WO2016179938A1 - Method and device for question recommendation - Google Patents

Abstract

A method and device for question recommendation. The question recommendation method comprises: receiving a search question (S11); acquiring question attribute information of the search question, and acquiring an initial search result according to the question attribute information (S12); acquiring user description information of a user, and sorting the initial search result according to the user description information to obtain a sorted result (S13); and selecting a preset number of questions from the sorted result, and determining the same as recommended questions (S14). The question recommendation method is able to determine the relevance between the recommended questions and the search questions, thus improving the recommendation accuracy.

Description

Title recommendation method and topic recommendation device

Cross-reference to related applications

This application claims the priority of the Chinese patent application number "201510246727.2", which was submitted by Baidu Online Network Technology (Beijing) Co., Ltd. on May 14, 2015, and whose invention title is "Title Recommendation Method and Title Recommendation Device".

Technical field

The present invention relates to the field of Internet technologies, and in particular, to a topic recommendation method and a topic recommendation device.

Background technique

There are three main methods for recommending related topics: one is to find other similar descriptions based on the description text of the current topic, and present a list to the user, and the list is sorted according to text similarity; the second is According to the knowledge point information of the current topic, try to find other topics with the same or similar knowledge points; the third is to find two or more questions that the user frequently views in order according to the user's retrieval and click behavior, and to mine the frequent items of the behavior. The relationship between the topics is established, and the user is presented with a list, and the list is sorted according to the strength of the association.

However, the relevance of the questions recommended by the above methods is poor, and the recommended effect is not satisfactory.

Summary of the invention

The present invention aims to solve at least one of the technical problems in the related art to some extent.

To this end, an object of the present invention is to provide a method for recommending a topic, which can improve the correlation between a recommended topic and a search topic, and improve the recommendation effect.

Another object of the present invention is to provide a title recommendation device.

The method for recommending a topic according to the first aspect of the present invention includes: receiving a search topic; acquiring topic attribute information of the search title, and obtaining a preliminary search result according to the topic attribute information; acquiring a user of the user Descriptive information, and sorting the preliminary search results according to the user description information to obtain a sorted result; and selecting a preset number of results from the sorted result to determine a recommended title.

The topic recommendation method proposed by the first aspect of the present invention obtains the preliminary search result by acquiring the topic attribute information and the reference attribute information, and not only the text similarity, but also the recommended title and the search title. Correlation; in addition, by obtaining the user description information and sorting the preliminary search results according to the user description information, the user information can be referred to in the recommendation, and the correlation with the user is improved. Improve the recommendation.

In order to achieve the above object, a topic recommendation device according to a second aspect of the present invention includes: a receiving module, configured to receive a search topic; and an obtaining module, configured to acquire topic attribute information of the search topic, and according to the topic attribute Obtaining a preliminary search result; the sorting module is configured to obtain user description information of the user, and sort the preliminary search result according to the user description information to obtain a sorted result; and a determining module, configured to After the result, the result of selecting the preset number is determined as the recommended title.

The title recommendation device according to the second aspect of the present invention obtains the preliminary attribute search information according to the title attribute information, and obtains the preliminary search result according to the title attribute information. Since the title attribute information is referenced, not only the text similarity, but also the recommended title and the search title can be improved. Correlation; in addition, by obtaining the user description information and sorting the preliminary search results according to the user description information, the user information can be referred to in the recommendation, the relevance to the user is improved, and the recommendation effect is improved.

An embodiment of the present invention further provides an electronic device, including: one or more processors; a memory; one or more programs, the one or more programs being stored in the memory when the one or more When the processor is executed: the method according to any of the first aspect of the invention is performed.

Embodiments of the present invention also provide a non-volatile computer storage medium having one or more modules stored when the one or more modules are executed: performing the first aspect of the present invention The method of any of the preceding claims. The additional aspects and advantages of the invention will be set forth in part in the description which follows.

DRAWINGS

The above and/or additional aspects and advantages of the present invention will become apparent and readily understood from

1 is a schematic flow chart of a method for recommending a topic according to an embodiment of the present invention;

2 is a schematic flowchart of acquiring topic attribute information in an embodiment of the present invention;

3 is a schematic diagram of an online implementation process in an embodiment of the present invention;

4 is a schematic flowchart of obtaining preliminary search results in an embodiment of the present invention;

FIG. 5 is a schematic flowchart of obtaining user description information in an embodiment of the present invention;

6 is a schematic diagram of an offline implementation process in an embodiment of the present invention;

7 is a schematic flowchart of establishing a topic structured information base in an embodiment of the present invention;

8 is a schematic flowchart of establishing a user model in an embodiment of the present invention;

FIG. 9 is a schematic structural diagram of a title recommendation apparatus according to another embodiment of the present invention; FIG.

