Data processing method and device

Abstract

The invention discloses a data processing method and device, and the method comprises the steps: obtaining a first operation triggered by a user for the current display content, and enabling the first operation to have related content; user behavior data in a set time period are obtained based on the first operation, the user behavior data are implicit interaction data collected by a user based on control operation of an input device in the process of accessing the current display content, and the implicit interaction data can represent the concerned content of the user; and determining user intention data based on the user behavior data and the associated content.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a data processing method and apparatus. Background Art

[0002] During the work process, there are often situations where additional information or data support is needed. For example, when viewing a technical document, if some technical terms are not understood, the current work task needs to be paused and relevant content needs to be retrieved, browsed, or searched using a search engine. The above process will seriously affect the user's work efficiency. Summary of the Invention

[0003] In view of this, this application provides the following technical solutions:

[0004] The first aspect of this application provides a data processing method, which includes:

[0005] Obtain a first operation triggered by the user for the currently displayed content, and the first operation has associated content;

[0006] Based on the first operation, obtain the user behavior data for a set period. The user behavior data is implicit interaction data collected based on the control operations of the input device during the process of the user accessing the currently displayed content, and the implicit interaction data can represent the content that the user is concerned about;

[0007] Determine the user intention data based on the user behavior data and the associated content.

[0008] In a possible implementation manner, the first operation is a content selection operation, and the content selection operation includes any one of the following: box selection operation, circle selection operation, swipe selection operation, line drawing operation;

[0009] The associated content at least includes the content selected by the content selection operation.

[0010] In a possible implementation manner, the obtaining the user behavior data for a set period based on the first operation includes:

[0011] Based on the first operation, obtain the user behavior data for a set period before the user triggers the first operation. The user behavior data includes at least one of the following: access status data of the current content, cursor movement data, page scrolling data, and the user behavior data has a fixed data structure.

[0012] In a possible implementation manner, the determining the user intention based on the user behavior data and the associated content includes:

[0013] Perform semantic analysis on the associated content to obtain semantic features;

[0014] Fuse the semantic features and the user behavior data to obtain fused features;

[0015] Perform intent recognition on the fused features and a local knowledge base to determine the user intent.

[0016] In a possible implementation, the user intent data includes at least one alternative intent, and the method further includes:

[0017] Obtain a selection operation of the user, where the selection operation is an operation of selecting at least one alternative intent from the at least one alternative intent;

[0018] Obtain reference data corresponding to the selected alternative intent based on the selection operation and output the reference data.

[0019] In a possible implementation, the method further includes:

[0020] Obtain intended intent data input by the user;

[0021] Obtain reference data corresponding to the intended intent data and output the reference data.

[0022] In a possible implementation, outputting the reference data includes:

[0023] Determine an output form of the reference data based on a type of the reference data or a relationship between the reference data and the associated content;

[0024] Output the reference data based on the output form.

[0025] In a possible implementation, it further includes:

[0026] Feed back the intended intent data to an intent generation model for outputting user intent data to enrich an output type of the intent generation model.

[0027] In a possible implementation, it further includes:

[0028] Construct a context feature vector, where the context feature vector is comprehensive information for expressing semantic features of user behavior and associated content, and the context feature vector includes a plurality of accumulation terms, including accumulation terms for various types of user behavior, and accumulation terms for different user behaviors have different weights and / or adjustment coefficients;

[0029] Train and optimize weights and / or adjustment parameters of the accumulation terms for various types of user behavior in the context feature vector based on feedback data of the user.

[0030] A second aspect of the present application provides a data processing device, and the device includes:

[0031] An operation acquisition module, configured to acquire a first operation triggered by a user for the currently displayed content, where the first operation has associated content;

[0032] A behavior acquisition module, configured to acquire user behavior data for a set time period based on the first operation, where the user behavior data is implicit interaction data collected based on control operations of an input device during the user's access to the currently displayed content, and the implicit interaction data can represent the content that the user is concerned about;

[0033] An intention determination module, configured to determine user intention data based on the user behavior data and the associated content.

[0034] A third aspect of the present application provides an electronic device, including at least one processor and a memory connected to the processor, where:

[0035] The memory is used to store a computer program;

[0036] The processor is configured to execute the computer program so that the electronic device can implement any of the above data processing methods. Description of the Drawings

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative efforts.

