degree between semantic features of upper and lower nodes is, the larger the product thereof is, and thus the differentiation degree error value after subtraction is smaller; otherwise, the discrimination error value is larger, which indicates that the semantic relevance of the upper layer and the lower layer is lower.
In addition, the normalized weight value of the semantic relevance between the semantic features of the upper node in the above formula and the semantic features of the ith item of the first item set may be determined according to all the semantic features of the ith item of the first item set. For example, the category of the article selects a movie, and selects the name, type, director and director data of the movie as semantic features for extraction, the weights of the above four semantic features in the semantic relevancy may be set to be the same, and the normalized weight value of the semantic relevancy may also be adjusted according to the actual situation. For example, the weight of the director can be adjusted to 4 times of the original weight, and the weights of the actors can be adjusted to 2 times of the original weight, on the basis of the adjustment, a normalized weight value of the semantic relevancy more fitting the actual requirement is obtained, and then the formula is calculated, so that the value of the semantic relevancy more conforms to the actual situation of the user, and the discrimination error value of the semantic relevancy has a reference value and actual pertinence. Therefore, the occupied weight of different recommended articles can be flexibly adjusted according to the characteristics of the categories to which the articles belong, so that the possibility that the recommended articles meet the interest of users is improved, for example, in the case of guidance of catering categories, the taste and the price can be used as semantic features with higher weight, and the environment and the position are lower in weight, so that the guidance which is more suitable for the actual situation is realized.
In step 1144, an article corresponding to the minimum value of the differentiation degree error values in each article set is obtained according to the differentiation degree error value of each article in the N article sets.
In step 1145, the article corresponding to the minimum value of the differentiation error values in each article set is determined as the lower node corresponding to each article set. That is, the item corresponding to the minimum value of the differentiation error value in each item set may be determined as the item of the lower node in the item set as the upper node.
Fig. 4 is a flowchart illustrating another problem determination method for item recommendation according to an exemplary embodiment, as shown in fig. 4, the method shown in fig. 1 further includes the following steps:
in step 130, an answer to the current guide question is received.
Illustratively, the guidance question determined in step 120 is "how do you like the movie of the" small era? After issuing such a guidance question to the user, waiting for the user to answer the question after obtaining the answer of the user, the determination of step 140 may be performed according to the answer.
In step 140, recommendation information is generated based on the answer to the current guidance question.
Exemplarily, in the guidance question "do you like the movie of the small era? ", the user has three available answers to the guide question: "like", "dislike" and "unknown", and then the item pointed to by the different answers in the decision tree is determined according to the decision tree established before, so as to recommend the item, for example, for the user answering "like", the recommended movie "youth", the recommended user answering "dislike" is the movie "chinese partnership", and the remaining users who do not answer preferences directly, the recommendable movie "fly by bullet". It can be seen from the above method that for the users who like the small era of movies, the movies of youth series should meet their preferences, so that the movies are recommended to be youth-oriented, but do not like the descriptions of the users of the small era, the interested categories may belong to the more mature movie contents, and the chinese partners are suitable items for them. In addition, it can be understood that the recommendation of the item and the acquisition of the next guide question are both based on the answer of the current guide question, that is, when a certain item is determined according to the answer of the user, the item can be recommended to the user, and the next guide question generated based on the item can also be provided. For example, a user who answers "dislike" recommends the movie "Chinese partnership", and generates the next guidance question: "do you like China partners this movie? ".
For example, after obtaining the answer to the current guidance question, generating recommendation information according to the answer to the current guidance question may further include:
when the answer to the current guidance question is an affirmative answer, item generation recommendation information similar to the node to which the current guidance question corresponds (i.e., the item on which the current guidance question is generated) is selected. When the answer of the user to the current guide question is a positive answer, the fact that the user likes or is interested is indicated, and therefore the item most similar to the item can be found out to be recommended according to the semantic features of the item.
Or when the answer of the current guide question is a negative answer or uncertain, acquiring the recommendation information by adopting a collaborative filtering algorithm. Different from the previous positive answer, when the answer of the user to the answer of the guide question is negative or the preference of the user cannot be clearly and directly identified, a collaborative filtering algorithm can be adopted to find out and recommend the articles selected by other users having the same preference with the user, and the collaborative filtering algorithm is a method for recommending articles with similar preferences by selecting the articles with similar preferences through the similarity among different users.