FIG. 10 is a schematic structural diagram of a title recommendation apparatus according to another embodiment of the present invention.

detailed description

The embodiments of the present invention are described in detail below, and the examples of the embodiments are illustrated in the accompanying drawings, in which the same or similar reference numerals indicate the same or similar modules or modules having the same or similar functions. The embodiments described below with reference to the accompanying drawings are intended to be illustrative of the invention and are not to be construed as limiting. Rather, the invention is to cover all modifications, variations and equivalents of the inventions.

FIG. 1 is a schematic flowchart of a method for recommending a topic according to an embodiment of the present invention, where the method includes:

S11: Receive a search question.

When the user needs to retrieve a question, a search question can be entered in the search box.

It can be understood that the embodiment can be executed by the server. When executed by the server, after the search box in the browser receives the search question input by the user, the search question can be sent to the server, and the server receives the message sent by the browser. Search for the title. or,

The embodiment may also be implemented by a web product having a search function, the web product includes a part of the front end interacting with the user and a background processing part. At this time, the search term input by the user may be received by the front end of the web product, for example, by searching. The box receives the search question entered by the user.

S12: Obtain topic attribute information of the search topic, and obtain a preliminary search result according to the topic attribute information.

Optionally, referring to FIG. 2, the acquiring the topic attribute information of the search topic includes:

S21: Obtain identification information of the search topic.

For example, when the search term input by the user is in the form of a picture, the image may be identified by Optical Character Recognition (OCR) to obtain the recognition result. In the pre-stored original question bank, the search result is the same as or similar to the recognition result. The title identifies the identification information (id) of the same or similar topic as the identification information (id) of the current search title.

S22: Acquire, in a pre-established topic structured information database, topic attribute information corresponding to the identifier information, where the title structured information database corresponds to the identification information of the saved topic and the topic attribute information.

After obtaining the id of the search title, the title attribute information corresponding to the id of the search title may be acquired in the pre-established topic structured information base.

The title structured information database stores the attribute attribute information corresponding to the identification information of the title. The title attribute information includes, for example, the title type, the title difficulty, the title structure, the topic knowledge point, the answer quality, and the normalized topic description.

Specifically, obtaining the recommended topic according to the search term input by the user is completed online. Referring to FIG. 3, the online system may include a topic feature acquisition module 31. After the topic feature acquisition module 31 obtains the retrieval topic, the topic attribute information may be obtained from the topic structured information library.

Referring to FIG. 3, the online system may further include: a text retrieval module 32, and the topic feature acquisition module 31 acquires the topic. After the attribute information, the text retrieval module 32 can obtain the preliminary search result according to the topic attribute information.

Optionally, referring to FIG. 4, the obtaining the preliminary search result according to the attribute information of the topic includes:

S41: Obtain a keyword of the search topic, and perform a text search according to the keyword to obtain a text search result;

For example, a general word segmentation technique can be used to segment a search term, and then a keyword can be obtained from the obtained segmentation according to a preset rule. The preset rules are, for example, based on the position of the participle in the search title, the degree of importance in the question bank, whether the subject word, and the like.

After obtaining the keyword, the keyword can be used as a query to retrieve the relevant topic in the existing database, and the text retrieval result related to the text description is obtained.

S42: Perform weight adjustment on the text retrieval result according to the attribute information of the topic, and obtain a search result after the weight adjustment;

For example, the text retrieval result can be adjusted according to the knowledge point, type, difficulty, and answer quality of the search topic. Specifically, the text search result may be limited to the same or similar to the knowledge point of the search title, and the text search result whose type and difficulty are similar to the search title is weighted, and the weight of the text search result with high quality is weighted. The specific weighted values can be preset according to actual needs.

By adjusting the weight, text retrieval results with different weights can be obtained.

In addition, the knowledge points used in this embodiment may be fine-grained knowledge points. For details, refer to related descriptions in the subsequent topic knowledge point extraction.

By using fine-grained knowledge points, the relevance of recommended topics to current search questions can be improved.

