CN104361063B - user interest discovery method and device - Google Patents

Abstract

Embodiments of the present invention provide a method and device for discovering user interests, the method comprising: acquiring user-input expression data or behavior data for recommended content; determining interest prediction results according to the expression data or behavior data; The client prompts the interest prediction result, and obtains the user's click data on the interest prediction result to determine the user's interest. The embodiments of the present invention can allow users to actively express interests through natural language, discover user interests in a timely and accurate manner, and improve user experience.

Description

User Interest Discovery Method and Device

technical field

The embodiments of the present invention relate to the field of information technology, and in particular, to a method and device for discovering user interests.

Background technique

Personalized information recommendation technology can deliver information in line with the user's interests to users, so this technology is gradually being used more and more in network access. In the personalized information recommendation technology, it is necessary to accurately and timely discover user interests.

Existing user interest discovery technology generally obtains positive feedback behaviors such as clicking, sharing, and favorites and/or negative feedback behaviors such as ignoring and disapproving of recommended content, and analyzes user interests from the feedback behaviors.

The above-mentioned user interest discovery technology has the following defects: On the one hand, the information reflected by the user’s feedback behavior is relatively vague. Given a recommended content, whether it is a positive feedback behavior or a negative feedback behavior, it is difficult to accurately determine the specific attributes of the user’s feedback. As a result, the accuracy of discovered user interests is low; on the other hand, it often requires a large amount of user feedback behaviors to accurately capture user interests. Since this process usually takes a long time, it is not conducive to timely discovery of user interests.

Contents of the invention

Embodiments of the present invention provide a method and device for discovering user interests, so as to improve the accuracy and timeliness of discovered user interests.

Embodiments of the present invention adopt the following technical solutions:

On the one hand, an embodiment of the present invention provides a method for discovering user interests, including:

Obtain the expression data or behavior data of the recommended content input by the user;

determining an interest prediction result according to the expression data or the behavior data;

The application client is prompted with the interest prediction result, and the click data of the user on the interest prediction result is acquired to determine the user interest.

Further, according to the expression data, determine interest prediction results, including:

Determining an object of interest and a user attitude corresponding to the expression data according to the expression data;

According to the object of interest and user attitude corresponding to the expression data, an interest prediction result is determined.

Further, before obtaining the expression data of the recommended content input by the user, it also includes:

Obtain the behavioral data of the recommended content input by the user;

Determine the user's satisfaction with the recommended content according to the behavior data;

If the satisfaction degree is less than the set threshold value, the operation of acquiring the expression data of the recommended content input by the user is triggered.

Further, according to the expression data, determining the object of interest corresponding to the expression data includes:

matching the words contained in the expression data in the object knowledge base;

The words that are successfully matched are used as the object of interest corresponding to the expression data.

Further, after using the successfully matched words as the object of interest corresponding to the expression data, it also includes:

According to the text distance between the object of interest and the word corresponding to the user attitude, the object of interest whose text distance is greater than a set value is filtered out.

Further, according to the expression data, determining the user attitude corresponding to the expression data includes:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

Further, before determining the interest prediction result according to the behavior data, it also includes:

Determine the user's satisfaction with the recommended content according to the behavior data;

If the satisfaction degree is less than the set threshold value, an operation of determining an interest prediction result according to the behavior data is triggered.

Further, according to the behavior data, determining the user's satisfaction with the recommended content includes at least one of the following:

Determine the user's satisfaction with the recommended content according to the user's refresh frequency of the recommended content;

Determine the user's satisfaction with the recommended content based on the user's click data and stay time on the recommended content;

According to the user's support feedback data and attention time on the recommended content, the user's satisfaction with the recommended content is determined.

Further, after prompting the application client for the interest prediction result, and obtaining the click data of the user on the interest prediction result to determine the user interest, the method further includes:

According to the determined user interests, the recommended content pushed to the user is revised.

On the other hand, an embodiment of the present invention also provides a user interest discovery device, including:

The user data acquisition module is used to acquire the expression data or behavior data of the recommended content input by the user;

An interest prediction result determination module, configured to determine an interest prediction result according to the expression data or the behavior data;

The interest determination module is configured to prompt the application client for the interest prediction result, and obtain user click data on the interest prediction result to determine the user interest.

Further, the interest prediction result determination module includes:

An interest object determining unit, configured to determine an interest object corresponding to the expression data according to the expression data;

a user attitude determining unit, configured to determine a user attitude corresponding to the expression data according to the expression data;

The interest prediction result determining unit is configured to determine the interest prediction result according to the object of interest and user attitude corresponding to the expression data.

Further, the device also includes:

The first satisfaction degree determining module is used to determine the user's satisfaction degree to the recommended content according to the behavior data before obtaining the expression data of the recommended content input by the user;

The expression data acquisition triggering module is configured to trigger the execution of the operation of acquiring the expression data of the recommended content input by the user when the satisfaction degree is less than the set threshold value.