The generating of the recommendation information according to the answer to the current guide question may be understood as generating the recommendation information according to the answer to the guide question each time the answer to one guide question is obtained, or may be generated according to the answer to the current guide question and the answers to one or more guide questions answered before the current guide question. That is to say, the problem determination method for item recommendation does not generate recommendation information after the user has to answer all designed guidance problems, but can generate corresponding recommendation information after any guidance problem, so that the problem that the user easily gives up answering questions when the user needs to answer many guidance problems before obtaining the recommendation information is avoided, each question has corresponding recommendation information, the recommendation range and content can be greatly expanded, and the experience satisfaction of the user is improved.
Further, after the user answers a plurality of questions, an interest map of the user can be determined according to the answer of the current guide question and the answers of one or more guide questions answered before the current guide question, and the interest map can be used for adjusting the item sequence in the recommendation information. The recommended articles can be more than two, so that when the recommended articles are provided for the user, the collected answers of a plurality of guide problems can be utilized to establish an interest map for a single user, and the article sequence of the recommended information is better optimized, so that the articles which best accord with the user preference are arranged in front of other recommended articles, the recommendation which is more fit with the user preference is realized, and the user satisfaction is increased.
In summary, the problem determination method for item recommendation provided by the embodiment of the present disclosure determines an item indicated by an answer to a previous guidance question by using a pre-established decision tree according to the answer to the previous guidance question, where the decision tree is established according to the scoring data of the items in the item total set and the relationship between semantic features; and acquiring the current guide question according to the article indicated by the answer of the previous guide question. Therefore, the problem determination method for item recommendation provided by the disclosure can solve the problem that when the guidance problem is generated, the relevance between the upper-level problem and the lower-level problem is low, so that the problem is relatively high in jumping performance, the relevance between the guidance problems can be improved, and the user experience of the answering user can be improved. And items meeting the user's preferences may also be recommended to the user based on the answers to the guide questions.
Fig. 5 is a block diagram illustrating an issue determination apparatus for item recommendation, according to an example embodiment, the apparatus 500 may be used to perform the method described in any of fig. 1-4. Referring to fig. 5, the apparatus 500 includes:
an item determination module 5010, configured to determine, according to the answer to the previous guide question, an item indicated by the answer to the previous guide question by using a pre-established decision tree, where the decision tree is established according to the scoring data of the items in the total set of items and the relationship between the semantic features;
the guide question acquiring module 5020 is used for acquiring the current guide question according to the item indicated by the answer of the previous guide question.
Optionally, fig. 6 is a block diagram illustrating another problem determination apparatus for item recommendation according to an exemplary embodiment, referring to fig. 6, the apparatus 500 further includes:
a score obtaining module 5030, configured to obtain score data of the items in the total item set, where the score data includes a score of each item in the total item set;
a root node determining module 5040, configured to determine a root node of the decision tree according to the score data of the items in the total item set, and use the root node as an upper node;
a semantic feature extraction module 5050, configured to extract semantic features of upper nodes and semantic features of each item in N item sets, where the N item sets are determined according to score data of each item in an item total set, and each item set includes one or more items;
a lower node determining module 5060, configured to determine, according to the semantic features of the upper node and the semantic features of each item in the N item sets, a lower node corresponding to each item set in the N item sets, to obtain N lower nodes corresponding to the N item sets;
a loop module 5070, configured to take the N lower nodes as upper nodes, respectively;
the loop module 5070 is further configured to repeat the steps of extracting semantic features of upper nodes and semantic features of each item in the N item sets to the step of using the N lower nodes as upper nodes, respectively, until determining a node at an M-th layer, where M is the total number of layers of the decision tree;
a decision tree building module 5080, configured to build a decision tree according to the nodes in the first layer to the mth layer.
Optionally, fig. 7 is a block diagram of a root node determining module according to the embodiment shown in fig. 6, and referring to fig. 7, the root node determining module 5040 includes:
the classification submodule 5041 is used for classifying the score of each item in the total item set according to the score of each item in the total item set;
a variance calculation sub-module 5042 for calculating the score variance for each class of scores for each item in the total set of items;
the variance sum calculation sub-module 5043 is used for obtaining the sum of the score variances of each article in the total article set according to the score variances of all the categories of each article in the total article set;
the root node determining submodule 5044 is configured to obtain an item corresponding to the minimum value of the sum of the scores and the variances in the item total set, and use the item as a root node of the decision tree.