S43: Acquire the knowledge point information in the attribute information of the topic, and perform the weight adjustment on the weighted search result, where the knowledge point information includes: a single knowledge point or a mixed knowledge point;

For example, if the search title is a single knowledge point, the text search result of the single knowledge point is weighted; or, if the search title is a mixed knowledge point, each knowledge point in the mixed knowledge point and the corresponding weight are determined, and then according to Each knowledge point and the corresponding weight determine the weight of the text retrieval result, and the text retrieval result close to the retrieval title is weighted.

In this embodiment, by distinguishing between a single knowledge point and a mixed knowledge point, the correlation between the recommended topic and the current search topic can be improved.

S44: Select a preset number of search results from the search result after the weight adjustment again, and determine the preliminary search result.

For example, according to the weight information, a preset number of search results having a larger weight is selected as the preliminary search result, and the preset number is, for example, 50.

S13: Acquire user description information of the user, and sort the preliminary search results according to the user description information to obtain a sorted result;

Optionally, referring to FIG. 5, the acquiring user description information of the user includes:

S51: Obtain identification information of the user.

The user can be assigned a corresponding identification information when the user is registered. When the user logs in, the login information carries the user's own identification information, and the system can obtain the user identification information (id) in the login information.

S52: The user description information corresponding to the identifier information of the user is obtained in the preset user model, where the user model correspondingly saves the identifier information of the user and the user description information.

Referring to FIG. 3, the online system may further include: a user feature obtaining module 33, and the user feature obtaining module 33 may obtain user description information corresponding to the user's identification information in the user model according to the user's identification information.

The user description information includes, for example, user preference difficulty, user preference type, user textbook version, user browsing of the topic, click, and collection status.

After the user description information is obtained, the preliminary search results may be sorted. For example, referring to FIG. 3, the online system further includes: an advanced sorting module 34, and the advanced sorting module 34 is configured to sort the preliminary search results according to the attribute information of the user. .

Specifically, the advanced sorting process may include: weighting preliminary search results that are consistent with the user description information, and sorting the preliminary search results according to the weighted weights.

For example, weight the preliminary search results as follows:

If the difficulty level of the topic is consistent with the user's preference difficulty level, weighting is performed;

If the topic type is consistent with the user preference type, weighting is performed;

The source of the question is consistent with the version of the user's textbook and weighted.

The title grade information is consistent with the user's current grade and weighted.

The current knowledge point user history behavior analysis is adjusted according to the difficulty and number of browsing of the current knowledge point in the user history, for example, a question slightly higher than the current difficulty level, and weighted.

After the above weighting is performed, the search results are sorted in descending order of weights, for example, 50 sorted results are obtained.

S14: Select a preset number of results from the sorted result to determine a recommended title.

The preset number may be set by the user or set by the system by default. After that, the result obtained by sorting in the previous step may be selected according to the order from the front to the back.

Further, when the main body of the embodiment is a web product, the web product can also display the recommended topic to the user.

For example, referring to Figure 3, the advanced ranking can show the user the recommendation results associated with the search topic.

The above topic structured information base and user model can be established offline.

When building the topic structured information base, referring to FIG. 6, the offline system may include a feature extraction distribution module 61 and a title. The title feature extraction module 62, the topic feature extraction module 62 may specifically include: a topic type classification module 621, a topic difficulty classification module 622, a topic structure splitting module 623, a topic knowledge point extraction module 624, an answer quality ranking module 625, and a topic description A module 626.

Referring to FIG. 7, the establishing a topic structured information base includes:

S71: Obtain a history topic.

Among them, the historical title refers to the title that is input when the topic structured information database is created. Since it is input before the current search topic, it can be called a history topic in order to distinguish it from the current search topic.

Referring to Figure 6, each new historical topic is represented by a new topic.

S72: Distribute the historical title to different classification modules.

For example, referring to FIG. 6, the feature extraction distribution module distributes each newly added topic to a classification module such as a topic type classification module and a topic difficulty classification module.

S73: Perform topic feature extraction on the historical topic, and obtain topic attribute information corresponding to each classification module.

Specifically, the process of extracting topic features may include:

(1) Classification of topic types

After the topic type classification, the title attribute information that can be obtained includes: the subject type and the type of the topic.

For example, a classification model built using a Support Vector Machine (SVM). The n-gram-based feature is used to classify the topic type, including the subject type (language, mathematics, physics, chemistry, etc.) of the topic, as well as the type of the topic (choose, fill in the blank, short answer, etc.).

(2) Classification of the difficulty of the topic

After the problem difficulty is graded, the obtained topic attribute information may include: the difficulty level to which each newly added topic belongs, and the number of difficulty levels may be preset.