Further, the interest object determination unit includes:

The matching subunit is used to match the words contained in the expression data in the object knowledge base;

The interested object determination subunit is configured to use the successfully matched word as the interested object corresponding to the expression data.

Further, the interest object determination unit also includes:

The interest object filtering subunit is used to filter out the text distance greater than the set value according to the text distance between the interest object and the word corresponding to the user attitude after the successfully matched word is used as the interest object corresponding to the expression data. Fixed-value object of interest.

Further, the user attitude determining unit is specifically used for:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

Further, the device also includes:

The second satisfaction determination module is used to determine the user's satisfaction with the recommended content according to the behavior data before determining the interest prediction result according to the behavior data;

The interest prediction result determining trigger module is configured to trigger determining the interest prediction result according to the behavior data if the satisfaction degree is less than a set threshold value.

Further, the first satisfaction determination module or the second satisfaction determination module includes at least one of the following:

A first satisfaction degree determining unit, configured to determine the user's satisfaction degree to the recommended content according to the user's refresh frequency on the recommended content;

The second satisfaction determination unit is used to determine the user's satisfaction with the recommended content according to the user's click data on the recommended content and the length of stay;

The third satisfaction degree determining unit is configured to determine the user's degree of satisfaction with the recommended content according to the user's support feedback data and attention time on the recommended content.

Further, the device also includes:

The push module is configured to prompt the application client for the interest prediction result, and obtain the click data of the user on the interest prediction result to determine the user interest, and modify the recommended content pushed to the user according to the determined user interest.

The beneficial technical effect of the technical solution proposed by the embodiment of the present invention is: by obtaining the expression data or behavior data input by the user for the recommended content, the interest prediction result is determined, and then the application client is prompted for the interest prediction result, and the user's interest prediction is obtained. Result click data to determine user interests. The embodiments of the present invention can allow users to actively express interests through natural language, discover user interests in a timely and accurate manner, and improve user experience.

Description of drawings

In order to illustrate the present invention more clearly, the accompanying drawings that need to be used in the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art , on the premise of not paying creative labor, other drawings can also be obtained based on these drawings.

FIG. 1 is a flowchart of a method for discovering user interests provided by Embodiment 1 of the present invention;

FIG. 2 is a flow chart of a method for discovering user interests provided in Embodiment 2 of the present invention;

FIG. 3 is a flow chart of a method for discovering user interests provided by Embodiment 3 of the present invention;

FIG. 4 is a flowchart of a method for discovering user interests provided by Embodiment 4 of the present invention;

FIG. 5 is a flowchart of a method for discovering user interests provided by Embodiment 5 of the present invention;

FIG. 6 is a flowchart of a method for discovering user interests provided by Embodiment 5 of the present invention;

FIG. 7 is a flow chart of a method for discovering user interests provided in Embodiment 5 of the present invention;

FIG. 8 is a structural block diagram of an apparatus for discovering user interests provided by Embodiment 6 of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Example. It can be understood that the specific embodiments described here are only used to explain the present invention, rather than limit the present invention. Based on the embodiments of the present invention, all those skilled in the art can obtain without creative work. Other embodiments all belong to the protection scope of the present invention. In addition, it should be noted that, for the convenience of description, only parts related to the present invention are shown in the drawings but not all content.

Embodiment one

Fig. 1 is a flow chart of a method for discovering user interests provided by this embodiment, the method comprising the following steps:

101. Obtain the expression data of the recommended content input by the user.

102. According to the expression data, determine an object of interest corresponding to the expression data, including:

matching the words contained in the expression data in the object knowledge base;

The words that are successfully matched are used as the object of interest corresponding to the expression data.

103. According to the expression data, determine the user attitude corresponding to the expression data, including:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

104. Determine an interest prediction result according to the object of interest and user attitude corresponding to the expression data.

105. Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

In this embodiment, the expression data may specifically be interest expression data expressed by the user through voice or text input using natural language to increase, decrease, or block some recommended content. For example: "I want to see more news about Type 99 tanks", "I will uninstall the news about Hammer mobile phone again!", "There are too many news about Xiaomi, have you charged money?" etc. Users can choose Text input mode or voice input mode, the text input mode is a common input mode, which will not be described in detail here; and the voice input mode, the user's voice will first be converted into text information by the voice recognition module, and there are many mature methods for completing this step. Solutions, such as the existing open platform for speech recognition. As long as the recommendation service uses the open software development kit (SDK: Software Development Kit) of these speech recognition service providers, the speech recognition service of these platforms can be used for free, and this technology has been very mature in practice, and the recognition accuracy rate is as high as 95%. .