Alternatively, fig. 8 is a block diagram of a lower node determining module according to the embodiment shown in fig. 6, and referring to fig. 8, the lower node determining module 5060 includes:
the semantic feature acquisition submodule 5061 is used for acquiring semantic features of upper nodes and semantic features of each item in the N item sets;
the semantic relevance determining submodule 5062 is configured to determine semantic relevance between the semantic features of the upper node and the semantic features of each item in the N item sets;
the discrimination error value determining submodule 5063 is configured to obtain a discrimination error value of each item in the N item sets according to semantic relevance between the semantic features of the upper node and the semantic features of each item in the N item sets;
the article determining submodule 5064 is configured to obtain, according to the discrimination error value of each article in the N article sets, an article corresponding to the minimum value of the discrimination error values in each article set;
the lower node determining submodule 5065 is configured to determine that the item corresponding to the minimum value of the discrimination error values in each item set is the lower node corresponding to each item set.
Optionally, the discrimination error value determination submodule 5063 in the apparatus of fig. 8 is configured to: determining a discrimination error value for each item in the N item sets according to the score variance of each item in the N item sets by using a discrimination error value formula, the discrimination error value formula comprising:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of the first item set, j represents the upper node, the first item set is any one of N item sets, Errt(i) A discrimination error value, e, for the ith item representing the first set of items2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the upper node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness of semantic features representing upper nodes to semantic features of the ith item of the first set of items, WiA normalized weight value representing a semantic relatedness.
Optionally, the normalized weight value of the semantic relevance between the semantic features of the upper node and the semantic features of the ith item of the first item set is determined according to all the semantic features of the ith item of the first item set.
Optionally, fig. 9 is a block diagram illustrating a further problem determination apparatus for item recommendation according to an exemplary embodiment, and referring to fig. 9, the apparatus 500 further includes:
an answer receiving module 5090, configured to receive an answer to the current guide question;
the recommending module 5100 is configured to generate recommendation information according to an answer of the current guidance question.
Optionally, the recommending module 5100 is further configured to:
when the answer of the current guide question is a positive answer, selecting an article similar to the node corresponding to the current guide question to generate recommendation information; or
And when the answer of the current guide question is a negative answer or uncertain, acquiring recommendation information by adopting a collaborative filtering algorithm.
Alternatively, fig. 10 is a block diagram of another problem determination apparatus for item recommendation according to an exemplary embodiment, referring to fig. 10, the apparatus further includes:
a determining module 5110, configured to determine an interest map of the user according to the answer of the current guide question and the answers of one or more guide questions previously answered to the current guide question, where the interest map is used to adjust the order of the items in the recommendation information.
In summary, the problem determination apparatus for item recommendation provided in the embodiments of the present disclosure determines an item indicated by an answer to a previous guide question by using a pre-established decision tree according to the answer to the previous guide question, where the decision tree is established according to the scoring data of the items in the total item set and the relationship between semantic features; and acquiring the current guide question according to the article indicated by the answer of the previous guide question. Therefore, the problem determination method for item recommendation provided by the disclosure can solve the problem that when the guidance problem is generated, the relevance between the upper-level problem and the lower-level problem is low, so that the problem is relatively high in jumping performance, the relevance between the guidance problems can be improved, and the user experience of the answering user can be improved. And items meeting the user's preferences may also be recommended to the user based on the answers to the guide questions.
The preferred embodiments of the present disclosure are described in detail with reference to the accompanying drawings, however, the present disclosure is not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solution of the present disclosure within the technical idea of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. In order to avoid unnecessary repetition, various possible combinations will not be separately described in this disclosure.
In addition, any combination of various embodiments of the present disclosure may be made, and the same should be considered as the disclosure of the present disclosure, as long as it does not depart from the spirit of the present disclosure.