Specifically, a classification model constructed by a Gradient Boosting Decision Tree (GBDT) may be used to divide the difficulty value of the problem and generate a difficulty level classification. In the classification module, the main features adopted include: the description keyword of the parsed content, the length of the answer, the ratio of the number of times the knowledge point is questioned to the number of resolutions, and the problem is obtained on the User Generated Content (UGC) platform. The number of answers, the level of respondents, and the time-consuming information.

(3) Topic structure split

After the topic structure is split, the obtained topic attribute information may include: a structure such as a stem, a question, an option, and a blank fill.

Specifically, it can be classified according to the sentence granularity, and then divided according to the sequence of the sentence, and then divided into various structures such as the stem (background description fragment, conditional fragment), questions, options, and blanks.

(4) Topic knowledge point extraction

Through the topic knowledge point extraction, the obtained topic attribute information includes: knowledge points, and the knowledge points can be represented by tags.

The knowledge point label can be obtained by fusing the following two main processes.

(I) Subject keyword extraction: In the case of topic structure splitting, keyword extraction is mainly performed from the conditional segment and the question segment. Keyword extraction is done using an SVM classification model. The features used include the part of speech, the position in the sentence, the importance in the question bank, and whether the subject is a feature. Classify words as keywords and non-keywords. In each question, keywords that reach the threshold can be selected, and the maximum value of the selected keywords can be limited, for example, up to 5 keywords are selected.

After extracting the keyword, the tag can be directly obtained from the keyword, for example, the keyword is determined as a tag or the like.

(II) Similar segment alignment: using similar segment alignment methods, the important parts of the current topic (conditional segments, question segments, and option segments) are compared with other important segments of the title tag that have been tagged in the question bank. . The K-Nearest Neighbor (KNN) method is used to mark the current topic to be processed.

(III) Label fusion: merging the first label and the second label, for example, de-duplicating the first label and the second label to form a set, wherein the first label is a label extracted by a keyword, and the second label is A label obtained by comparing similar fragments. In addition, when determining the label, the confidence level corresponding to each label may be determined, and finally, according to the confidence level, a preset number of labels with a larger degree of confidence is selected as the knowledge point label of the topic, and the preset number is, for example, ten.

(5) Answer quality rating

After answering the quality rating, the obtained topic attribute information includes the level of the answer quality, for example, high quality answer, general quality answer, low quality answer, and the like.

Specifically, the SVM model can be used to rank the quality of the responses, which are divided into high quality, general quality, and low quality answers. The features used in answering the quality grading are: the semantic relevance of the answer and the topic, the source, the length, the formatting information, and the user's click-through behavior.

(6) Title description

The title attribute information obtained by the title description includes, for example, a normalized description.

Specifically, the descriptions of different topics (especially formulas) are inconsistent. According to the definition, after the description mode is normalized, it is added to the database.

It can be understood that the above classification model uses SVM, GBDT as an example, and can also be the remaining classification models, for example, Logistic regression, linear regression, random forest, neural network, naive Bayes and other algorithms with classification ability. model.

S74: Acquire identification information of the historical topic, and correspondingly save the identification information of the topic and the topic attribute information.

Each of the historical topics may be assigned unique identification information. After obtaining the topic attribute information of the historical topic, the identification information and the topic attribute information of the historical title may be correspondingly saved in the topic structured information database.

The prior art acquires a recommended topic according to the description text of the current topic, due to the degree of structural analysis of the topic Not enough, there is no effective way to distinguish between the type of topic, the condition, the semantic scene of the content of the question, etc., resulting in low relevance of the search. In this embodiment, when different topic attribute information is saved, different topic attribute information such as the title type, the title difficulty, the title structure, the topic knowledge point, etc. are saved in a structured information form, so that the attribute information of each topic has a hierarchy. Relationships and interconnections to improve search relevance.

When the user model is established, referring to FIG. 6, the offline system further includes: a user modeling module 63, configured to establish a user model according to the user behavior log, the user attribute information, and the topic structured information library.

Optionally, referring to FIG. 8, the establishing a user model includes:

S81: Obtain a user behavior log and obtain user attribute information.

Among them, the user behavior log can record the user's browsing, clicking, and collecting questions in the question bank.

User attribute information refers to some meta attributes about the user, such as gender, type (parent/student/teacher), region, grade, school, etc.

S82: Perform user modeling according to the topic structured information base, and the user behavior log and user attribute information, to obtain a user model.