The main difficulty in identifying the intention expressed by the user in natural language is to determine the corresponding object of interest and user attitude based on the expression data, that is, step 102 and step 103 . Simplify the user's intent recognition problem to the recognition of a two-tuple:

<Object, Attitude>

Among them, Object represents the object of interest targeted by the user. In the above example, the objects of interest are "Type 99 tank", "Hammer mobile phone", and "Xiaomi (Xiaomi Technology Co., Ltd.)". Attitude represents the user's attitude, which can be represented by a limited set, such as: {'block', 'reduce recommendation', 'increase recommendation'}.

In practical applications, the determination of an Object needs to rely on an Object knowledge base, that is, an object knowledge base. The object knowledge base needs a large number of meaningful Objects confirmed by humans. The knowledge base contains the alias (ref), type (type) and name (name) of the object of interest corresponding to the object of interest. The type can be category (category) or Entity (entity: name of person, name of institution, title of work, etc.). E.g:

<ref:[Technology News, Technology Articles, Technology], type: category, name: Technology>

<ref:[Xiaomi, Xiaomi Technology], type: entity, name: Xiaomi Technology Co., Ltd.>

<ref:[Fan Bingbing], type: entity, name: Fan Bingbing>

With the object knowledge base, the Object in a paragraph can be identified according to the alias mapping.

The determination of Attitude is mainly based on some manually defined rules, such as:

Less XXX - 'Fewer Referrals'

More recommended points XXX--'increase recommendation'

Don't XXX--'shield'

It should be noted that in one case, a word in a paragraph may map to multiple Objects. For example, 'Xiaomi' may refer to Xiaomi Technology Co., Ltd., or it may refer to edible millet. In this case, the user can be prompted , allowing the user to confirm which meaning they want to express, or judge which Object is more relevant to the user through the user's usage records. For example, if the user expresses a negative attitude, the Object that appears in the user's browsing history is more likely to be the object that the user wants to express, so as to confirm the Object that the user really wants to express.

In another case, the Object in the user's intention is not a specific concept, but rather an abstract concept. For example, expressions such as "Let's have something interesting" and "Don't make any more vulgar ones". Here 'interesting' and 'vulgar' are relatively abstract concepts that cannot be mapped to a specific Object. In this case, you can pre-define some special tags, such as 'dance', 'vulgar content' as Object. These tags can be marked by operators, or they can be automatically identified by machine learning algorithms. In this way, this kind of abstract demand can also be identified, and then the Object can be determined.

Next, an interest prediction result is determined according to the object of interest and user attitude corresponding to the expression data. The application client is prompted with the interest prediction results, the user selects and clicks one of the interest prediction results, and the click data of the user on the interest prediction results is obtained to determine the user interest.

For example: in the example above, "I will uninstall the news of Hammer phone!", it can be determined that the Object is "Hammer phone" and the Attitude is "Block", and the application client can be prompted that the interest prediction result is "Block Hammer phone news". For the user to click, if the prediction result does not meet the user's real intention, the user can ignore it; or prompt "Block Smartphone News?", "Yes" and "No" will appear for the user to choose, and the user chooses to click one of the interest prediction results , to obtain the user's click data on the interest prediction result, so as to determine the user's interest.

Another example: in the above example "I want to read more news about Type 99 tanks", it can be determined that the Object is "Type 99 tanks" and the Attitude is "Add recommendation", then the application client can be prompted that the interest prediction result is "Recommended Type 99 Tank R&D Progress News", "Recommended Type 99 Tank Putting into Use News" and "Recommended foreign media news about Type 99 Tank" and other options for users to choose. The user chooses to click on one of the interest prediction results to obtain the user's opinion on the Click data of the above interest prediction results to determine user interests.

This embodiment provides a method for discovering user interests. The method determines the interest prediction result by obtaining the expression data of the recommended content input by the user, then prompts the application client for the interest prediction result, and obtains the user's click on the interest prediction result. data to determine user interests. Users can actively express their requirements for recommended content instead of passively waiting for the system to understand their interests through a long data mining process, and users can express their needs through natural language, and the operation cost is very low. This embodiment can timely and accurately discover user interests and improve user experience.

Embodiment two

Fig. 2 is a flow chart of a method for discovering user interest provided by this embodiment, the method includes the following steps:

201. Acquire the expression data of the recommended content input by the user.

202. Determine an object of interest corresponding to the expression data according to the expression data, including:

matching the words contained in the expression data in the object knowledge base;

The words that are successfully matched are used as the object of interest corresponding to the expression data.

203. According to the expression data, determine a user attitude corresponding to the expression data, including:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

204. According to the text distance between the interest object and the word corresponding to the user attitude, filter out the interest object whose text distance is greater than a set value.

205. Determine an interest prediction result according to the object of interest and user attitude corresponding to the expression data.

206. Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

In this embodiment, the expression data may specifically be interest expression data expressed by the user through voice or text input using natural language to increase, decrease, or block some recommended content. The user can use text input mode or voice input mode.

The main difficulty in identifying the intention expressed by the user in natural language is to determine the corresponding object of interest and user attitude based on the expression data, that is, step 202 and step 203 . Simplify the user's intent recognition problem to the recognition of a two-tuple:

<Object, Attitude>

Among them, Object represents the object of interest targeted by the user, and Attitude represents the user's attitude, which can be represented by a limited set, such as: {'block', 'reduce recommendation', 'increase recommendation'}.

In practical applications, the determination of an Object needs to rely on an Object knowledge base, that is, an object knowledge base. The object knowledge base needs a large number of manually confirmed meaningful Objects. The knowledge base contains the alias (ref), type (type) and name (name) of the object of interest corresponding to the object of interest. The type can be category (category) or Entity (entity: name of person, name of institution, title of work, etc.). With the object knowledge base, the Object in a paragraph can be identified according to the alias mapping.

The determination of Attitude is mainly based on some manually defined rules, such as:

Less XXX - 'Fewer Referrals'

More recommended points XXX--'increase recommendation'

Don't XXX--'shield'

It should be noted that in one case, multiple objects can be identified in a sentence, such as "There are too many Xiaomi news, is it an advertisement?" In this sentence, both "Xiaomi news" and "advertisement" can be found The corresponding Object. In this case, it can be judged in conjunction with the identification of Attitude, that is, step 204. If the text distance between an Object and an Attitude is greater than the set value, the Object can be filtered out, thereby confirming the Object that the user really wants to express, and setting The value can be preset by the user's interest discovery device, or it can be set by the user. If the set value is "2", then when the text distance between an Object and an Attitude is greater than 2, the Object can be filtered out. The above example "Too much" can be determined as Attitude, the text distance between "Xiaomi News" and "Too much" is 0, and the text distance between "Advertising" and "Too much" is 3, then it can be filtered to "Advertising "This Object, to confirm the "Xiaomi news" that the user really wants to express.

Next, an interest prediction result is determined according to the object of interest and user attitude corresponding to the expression data. The application client is prompted with the interest prediction results, the user selects and clicks one of the interest prediction results, and the click data of the user on the interest prediction results is obtained to determine the user interest.

This embodiment provides a method for discovering user interests. On the basis of Embodiment 1, the method increases the text distance between the words corresponding to the object of interest and the attitude of the user, and filters out the objects of interest whose text distance is greater than the set value. , to improve the accuracy of determining the object of interest. This embodiment can timely and accurately discover user interests and improve user experience.

Embodiment three

FIG. 3 is a flow chart of a method for discovering user interests provided in this embodiment, the method comprising the following steps:

301. Obtain the behavior data of the recommended content input by the user;

302. Determine user satisfaction with the recommended content according to the behavior data. Wherein, determining the user's satisfaction with the recommended content may be implemented in multiple ways, including at least one of the following:

Determine the user's satisfaction with the recommended content according to the user's refresh frequency of the recommended content;

Determine the user's satisfaction with the recommended content based on the user's click data and stay time on the recommended content;

According to the user's support feedback data and attention time on the recommended content, the user's satisfaction with the recommended content is determined.

303 . Determine whether the satisfaction degree is less than the set threshold value. If the satisfaction degree is less than the set threshold value, trigger the execution of the operation of obtaining the expression data of the recommended content input by the user, that is, execute step 304 .

304. Obtain the expression data of the recommended content input by the user.

305. Determine an object of interest corresponding to the expression data according to the expression data, including:

matching the words contained in the expression data in the object knowledge base;

The words that are successfully matched are used as the object of interest corresponding to the expression data.

306. Determine the user attitude corresponding to the expression data according to the expression data, including:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

307. According to the text distance between the interest object and the word corresponding to the user attitude, filter out the interest object whose text distance is greater than a set value.

308. Determine an interest prediction result according to the object of interest and user attitude corresponding to the expression data.

309 . Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

In this embodiment, the behavior data can specifically be data corresponding to the user’s behavior during use, such as the refresh frequency of the recommended content, the click data and length of stay on the recommended content, the support feedback data and attention time for the recommended content, etc. , determine the user’s satisfaction with the recommended content according to these behavior data, set the satisfaction threshold, if the determined satisfaction is higher than the threshold, the recommendation can be continued, if the satisfaction is lower than the threshold, then Trigger the execution of the operation of obtaining the expression data of the recommended content input by the user. Subsequent operations are exactly the same as those in Embodiment 2, and will not be described in detail here.