Claims (16)
1. A problem determination method for item recommendation, the method comprising:
determining an article indicated by the answer of the previous guide question by utilizing a pre-established decision tree according to the answer of the previous guide question, wherein the decision tree is established according to the scoring data of the articles in the total article set and the relation between semantic features;
acquiring a current guide question according to the article indicated by the answer of the last guide question; the method for establishing the decision tree comprises the following steps:
obtaining scoring data of the items in the total item set, wherein the scoring data comprises a score of each item in the total item set;
determining a root node of the decision tree according to the scoring data of the items in the total item set, and taking the root node as an upper node;
extracting semantic features of the upper node and semantic features of each item in N item sets, wherein the N item sets are determined according to the score of each item in the total item set, and each item set comprises one or more items;
determining a lower layer node corresponding to each item set in the N item sets according to the semantic features of the upper layer nodes and the semantic features of each item in the N item sets to obtain N lower layer nodes corresponding to the N item sets;
taking the N lower-layer nodes as the upper-layer nodes respectively;
repeating the steps from the step of extracting the semantic features of the upper-layer nodes and the semantic features of each article in the N article sets to the step of taking the N lower-layer nodes as the upper-layer nodes respectively until nodes of an Mth layer are determined, wherein M is the total number of layers of the decision tree;
and establishing the decision tree according to the nodes from the first layer to the Mth layer.
2. The method according to claim 1, wherein the determining a root node of the decision tree according to the score data of the items in the total item set, and regarding the root node as an upper node comprises:
classifying the score of each item in the total item set according to the score of each item in the total item set;
calculating a score variance for each class of scores for each item in the total set of items;
obtaining the sum of the score variances of each article in the total article set according to the score variances of all the categories of each article in the total article set;
and acquiring the article corresponding to the minimum value of the sum of the scoring variances in the total article set as the root node of the decision tree.
3. The method according to claim 1, wherein the determining a lower level node corresponding to each of the N item sets according to the semantic features of the upper level node and the semantic features of each of the N item sets to obtain N lower level nodes corresponding to the N item sets comprises:
acquiring semantic features of the upper node and semantic features of each item in the N item sets;
determining semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets respectively;
obtaining a discrimination error value of each article in the N article sets according to the semantic correlation between the semantic features of the upper node and the semantic features of each article in the N article sets;
acquiring an article corresponding to the minimum value of the discrimination error values in each article set according to the discrimination error value of each article in the N article sets;
and determining the article corresponding to the minimum value of the discrimination error values in each article set as a lower-layer node corresponding to each article set.
4. The method according to claim 3, wherein the obtaining a discrimination error value for each item in the N item sets according to the semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets comprises: determining a discrimination error value for each item in the N item sets according to the score variance of each item in the N item sets by using a discrimination error value formula, the discrimination error value formula comprising:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of a first item set, j represents the upper node, the first item set is any one of the N item sets, Errt(i) A discrimination error value, e, representing the ith item of the first set of items2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the top node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness, W, of semantic features representing the upper node to semantic features of an ith item of the first set of itemsiA normalized weight value representing a semantic relatedness.
5. The method according to claim 4, wherein the normalized weight value of the semantic relevance of the semantic features of the upper node to the semantic features of the ith item of the first item set is determined according to all semantic features of the ith item of the first item set.
6. The method of claim 1, further comprising:
receiving an answer to the current guide question;
and generating recommendation information according to the answer of the current guide question.
7. The method of claim 6, wherein generating recommendation information based on the answer to the current guide question further comprises:
when the answer of the current guide question is a positive answer, selecting an article similar to the node corresponding to the current guide question to generate the recommendation information; or
And when the answer of the current guide question is a negative answer or uncertain, acquiring the recommendation information by adopting a collaborative filtering algorithm.
8. The method of claim 6, further comprising:
and determining an interest map of the user according to the answer of the current guide question and the answers of one or more guide questions answered before the current guide question, wherein the interest map is used for adjusting the item sequence in the recommendation information.