For example, through the user behavior log, the user can browse, click or collect the title, and according to the structured information database of the topic, the topic attribute information of the corresponding topic can be obtained, so that the user preference information can be obtained, for example, the subject concerned by the user can be known. Knowledge point information, topic information, knowledge point difficulty level, title type information, etc. And, after obtaining the user attribute information, it can also be recorded in the user model.

In addition, the user identification information may be allocated when the user registers, and the user identification information and the user preference information and the user attribute information may be saved in the user model.

In this embodiment, by obtaining the topic attribute information and obtaining the preliminary search result according to the topic attribute information, since the title attribute information is referred to, not only the text similarity, the correlation between the recommended topic and the search topic can be improved; The user describes the information, and sorts the preliminary search results according to the user description information, and can refer to the user information in the recommendation, thereby improving the correlation with the user and improving the recommendation effect. In this embodiment, the fine-grained processing is performed on the knowledge point extraction, and the difference between the single knowledge point and the mixed knowledge point is considered, and the correlation between the recommended related topic and the currently retrieved topic is effectively improved. This embodiment introduces more topic attribute information, such as difficulty, type, etc., which cannot be processed from the perspective of text similarity. The introduction of these attribute information improves the correlation between the recommended topic and the search topic. In this embodiment, by referring to the user attribute information, the user's personalized recommendation can be supported, and different users can be seen to see different related topics, thereby improving the user experience.

FIG. 9 is a schematic structural diagram of a title recommendation apparatus according to another embodiment of the present invention. The apparatus 90 includes a receiving module 91, an obtaining module 92, a sorting module 93, and a determining module 94.

The receiving module 91 is configured to receive a search question;

When the user needs to retrieve a question, a search question can be entered in the search box.

It can be understood that the embodiment may be a server. The receiving module at this time is configured to receive a search question sent by the browser, and the browser may obtain the search question input by the user from the search box. or,

The embodiment may also be a web product device with a search function, and the receiving module at this time is configured to receive a search question input by the user.

The obtaining module 92 is configured to acquire the topic attribute information of the search topic, and obtain a preliminary search result according to the topic attribute information;

Optionally, the obtaining module 92 is configured to obtain the topic attribute information of the search topic, including:

Obtaining identification information of the search topic;

In the pre-established topic structured information database, the title attribute information corresponding to the identifier information is obtained, wherein the title structured information database corresponds to the identification information of the saved topic and the topic attribute information.

For example, when the search term input by the user is in the form of a picture, the image may be identified by Optical Character Recognition (OCR) to obtain the recognition result. In the pre-stored original question bank, the search result is the same as or similar to the recognition result. The title identifies the identification information (id) of the same or similar topic as the identification information (id) of the current search title.

After obtaining the id of the search title, the title attribute information corresponding to the id of the search title may be acquired in the pre-established topic structured information base.

The title structured information database stores the attribute attribute information corresponding to the identification information of the title. The title attribute information includes, for example, the title type, the title difficulty, the title structure, the topic knowledge point, the answer quality, and the normalized topic description.

Optionally, referring to FIG. 10, the apparatus 90 further includes: a first establishing module 95, configured to establish a topic structured information base, where the first establishing module 95 is specifically configured to:

Obtain historical topics;

Distributing the historical title to different classification modules;

Performing topic feature extraction on the historical topic, and acquiring topic attribute information corresponding to each classification module;

Obtaining the identification information of the historical topic, and correspondingly storing the identification information of the topic and the topic attribute information.

For the process of establishing the topic structured information base, refer to the related description in the method embodiment, and details are not described herein again.

Optionally, the first establishing module 95 is configured to perform topic feature extraction on the historical topic, and obtain topic attribute information corresponding to each classification module, including:

When the classification module is a topic knowledge point extraction module, the topic keyword extraction is performed, and the similar segment comparison is performed, and the first label is extracted according to the topic keyword, and the second label is obtained according to the similar segment comparison, and The first label and the second label are merged, and a preset number of labels are selected from the merged labels to be determined as a knowledge point label.

Optionally, the obtaining module 92 is configured to obtain a preliminary search result according to the topic attribute information, including:

Obtaining a keyword of the search topic, and performing a text search according to the keyword to obtain a text search result;

And adjusting the weight of the text search result according to the attribute information of the topic to obtain a search result after the weight adjustment;

Obtaining the knowledge point information in the attribute information of the topic, and re-adjusting the weighted search result, where the knowledge point information includes: a single knowledge point or a mixed knowledge point;

From the search result after the weight adjustment again, a preset number of search results are selected and determined as the preliminary search result.