[0038] Figure 1 It is a flowchart of a data processing method disclosed in an embodiment of the present application;

[0039] Figure 2 It is an example diagram of various first operation selection contents disclosed in an embodiment of the present application;

[0040] Figure 3 It is a flowchart of determining user intention disclosed in an embodiment of the present application;

[0041] Figure 4 It is a schematic diagram of the overall implementation architecture of user intention inference disclosed in an embodiment of the present application;

[0042] Figure 5 It is an example diagram of the display of a reference data disclosed in an embodiment of the present application;

[0043] Figure 6 It is an example diagram of the display of another reference data disclosed in an embodiment of the present application;

[0044] Figure 7 The flowchart of personalized adaptation of the data processing solution disclosed in the embodiments of the present application;

[0045] Figure 8 The structural schematic diagram of a data processing device disclosed in the embodiments of the present application;

[0046] Figure 9 The structural schematic diagram of an electronic device disclosed in the embodiments of the present application. Detailed implementation manners

[0047] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0048] The embodiments of the present application can be applied to an electronic device. The present application does not limit the product form of the electronic device, which may include but is not limited to smart phones, tablet computers, wearable devices, personal computers (PCs), netbooks, etc., and can be selected according to application requirements.

[0049] In view of the situation that when a user accesses display content using an electronic device, there may be doubts about some text in the content and other supporting content is needed to help understand or sort it out, the present application discloses a data processing method to better support the user's understanding or sorting of the display content during the process of the user accessing the display content and improve the access efficiency.

[0050] Figure 1 The flowchart of a data processing method disclosed in the embodiments of the present application. Refer to Figure 1 As shown, the data processing method may include:

[0051] Step 101: Obtain a first operation triggered by the user for the current display content, and the first operation has associated content.

[0052] Among them, the current display content may be document content, web page display content, or display content of other relevant applications that can display text. The present application does not fixedly limit this. The first operation triggered by the user for the current display content may be an operation that can select or mark a certain or certain text content, that is, the first operation may be a content selection operation, and the content selection operation may include but is not limited to: a selection operation, a circle selection operation, a swipe selection operation, a line drawing operation, etc. Figure 2 The exemplary diagram of various first operation selection contents disclosed in the embodiments of the present application.

[0053] The first operation has associated content, which at least includes the content selected by the content selection operation, such as Figure 2 "cosine similarity" in Figure 2 . Currently, in other implementations, in addition to the content selected by the first operation, the associated content may further include the context of the selected content, such as the paragraph where the selected content is located, or the paragraph where the selected content is located, the previous paragraph, and the next paragraph of the paragraph where the selected content is located.

[0054] In the actual application process, the data processing solution disclosed in this application can be configured to be enabled or disabled. Only when the solution is enabled, the user will trigger subsequent processing when performing the first operation, and a selection mark (such as a square, a straight line, etc.) will be displayed on the current displayed content based on the first operation triggered by the user. When the data processing solution is not enabled, the same first operation of the user will not trigger the execution of subsequent related processing, nor will a selection mark be displayed in the current displayed content.

[0055] Step 102: Obtain the user behavior data for a set period based on the first operation. The user behavior data is implicit interaction data collected based on the control operation of the input device during the user's access to the current displayed content, and the implicit interaction data can characterize the content that the user is concerned about.

[0056] When the user browses the displayed content, the user usually makes some unconscious actions. For example, when browsing a certain paragraph of text, the user usually controls the cursor to slide synchronously below the text being browsed while silently reading the text content in mind; or, when the user views a certain piece of text, and suddenly associates with the relevant content where this piece of text appeared before, the user may move the cursor from the current position to the relevant content before and click the left mouse button; or, after the user views the content of the current paragraph for a long time, the user controls the scroll wheel on the mouse to switch the displayed content, etc. Although these control operations on the input device do not actually interact with the displayed content, they can to a certain extent reflect the user's viewing position or viewing focus, and can be called the user's implicit interaction data. The implicit interaction data can be automatically collected by the device when the user accesses the displayed content to facilitate subsequent determination of the content that the user is concerned about.

[0057] It should be noted that the user behavior data will be collected in real time during the user's browsing or editing of the displayed content. However, considering the data storage cost and the fact that the user behavior data a long time ago has no correlation with the user intention corresponding to the first operation triggered by the current user, which may affect the accuracy of the intention judgment for the first operation triggered by the current user. Therefore, in the embodiments of this application, after the user triggers the first operation, only the user behavior data for a set period is obtained. The set period can be a period of a fixed duration before the current moment, such as within the previous 3 minutes, or within the previous 2 minutes.

[0058] Step 103: Determine user intent data based on the user behavior data and the associated content.

[0059] After obtaining the user behavior data, the user intent data can be determined by combining the user behavior data and the associated content. For example, if the associated content of the first operation triggered by the user is "deep learning", and the user behavior data indicates that the user controls the cursor to cross "deep learning" twice successively, it can be determined that the user may not fully understand the meaning of "deep learning", and the user intent data is determined to be the interpretation of the concept of "deep learning". Another example, if the associated content of the first operation triggered by the user is the sales data of a certain manufacturer, the user intent data can be determined to be obtaining the sales data of similar enterprises for comparison.