9. A problem determination apparatus for item recommendation, the apparatus comprising:
the article determination module is used for determining an article indicated by the answer of the previous guide question by utilizing a pre-established decision tree according to the answer of the previous guide question, wherein the decision tree is established according to the scoring data of the articles in the total article set and the relation between semantic features;
the guide question acquisition module is used for acquiring a current guide question according to the article indicated by the answer of the previous guide question; wherein the content of the first and second substances,
the device further comprises:
the score acquisition module is used for acquiring score data of the items in the total item set, wherein the score data comprises a score of each item in the total item set;
a root node determining module, configured to determine a root node of the decision tree according to the scores of the items in the total item set, and use the root node as an upper node;
a semantic feature extraction module, configured to extract semantic features of the upper node and semantic features of each item in N item sets, where the N item sets are determined according to score data of each item in the total item set, and each item set includes one or more items;
a lower layer node determining module, configured to determine, according to the semantic features of the upper layer node and the semantic features of each item in the N item sets, a lower layer node corresponding to each item set in the N item sets, to obtain N lower layer nodes corresponding to the N item sets;
a circulation module, configured to use the N lower layer nodes as the upper layer nodes, respectively;
the loop module is further configured to repeat the step of extracting the semantic features of the upper node and the semantic features of each item in the N item sets to the step of using the N lower nodes as the upper nodes, respectively, until determining a node on an M-th layer, where M is the total number of layers of the decision tree;
and the decision tree establishing module is used for establishing the decision tree according to the nodes from the first layer to the Mth layer.
10. The apparatus of claim 9, wherein the root node determining module comprises:
the classification submodule is used for classifying the score of each article in the total article set according to the score of each article in the total article set;
a variance calculation sub-module for calculating a score variance for each class of scores for each item in the total set of items;
the variance sum calculation submodule is used for obtaining the sum of the score variances of each article in the total article set according to the score variances of all the categories of each article in the total article set;
and the root node determining submodule is used for acquiring the article corresponding to the minimum value of the sum of the scoring variances in the total article set and taking the article as the root node of the decision tree.
11. The apparatus of claim 9, wherein the lower node determining module comprises:
a semantic feature obtaining submodule, configured to obtain a semantic feature of the upper node and a semantic feature of each item in the N item sets;
a semantic relevance determining submodule for determining semantic relevance of the semantic features of the upper node to the semantic features of each item in the N item sets respectively;
a discrimination error value determining submodule, configured to obtain a discrimination error value of each item in the N item sets according to semantic relevance between the semantic features of the upper node and the semantic features of each item in the N item sets;
the article determining submodule is used for acquiring an article corresponding to the minimum value of the distinguishing degree error value in each article set according to the distinguishing degree error value of each article in the N article sets;
and the lower-layer node determining submodule is used for determining the article corresponding to the minimum value of the discrimination error value in each article set as the lower-layer node corresponding to each article set.
12. The apparatus of claim 11, wherein the discrimination error value determination submodule is configured to: determining a discrimination error value for each item in the N item sets according to the score variance of each item in the N item sets by using a discrimination error value formula, the discrimination error value formula comprising:
Errt(i)＝e2(t)–Wi*dis[TF(i,j)]
wherein i represents the ith item of a first item set, j represents the upper node, the first item set is any one of the N item sets, Errt(i) Representing the first articleDifference error value of ith item of set, e2(t) represents the score variance of the ith item of the first set of items, TF (i, j) represents the semantic features of the top node and the semantic features of the ith item of the first set of items, dis [ TF (i, j)]Semantic relatedness, W, of semantic features representing the upper node to semantic features of an ith item of the first set of itemsiA normalized weight value representing a semantic relatedness.
13. The apparatus of claim 12, wherein the normalized weight value of the semantic relevance of the semantic features of the upper node to the semantic features of the ith item of the first item set is determined from all semantic features of the ith item of the first item set.
14. The apparatus of claim 9, further comprising:
the answer receiving module is used for receiving the answer of the current guide question;
and the recommending module is used for generating recommending information according to the answer of the current guide question.
15. The apparatus of claim 14, wherein the recommendation module is further configured to:
when the answer of the current guide question is a positive answer, selecting an article similar to the node corresponding to the current guide question to generate the recommendation information; or
And when the answer of the current guide question is a negative answer or uncertain, acquiring the recommendation information by adopting a collaborative filtering algorithm.