For example, a general word segmentation technique can be used to segment a search term, and then a keyword can be obtained from the obtained segmentation according to a preset rule. The preset rules are, for example, based on the position of the participle in the search title, the degree of importance in the question bank, whether the subject word, and the like.

After obtaining the keyword, the keyword can be used as a query to retrieve the relevant topic in the existing database, and the text retrieval result related to the text description is obtained.

For example, the text retrieval result can be adjusted according to the knowledge point, type, difficulty, and answer quality of the search topic. Specifically, the text search result may be limited to the same or similar to the knowledge point of the search title, and the text search result whose type and difficulty are similar to the search title is weighted, and the weight of the text search result with high quality is weighted. The specific weighted values can be preset according to actual needs.

By adjusting the weight, text retrieval results with different weights can be obtained.

In addition, the knowledge points used in this embodiment may be fine-grained knowledge points. For details, refer to related descriptions in the subsequent topic knowledge point extraction.

By using fine-grained knowledge points, the relevance of recommended topics to current search questions can be improved.

For example, if the search title is a single knowledge point, the text search result of the single knowledge point is weighted; or, if the search title is a mixed knowledge point, each knowledge point in the mixed knowledge point and the corresponding weight are determined, and then according to Each knowledge point and the corresponding weight determine the weight of the text retrieval result, and the text retrieval result close to the retrieval title is weighted.

In this embodiment, by distinguishing between a single knowledge point and a mixed knowledge point, the correlation between the recommended topic and the current search topic can be improved.

For example, according to the weight information, a preset number of search results having a larger weight is selected as the preliminary search result, and the preset number is, for example, 50.

The sorting module 93 is configured to obtain user description information of the user, and sort the preliminary search results according to the user description information to obtain a sorted result;

Optionally, the sorting module 93 is configured to obtain user description information of the user, including:

Obtain identification information of the user;

Obtaining, in a preset user model, user description information corresponding to the identifier information of the user, where The user model correspondingly stores the identification information of the user and the user description information.

The user can be assigned a corresponding identification information when the user is registered. When the user logs in, the login information carries the user's own identification information, and the system can obtain the user identification information (id) in the login information.

The user description information includes, for example, user preference difficulty, user preference type, user textbook version, user browsing of the topic, click, and collection status.

Optionally, referring to FIG. 10, the apparatus 90 further includes: a second establishing module 96, configured to establish a user model, where the second establishing module 96 is specifically configured to:

Obtain a user behavior log and obtain user attribute information;

According to the topic structured information base, and the user behavior log and user attribute information, user modeling is performed to obtain a user model.

For the process of establishing the user model, refer to the related description in the method embodiment, and details are not described herein again.

Optionally, the sorting module 93 is configured to sort the preliminary search results according to the user description information, including:

The preliminary search results that are consistent with the user description information are weighted, and the preliminary search results are sorted according to the weighted weights.

For example, weight the preliminary search results as follows:

If the difficulty level of the topic is consistent with the user's preference difficulty level, weighting is performed;

If the topic type is consistent with the user preference type, weighting is performed;

The source of the question is consistent with the version of the user's textbook and weighted.

The title grade information is consistent with the user's current grade and weighted.

The current knowledge point user history behavior analysis is adjusted according to the difficulty and number of browsing of the current knowledge point in the user history, for example, a question slightly higher than the current difficulty level, and weighted.

After the above weighting is performed, the search results are sorted in descending order of weights, for example, 50 sorted results are obtained.

The determining module 94 is configured to select a preset number of results from the sorted result and determine the recommended title.

The preset number may be set by the user or set by the system by default. After that, the result obtained by sorting in the previous step may be selected according to the order from the front to the back.

Further, when the device is a web product device, referring to FIG. 10, the search topic is input by the user, and the device includes:

The display module 97 is configured to display the recommended topic to the user.

In this embodiment, by obtaining the title attribute information and obtaining the preliminary search result according to the title attribute information, since the title attribute information is referred to, not only the text similarity, but the recommended title and the search title can be improved. Correlation; In addition, by obtaining the user description information and sorting the preliminary search results according to the user description information, the user information may be referred to during the recommendation, thereby improving the relevance with the user and improving the recommendation effect. In this embodiment, the fine-grained processing is performed on the knowledge point extraction, and the difference between the single knowledge point and the mixed knowledge point is considered, and the correlation between the recommended related topic and the currently retrieved topic is effectively improved. This embodiment introduces more topic attribute information, such as difficulty, type, etc., which cannot be processed from the perspective of text similarity. The introduction of these attribute information improves the correlation between the recommended topic and the search topic. In this embodiment, by referring to the user attribute information, the user's personalized recommendation can be supported, and different users can be seen to see different related topics, thereby improving the user experience.