As a further illustration, in this embodiment, according to the user's refresh frequency of the recommended content, the method of determining the user's satisfaction with the recommended content may be as follows: when the user is not satisfied with the recommended content, the time for the user to read the recommended content will decrease Short, and the length of reading time is basically inversely proportional to the refresh rate. When the refresh rate is high, the corresponding satisfaction is low, and when the refresh rate is low, the corresponding satisfaction is high. Assume that the refresh rate is 5 times per hour. Satisfaction is the threshold value. When the user's refresh frequency is 6 times per hour, the operation of obtaining the expression data of the recommended content input by the user will be triggered;

According to the user's click data and length of stay on the recommended content, the way to determine the user's satisfaction with the recommended content can be: use a scoring system, when the user clicks on a recommended article, it means that the user is interested in the article, and can be used for The content is scored, such as 1 point. When the user reads the article, the reading time is scored, such as 1 point for every 30 seconds of reading, and 4 points for reading for 2 minutes. Of course, the score can be calculated according to the length of the article content. The adjustment will not be described in detail here. Assume that the cumulative score of 10 recommended articles is the satisfaction score, and the threshold value is 20 points. When the user ignores many recommended articles, the satisfaction score will be very low. When When the satisfaction score is lower than 20 points, it will trigger the execution of the operation of obtaining the expression data of the recommended content input by the user;

According to the user's support feedback data and attention time for the recommended content, the way to determine the user's satisfaction with the recommended content can be as follows: each recommended article can be equipped with user satisfaction survey options and set corresponding scores, such as "Very satisfied" is 5 points, "very satisfied" is 4 points, "general" is 3 points, "dissatisfied" is 2 points, and "very dissatisfied" is 1 point. Users can recommend for this article by choosing one of the options Score the article to get the user’s support feedback data for the recommended content. The attention time can be the user’s attention time for the recommended content in a certain field, such as entertainment, sports, economy, etc. If the attention time is long, the recommendation in this field can be increased. The satisfaction score of the content, and finally the final satisfaction score is obtained by combining the above two scores. When the satisfaction score is lower than the threshold value, it will trigger the execution of the operation of obtaining the expression data of the recommended content input by the user. The above are only examples of specific implementation manners, and are not limited to the above implementation manners in practical applications.

This embodiment provides a method for discovering user interests. On the basis of Embodiment 2, the method adds the behavior data of the recommended content input by the user, and determines the user's preference for the recommended content according to the behavior data. Satisfaction steps. If the satisfaction is high, the steps for users to actively express data can be reduced to further improve user experience.

Embodiment four

FIG. 4 is a flow chart of a method for discovering user interests provided in this embodiment, the method comprising the following steps:

401. Obtain the behavior data of the recommended content input by the user;

402. Determine user satisfaction with the recommended content according to the behavior data.

403. Determine whether the satisfaction degree is less than a set threshold value, and if the satisfaction degree is less than the set threshold value, trigger the execution of an operation of acquiring the expression data of the recommended content input by the user.

404. Obtain the expression data of the recommended content input by the user.

405. Determine an object of interest corresponding to the expression data according to the expression data, including:

matching the words contained in the expression data in the object knowledge base;

The words that are successfully matched are used as the object of interest corresponding to the expression data.

406. Determine the user attitude corresponding to the expression data according to the expression data, including:

Matching the words contained in the expression data in the preset emotional attitude template;

According to the matching result, the user attitude corresponding to the expression data is determined.

407. According to the text distance between the interest object and the word corresponding to the user attitude, filter out the interest object whose text distance is greater than a set value.

408. Determine an interest prediction result according to the object of interest and user attitude corresponding to the expression data.

409. Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

410. Correct the recommended content pushed to the user according to the determined interest of the user.

A correction rule may be formulated for the correction, and there may be many kinds of correction rules, which are only described here as examples. For example, you can set an initial score for each article, and adjust the score of the article according to the user's interest. Articles with high scores are more likely to be recommended, while articles with low scores are more difficult to be recommended, and articles with zero scores will be recommended. Hidden. For example: for the user interest <'Xiaomi', 'reduce recommendation'>, two correction rules will be triggered: "The score of articles containing Xiaomi in the title will be reduced by 50%" and "The score of articles containing Xiaomi in the keyword will be reduced by 20%". After the revised rules take effect, the scores of articles that hit "Xiaomi" will be penalized, making it harder to be recommended. The effective time of the modification rule can be determined according to the number of times the user expresses it. If it is the user's first expression, the effective time can be three days, the second time can be one week, and the third time can be one month; and for <'Huawei','add recommendation'>, it can trigger "the title contains Huawei Of course, the correction rule can also be in other forms, such as: "If the number of articles containing Huawei in the title of the recommended candidate article is less than 5, search the database for 5 such articles and add them to the recommendation candidate ". In addition, the correction rules can be adjusted according to the frequency of the user's active expression, so that the user can see the improvement of the recommendation result faster.