16. The apparatus of claim 14, further comprising:
and the determining module is used for determining an interest map of the user according to the answer of the current guide question and the answers of one or more guide questions answered before the current guide question, wherein the interest map is used for adjusting the order of the items in the recommendation information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611000826.3A CN106776701B (en) | 2016-11-14 | 2016-11-14 | Problem determination method and device for item recommendation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611000826.3A CN106776701B (en) | 2016-11-14 | 2016-11-14 | Problem determination method and device for item recommendation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106776701A CN106776701A (en) | 2017-05-31 |
| CN106776701B true CN106776701B (en) | 2020-04-07 |
Family
ID=58973597
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611000826.3A Active CN106776701B (en) | 2016-11-14 | 2016-11-14 | Problem determination method and device for item recommendation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106776701B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107357849B (en) * | 2017-06-27 | 2020-11-03 | 北京百度网讯科技有限公司 | Interaction method and device based on test application |
| CN107992610A (en) * | 2017-12-15 | 2018-05-04 | 泰康保险集团股份有限公司 | A kind of automatic question-answering method, device, equipment and storage medium |
| CN109166631A (en) * | 2018-07-10 | 2019-01-08 | 武汉海云健康科技股份有限公司 | The construction method of map is associated with the drug of convolutional neural networks based on Network Science |
| CN110265153A (en) * | 2019-05-16 | 2019-09-20 | 平安科技（深圳）有限公司 | Chronic disease follow-up method and electronic device |
| CN110930969B (en) * | 2019-10-14 | 2024-02-13 | 科大讯飞股份有限公司 | Background music determining method and related equipment |
| CN111353784A (en) * | 2020-02-25 | 2020-06-30 | 支付宝(杭州)信息技术有限公司 | Transfer processing method, system, device and equipment |
| CN112288531A (en) * | 2020-10-30 | 2021-01-29 | 广州富港万嘉智能科技有限公司 | Recommended dish generation method, computer-readable storage medium, intelligent cooking equipment and server |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572982A (en) * | 2014-12-31 | 2015-04-29 | 东软集团股份有限公司 | Personalized recommendation method and system based on question guide |
| CN104794242A (en) * | 2015-05-11 | 2015-07-22 | 何杨洲 | Searching methods |
-
2016
- 2016-11-14 CN CN201611000826.3A patent/CN106776701B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572982A (en) * | 2014-12-31 | 2015-04-29 | 东软集团股份有限公司 | Personalized recommendation method and system based on question guide |
| CN104794242A (en) * | 2015-05-11 | 2015-07-22 | 何杨洲 | Searching methods |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106776701A (en) | 2017-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106776701B (en) | Problem determination method and device for item recommendation | |
| US11514333B2 (en) | Combining machine-learning and social data to generate personalized recommendations | |
| Cufoglu | User profiling-a short review | |
| WO2017041370A1 (en) | Human-computer chatting method and device based on artificial intelligence | |
| CN106709040B (en) | Application search method and server | |
| CN104081429B (en) | Video recommendations based on video co-occurrence statistics | |
| US9754306B2 (en) | Recommendation engine with profile analysis | |
| Selke et al. | Pushing the boundaries of crowd-enabled databases with query-driven schema expansion | |
| CN105339933B (en) | Pass through the news results of query expansion | |
| EP2560111A2 (en) | Systems and methods for facilitating the gathering of open source intelligence | |
| CN107924401A (en) | Video recommendations based on video title | |
| CN110321291A (en) | Test cases intelligent extraction system and method | |
| KR20160057475A (en) | System and method for actively obtaining social data | |
| JP4370850B2 (en) | Information processing apparatus and method, program, and recording medium | |
| CN107767279A (en) | A kind of average weighted personalized friend recommendation method based on LDA | |
| KR101543780B1 (en) | System and method for expert search by dynamic profile and social network reliability | |
| CN101551806A (en) | Personalized website navigation method and system | |
| CN109241451B (en) | Content combination recommendation method and device and readable storage medium | |
| CN109902823B (en) | Model training method and device based on generation countermeasure network | |
| CN102999507A (en) | Recommendation processing method and device for information of network microblog celebrities | |
| US20150248720A1 (en) | Recommendation engine | |
| WO2014107194A1 (en) | Identifying relevant user content | |
| Nguyen et al. | NowAndThen: a social network-based photo recommendation tool supporting reminiscence | |
| US20160094887A1 (en) | Tv program-based shopping guide system and tv program-based shopping guide method | |
| KR20090046353A (en) | Method for personalized multi recommendation learning |
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 |