An embodiment of the present invention further provides an electronic device, including: one or more processors; a memory; one or more programs, the one or more programs being stored in the memory when the one or more When the processor executes: receiving the search topic; acquiring the topic attribute information of the search topic, and obtaining a preliminary search result according to the topic attribute information; acquiring user description information of the user, and performing the preliminary according to the user description information The search results are sorted to obtain the sorted result; after the sorted result, the preset number of results is selected and determined as the recommended title.

Embodiments of the present invention also provide a non-volatile computer storage medium storing one or more modules, when the one or more modules are executed: receiving a retrieval title; acquiring the retrieval a topic attribute information of the topic, and obtaining a preliminary search result according to the topic attribute information; acquiring user description information of the user, and sorting the preliminary search result according to the user description information to obtain a sorted result; After the sorted result, the result of the preset number is selected and determined as the recommended title. .

It should be noted that in the description of the present invention, the terms "first", "second" and the like are used for descriptive purposes only, and are not to be construed as indicating or implying relative importance. Further, in the description of the present invention, the meaning of "a plurality" is two or more unless otherwise specified.

Any process or method description in the flowcharts or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a particular logical function or process. And the scope of the preferred embodiments of the invention includes additional implementations, in which the functions may be performed in a substantially simultaneous manner or in an opposite order depending on the functions involved, in the order shown or discussed. It will be understood by those skilled in the art to which the embodiments of the present invention pertain.

It should be understood that portions of the invention may be implemented in hardware, software, firmware or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques well known in the art: having logic gates for implementing logic functions on data signals. Discrete logic circuits, application specific integrated circuits with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

One of ordinary skill in the art can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by a program to instruct related hardware, and the program can be stored in a computer readable storage medium. When executed, one or a combination of the steps of the method embodiments is included.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. The integrated modules, if implemented in the form of software functional modules and sold or used as stand-alone products, may also be stored in a computer readable storage medium.

The above mentioned storage medium may be a read only memory, a magnetic disk or an optical disk or the like.

In the description of the present specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like means a specific feature described in connection with the embodiment or example. A structure, material or feature is included in at least one embodiment or example of the invention. In the present specification, the schematic representation of the above terms does not necessarily mean the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples.

Although the embodiments of the present invention have been shown and described, it is understood that the above-described embodiments are illustrative and are not to be construed as limiting the scope of the invention. The embodiments are subject to variations, modifications, substitutions and variations.

Claims (21)