This embodiment provides a method for discovering user interest. On the basis of Embodiment 3, the method adds a step of correcting the recommended content pushed to the user according to the determined user interest. This step can make the user's expression take effect immediately. After the expression is completed, the user can see the improvement of the recommendation result in the next recommendation, which can further improve the user experience.

Embodiment five

FIG. 5 is a flowchart of a method for discovering user interests provided by this embodiment, the method includes the following steps:

501. Acquire behavior data on recommended content input by a user.

502. Determine an interest prediction result according to the behavior data.

503. Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

In this embodiment, the behavior data can specifically be data corresponding to the user’s behavior during use, such as the refresh frequency of the recommended content, the click data and length of stay on the recommended content, the support feedback data and attention time for the recommended content, etc. , as a further illustration, when the user is dissatisfied with the recommended content, the time for the user to read the recommended content will be short, and the length of reading time is basically inversely proportional to the refresh frequency. When the refresh frequency is high, it means that the user is not interested in the recommended content. , and when the refresh frequency is low, it means that the user is interested in the recommended content, so as to determine the interest prediction result for the user to choose; when the user clicks on a recommended article, it means that the user is interested in the article, and when the user spends a long time reading the article, I am very interested in the article, so as to determine the interest prediction results for the user to choose; in addition, user satisfaction survey options can be set in each recommended article, such as "very satisfied", "very satisfied", "general", "not Satisfied" and "Very Dissatisfied", the user can get the user's support feedback data for the recommended content by selecting one of the options, and the attention time can be the user's attention time for the recommended content in a certain field, such as entertainment, sports and Economy, etc., if the attention time is long, it means that the user is interested in the recommended content in this field, and finally the interest prediction result is determined for the user to choose. The above are only examples of specific implementation manners, and are not limited to the above implementation manners in practical applications.

For example: If the user ignores many recommended articles about Xiaomi, then the user is likely not interested in Xiaomi. If a user clicks more on articles about electric cars, then the user is likely to be interested in news about Tesla founder Musk. Traditionally, this information would be directly used to recommend less or more certain types of news. However, since user behavior is complex, such predictions may be wrong. Blindly recommending articles based on these predictions may not improve user satisfaction. Therefore, according to the behavior data, the interest prediction result can be determined, the application client can be prompted with the interest prediction result, and the click data of the user on the interest prediction result can be obtained to determine the user interest. For example, the following interest prediction results can be prompted to the client,

Do you want to:

Reduce the recommendation about Xiaomi

Recommend Musk's news

Block firewood news

If one of the above prediction results meets the user's wishes, the user can click on the corresponding interest prediction result to determine the user's interest.

This embodiment provides a user interest discovery method, the method determines the interest prediction result according to the user behavior data, prompts the application client for the interest prediction result, and obtains the user's click data on the interest prediction result to determine user interest. This embodiment can predict the user's possible recommendation needs according to the user behavior data. Once the prediction is successful, the user only needs to confirm, which can save the voice or text input step, and further improve the user experience.

Embodiment six

FIG. 6 is a flowchart of a method for discovering user interests provided by this embodiment, the method includes the following steps:

601. Acquire behavior data on recommended content input by a user.

602. Determine user satisfaction with the recommended content according to the behavior data.

603 . Determine whether the satisfaction degree is less than a set threshold value, and if so, trigger an operation of determining an interest prediction result according to the behavior data, that is, perform step 604 .

604. Determine an interest prediction result according to the behavior data.

605. Prompt the application client for the interest prediction result, and acquire click data of the user on the interest prediction result, so as to determine the user's interest.

In this embodiment, the behavior data can specifically be data corresponding to the user’s behavior during use, such as the refresh frequency of the recommended content, the click data and length of stay on the recommended content, the support feedback data and attention time for the recommended content, etc. , determine the user’s satisfaction with the recommended content according to these behavior data, set the satisfaction threshold, if the determined satisfaction is higher than the threshold, the recommendation can be continued, if the satisfaction is lower than the threshold, then It is triggered to determine the interest prediction result according to the behavior data.

This embodiment provides a user interest discovery method, which determines the user's satisfaction with the recommended content according to the behavior data, and if the satisfaction is less than the set threshold value, triggers the execution of determining the interest prediction result based on the behavior data, The application client is prompted with the interest prediction result, and the click data of the user on the interest prediction result is acquired to determine the user interest. It can reduce the number of steps for users to actively perform data expression, greatly reduce the cost of use, and further improve user experience.

Embodiment seven

FIG. 7 is a flowchart of a method for discovering user interests provided by this embodiment, the method includes the following steps:

701. Acquire behavior data on recommended content input by a user.

702. Determine an interest prediction result according to the behavior data.

703. Prompt the application client for the interest prediction result, and acquire user click data on the interest prediction result, so as to determine the user interest.