A method for recommending a topic, characterized in that it comprises: Receiving a search question; Obtaining topic attribute information of the search topic, and obtaining preliminary search results according to the topic attribute information; Obtaining user description information of the user, and sorting the preliminary search results according to the user description information to obtain a sorted result; A result of selecting a preset number from the sorted result is determined as a recommended title. The method according to claim 1, wherein the obtaining the topic attribute information of the search topic comprises: Obtaining identification information of the search topic; In the pre-established topic structured information database, the title attribute information corresponding to the identifier information is obtained, wherein the title structured information database corresponds to the identification information of the saved topic and the topic attribute information. The method of claim 2, further comprising: establishing a topic structured information base, the establishing the topic structured information base, comprising: Obtain historical topics; Distributing the historical title to different classification modules; Performing topic feature extraction on the historical topic, and acquiring topic attribute information corresponding to each classification module; Obtaining the identification information of the historical topic, and correspondingly storing the identification information of the topic and the topic attribute information. The method of claim 3 wherein said classification module comprises at least one of the following: The topic type classification module, the topic difficulty classification module, the topic structure split module, the topic knowledge point extraction module, the answer quality classification module, and the topic description normalization module. The method of claim 4, wherein the topic feature extraction is performed on the historical topic, and the topic attribute information corresponding to each classification module is obtained, including: When the classification module is a topic knowledge point extraction module, the topic keyword extraction is performed, and the similar segment comparison is performed, and the first label is extracted according to the topic keyword, and the second label is obtained according to the similar segment comparison, and The first label and the second label are merged, and a preset number of labels are selected from the merged labels to be determined as a knowledge point label. The method according to any one of claims 1 to 5, wherein the obtaining the preliminary search result according to the title attribute information comprises: Obtaining a keyword of the search topic, and performing a text search according to the keyword to obtain a text search result; And adjusting the weight of the text search result according to the attribute information of the topic to obtain a search result after the weight adjustment; Obtaining the knowledge point information in the attribute information of the topic, and re-adjusting the weighted search result, where the knowledge point information includes: a single knowledge point or a mixed knowledge point; From the search result after the weight adjustment again, a preset number of search results are selected and determined as the preliminary search result. The method according to any one of claims 1-6, wherein the obtaining user description information of the user comprises: Obtain identification information of the user; The user description information corresponding to the identifier information of the user is obtained in a preset user model, where the user model correspondingly stores the identifier information of the user and the user description information. The method of claim 7, further comprising: establishing a user model, the establishing a user model, comprising: Obtain a user behavior log and obtain user attribute information; According to the topic structured information base, and the user behavior log and user attribute information, user modeling is performed to obtain a user model. The method according to any one of claims 1-8, wherein the sorting the preliminary search results according to the user description information comprises: The preliminary search results that are consistent with the user description information are weighted, and the preliminary search results are sorted according to the weighted weights. The method according to any one of claims 1 to 9, wherein the search term is input by the user, the method further comprising: Presenting the recommended topic to the user. A title recommendation device, comprising: a receiving module, configured to receive a search question; An obtaining module, configured to acquire topic attribute information of the search topic, and obtain a preliminary search result according to the topic attribute information; a sorting module, configured to acquire user description information of the user, and sort the preliminary search results according to the user description information, to obtain a sorted result; The determining module is configured to select a preset number of results from the sorted result and determine the recommended title. The device according to claim 11, wherein the obtaining module is configured to acquire the title attribute information of the search topic, including: Obtaining identification information of the search topic; In the pre-established topic structured information database, the title attribute information corresponding to the identifier information is obtained, wherein the title structured information database corresponds to the identification information of the saved topic and the topic attribute information. The device according to claim 12, further comprising: a first establishing module, configured to establish a topic structured information base, wherein the first establishing module is specifically configured to: Obtain historical topics; Distributing the historical title to different classification modules; Performing topic feature extraction on the historical topic, and acquiring topic attribute information corresponding to each classification module; Obtaining the identification information of the historical topic, and correspondingly storing the identification information of the topic and the topic attribute information. The device according to claim 13, wherein the first establishing module is configured to perform topic feature extraction on the historical topic, and obtain topic attribute information corresponding to each classification module, including: When the classification module is a topic knowledge point extraction module, the topic keyword extraction is performed, and the similar segment comparison is performed, and the first label is extracted according to the topic keyword, and the second label is obtained according to the similar segment comparison, and The first label and the second label are merged, and a preset number of labels are selected from the merged labels to be determined as a knowledge point label. The device according to any one of claims 11 to 14, wherein the obtaining module is configured to obtain preliminary search results according to the topic attribute information, including: Obtaining a keyword of the search topic, and performing a text search according to the keyword to obtain a text search result; And adjusting the weight of the text search result according to the attribute information of the topic to obtain a search result after the weight adjustment; Obtaining the knowledge point information in the attribute information of the topic, and re-adjusting the weighted search result, where the knowledge point information includes: a single knowledge point or a mixed knowledge point; From the search result after the weight adjustment again, a preset number of search results are selected and determined as the preliminary search result. The device according to any one of claims 11 to 15, wherein the sorting module is configured to acquire user description information of the user, including: Obtain identification information of the user; The user description information corresponding to the identifier information of the user is obtained in a preset user model, where the user model correspondingly stores the identifier information of the user and the user description information. The device according to claim 16, further comprising: a second establishing module, configured to establish a user model, wherein the second establishing module is specifically configured to: Obtain a user behavior log and obtain user attribute information; According to the topic structured information base, and the user behavior log and user attribute information, user modeling is performed to obtain a user model. The device according to any one of claims 11-17, wherein the sorting module is configured to sort the preliminary search results according to the user description information, including: Pivoting the preliminary search results consistent with the user description information, and based on the weighted weights, the preliminary search The results are sorted. The device according to any one of claims 11 to 18, wherein the search term is input by the user, the device further comprising: And a display module, configured to display the recommended topic to the user. An electronic device, comprising: One or more processors; Memory One or more programs, the one or more programs being stored in the memory, when executed by the one or more processors: Performing the method of any of claims 1-10. A non-volatile computer storage medium characterized in that the computer storage medium stores one or more modules when the one or more modules are executed: Performing the method of any of claims 1-10.

Images