704. Correct the recommended content pushed to the user according to the determined interest of the user.

In this embodiment, on the basis of the fifth embodiment, a step of modifying the recommended content pushed to the user according to the determined user interest is added, which can further improve the user experience.

Embodiment Eight

Based on the above-mentioned embodiments, this embodiment provides a user interest discovery device, as shown in FIG. 8 , which is a structural block diagram of the device of this embodiment. The device may include:

The user data acquisition module 801 is configured to acquire the expression data or behavior data of the recommended content input by the user.

An interest prediction result determination module 802, configured to determine an interest prediction result according to the expression data or the behavior data, including: an interest object determination unit, configured to determine an interest object corresponding to the expression data according to the expression data The user attitude determination unit is used to determine the user attitude corresponding to the expression data according to the expression data; the interest prediction result determination unit is used to determine the interest prediction according to the interest object and user attitude corresponding to the expression data result. Wherein, the interest object determination unit specifically includes: a first matching subunit, used to match the words contained in the expression data in the object knowledge base; an interest object determination subunit, used to use the successfully matched words as the expression data Corresponding interest object; the interest object filtering subunit, used to filter the word according to the text distance between the interest object and the word corresponding to the user attitude after using the successfully matched word as the interest object corresponding to the expression data Objects of interest whose text distance is greater than the set value are discarded. Wherein, the user attitude determination unit includes: a second matching subunit, used to match the words contained in the expression data in the preset emotional attitude template; The user attitude corresponding to the data.

The interest determination module 803 is configured to prompt the application client for the interest prediction result, and acquire user click data on the interest prediction result to determine the user interest.

The first satisfaction degree determination module 804 is configured to determine the user's satisfaction level with the recommended content according to the behavior data before acquiring the expression data of the recommended content input by the user, including at least one of the following items: the first satisfaction The degree determination unit is used to determine the user's satisfaction degree of the recommended content according to the user's refresh frequency of the recommended content; the second satisfaction degree determination unit is used to determine the user's satisfaction with the recommended content according to the user's click data on the recommended content and the length of stay. Satisfaction level of the recommended content; a third satisfaction level determination unit, configured to determine the user's satisfaction level with the recommended content according to the user's support feedback data and attention time on the recommended content.

The expression data acquisition triggering module 805 is configured to trigger the execution of the operation of acquiring the expression data of the recommended content input by the user if the satisfaction degree is less than the set threshold value.

The second satisfaction determination module 806 is configured to determine the user's satisfaction with the recommended content according to the behavior data before determining the interest prediction result according to the behavior data, including at least one of the following: first satisfaction The degree determination unit is used to determine the user's satisfaction degree of the recommended content according to the user's refresh frequency of the recommended content; the second satisfaction degree determination unit is used to determine the user's satisfaction with the recommended content according to the user's click data on the recommended content and the length of stay. Satisfaction level of the recommended content; a third satisfaction level determination unit, configured to determine the user's satisfaction level with the recommended content according to the user's support feedback data and attention time on the recommended content.

The interest prediction result determination triggering module 807 is configured to trigger determining the interest prediction result according to the behavior data if the satisfaction degree is less than a set threshold.

The push module 808 is configured to prompt the application client for the interest prediction result, and obtain the user's click data on the interest prediction result to determine the user's interest, and modify the recommended content pushed to the user according to the determined user interest .

A device for discovering user interest provided by an embodiment of the present invention can execute the method for discovering user interest provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; preferred implementations in the examples are not to limit them, and for those skilled in the art, this The invention is capable of various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims (16)

1. A kind of 1. user interest discovery method, it is characterised in that including：

   The expression data or behavioral data to content recommendation of user's input are obtained, wherein the expression data are that user uses certainly The interest expression data of increase, reduction or shielding content recommendation expressed by right language；

   According to the expression data or the behavioral data, interest prediction result is determined；

   The interest prediction result is prompted to applications client, and obtains click data of the user to the interest prediction result, To determine user interest；

   Wherein, according to the expression data, interest prediction result is determined, including：

   According to the expression data, it is determined that object of interest corresponding with the expression data and user's attitude；

   According to object of interest corresponding with the expression data and user's attitude, interest prediction result is determined.
2. 2. according to the method for claim 1, it is characterised in that obtaining the expression data to content recommendation of user's input Before, in addition to：

   Obtain the behavioral data to content recommendation of user's input；

   According to the behavioral data, satisfaction of the user to the content recommendation is determined；

   If the satisfaction is less than setting threshold value, triggering performs the expression data to content recommendation for obtaining user's input Operation.
3. 3. according to the method for claim 1, it is characterised in that according to the expression data, it is determined that with the expression data Corresponding object of interest, including：

   The word that the expression data are included is matched in object repository；

   Using the word that the match is successful as object of interest corresponding to the expression data.
4. 4. according to the method for claim 3, it is characterised in that the word that the match is successful is corresponding as the expression data Object of interest after, in addition to：

   According to the text distance between object of interest word corresponding with user's attitude, filter out text distance and be more than and set The object of interest of definite value.
5. 5. according to the method for claim 1, it is characterised in that according to the expression data, it is determined that with the expression data Corresponding user's attitude, including：

   The word that the expression data are included is matched in default emotional attitude template；

   According to matching result, it is determined that user's attitude corresponding with the expression data.
6. 6. according to the method for claim 1, it is characterised in that according to the behavioral data, determine interest prediction result Before, in addition to：

   According to the behavioral data, satisfaction of the user to the content recommendation is determined；

   If the satisfaction is less than setting threshold value, triggering is performed according to the behavioral data, determines interest prediction result Operation.
7. 7. the method according to claim 2 or 6, it is characterised in that according to the behavioral data, determine that user pushes away to described The satisfaction of content is recommended, including it is at least one of following：

   Refreshing frequency according to user to content recommendation, determine satisfaction of the user to the content recommendation；

   Click data and stay time according to user to content recommendation, determine satisfaction of the user to the content recommendation；

   According to support feedback data of the user to content recommendation and concern time, satisfaction of the user to the content recommendation is determined Degree.
8. 8. according to any described methods of claim 1-6, it is characterised in that prompting the interest to predict to applications client As a result, and click data of the user to the interest prediction result is obtained, after determining user interest, in addition to：

   According to the user interest of determination, the content recommendation pushed to user is corrected.
9. 9. a kind of user interest finds device, it is characterised in that including：

   User data acquisition module, for obtaining the expression data or behavioral data to content recommendation of user's input, wherein institute It is that user uses the increase expressed by natural language, reduction or the interest expression data for shielding content recommendation to state expression data；

   Interest prediction result determining module, for according to the expression data or the behavioral data, determining interest prediction result；

   Interest determination module, for prompting the interest prediction result to applications client, and it is pre- to the interest to obtain user The click data of result is surveyed, to determine user interest；

   Wherein, the interest prediction result determining module includes：

   Object of interest determining unit, for according to the expression data, it is determined that object of interest corresponding with the expression data；

   User's attitude determining unit, for according to the expression data, it is determined that user's attitude corresponding with the expression data；

   Interest prediction result determining unit, for basis object of interest corresponding with the expression data and user's attitude, it is determined that Interest prediction result.
10. 10. device according to claim 9, it is characterised in that described device also includes：

    First satisfaction determining module, for obtain user input the expression data to content recommendation before, according to described Behavioral data, determine satisfaction of the user to the content recommendation；

    Data acquisition trigger module is expressed, for when the satisfaction is less than setting threshold value, it is defeated that triggering execution obtains user The operation of the expression data to content recommendation entered.
11. 11. device according to claim 9, it is characterised in that object of interest determining unit includes：

    First coupling subelement, the word included for matching the expression data in object repository；

    Object of interest determination subelement, for using the word that the match is successful as object of interest corresponding to the expression data.
12. 12. device according to claim 11, it is characterised in that object of interest determining unit also includes：

    Object of interest filters subelement, for using the word that the match is successful as object of interest corresponding to the expression data it Afterwards, according to the text distance between object of interest word corresponding with user's attitude, filter out text distance and be more than and set The object of interest of definite value.
13. 13. device according to claim 9, it is characterised in that user's attitude determining unit includes：

    Second coupling subelement, the word included for matching the expression data in default emotional attitude template；

    User's attitude determination subelement, for according to matching result, it is determined that user's attitude corresponding with the expression data.
14. 14. device according to claim 9, it is characterised in that described device also includes：

    Second satisfaction determining module, for according to the behavioral data, before determining interest prediction result, according to the row For data, satisfaction of the user to the content recommendation is determined；

    Interest prediction result determines trigger module, for when the satisfaction is less than setting threshold value, triggering to be performed according to institute Behavioral data is stated, determines the operation of interest prediction result.
15. 15. device according to claim 10, it is characterised in that the first satisfaction determining module includes following at least one ：

    First satisfaction determining unit, for the refreshing frequency according to user to content recommendation, determine user in the recommendation The satisfaction of appearance；

    Second satisfaction determining unit, for the click data and stay time according to user to content recommendation, determine user couple The satisfaction of the content recommendation；

    3rd satisfaction determining unit, for according to support feedback data of the user to content recommendation and concern time, it is determined that with Satisfaction of the family to the content recommendation.
16. 16. according to any described devices of claim 9-15, it is characterised in that described device also includes：

    Pushing module, for prompting the interest prediction result to applications client, and obtain user and the interest is predicted As a result click data, after determining user interest, according to the user interest of determination, correct in the recommendation pushed to user Hold.