[0060] In the data processing method of this embodiment, after the user triggers a first operation on the current displayed content, the user behavior data for a set time period will be obtained, and the user intent will be comprehensively determined by combining the user behavior data and the associated content of the first operation; since the user behavior data is implicit interaction data that can represent the content concerned by the user, using the user behavior data as the basis for determining the user intent can further improve the accuracy of intent inference.

[0061] In the above embodiment, the obtaining of the user behavior data for a set time period based on the first operation may include: based on the first operation, obtaining the user behavior data of the user in a set time period before triggering the first operation, and the user behavior data includes at least one of the following: access status data of the current content, cursor movement data, page scrolling data, and the user behavior data has a fixed data structure.

[0062] Considering that the user behavior data a long time ago has a relatively small relevance to the intent at the moment when the user triggers the first operation and may reduce the accuracy of intent inference based on the first operation, in this application, the set time period may be a time period of a fixed duration forward from the moment when the user triggers the first operation.

[0063] Among them, the access status content of the current content may include, but is not limited to, the browsing duration per unit character of the current displayed content, the relative browsing duration per unit character, the input duration per unit character, the relative input duration per unit character, etc.; the cursor movement data may include, but is not limited to, cursor Hover (slide over but do not click), cursor word selection, cursor position switching, etc.

[0064] Specifically, the user behavior involved in the user's access to the current displayed content may include:

[0065] Browsing duration per unit character: browsing duration of the whole document or content / number of text characters; average speed accumulated under the same type of task;

[0066] Relative browsing duration per character: Browsing duration of a certain piece of text / number of characters in the text; if the relative browsing duration per character is greater than the baseline level, it indicates in-depth processing and thinking, and here there may be intentions related to this piece of text.

[0067] Input duration per character: Input duration of the entire document or content / number of characters in the text; the cumulative average speed under the same type of task.

[0068] Relative input duration per character: Input duration of a certain piece of text / number of characters in the text; if the relative input duration per character is greater than the baseline level, it indicates in-depth processing and thinking, and here there may be intentions related to this piece of text.

[0069] Mouse Hover (hovering but not clicking): It is a behavior manifestation of reading word by word. It indicates deeper processing and careful thinking, and here there is a greater possibility of intentions related to this piece of text.

[0070] Mouse word selection: A trigger for intention speculation, determining the intention speculation related to the selected word content.

[0071] Page scrolling: Repeatedly browsing a certain piece of content, the relative browsing duration per character needs to be accumulated, and the number of browsing times needs to be added.

[0072] Cursor position switching: It may mean the switching of the theme (sub-theme), that is, if the main ideas of the context before and after the cursor position are inconsistent, then consider switching the content. At the same time, the cursor position reflects the repeated browsing and thinking of the context.

[0073] Of course, in addition to the above various user behavior data, in practical applications, other implicit interaction data can also be included, such as the user's eye movement data. Eye movement data can reflect the user's eye focus points and the content the user is concerned about. For example, the eye movement behaviors recorded by an external eye tracking device of an electronic device: fixation, saccade, etc.

[0074] In order to apply the above user behavior data to the relevant analysis and calculation of subsequent intention inference, a data structure that can be recognized by the machine is needed to represent the user behavior data. For example, the behavior data corresponding to each piece of text is arranged in the following order:

[0075] <Browsing duration (seconds), Browsing duration per character (seconds / character), Relative browsing duration per character (multiple), Mouse Hover (times), Mouse word selection (content), Page scrolling (number of browsing times + total duration), Cursor switching times, Eye movement (number of fixations + total duration + number of saccades), Other behaviors (custom fields)>

[0076] Example

[0077] Text: "The user is reading a technical document about the optimization solution of the AI model"

[0078] Behavior data: <60, 0.6, 0.8, 2, "None", "3 times + 20 seconds", 1, "8 times + 40 seconds + 12 times", "None">

[0079] Text: "The user selected 'deep learning' by word selection"

[0080] Behavior data: <10, 0.2, 0.3, 1, "Deep learning", "None", 0, "1 time + 5 seconds + 2 times", "None">

[0081] Text: "The user returned to the previous paragraph to confirm again"

[0082] Behavior data: <40, 0.4, 0.5,, "None", "1 time + 10 seconds", 0, "3 times + 15 seconds + 4 times", "None">

[0083] The above user behavior data can be used for subsequent user intention inference tasks. Since the user behavior data can represent the content that the user is concerned about to a certain extent, it helps to better understand the user intention.

[0084] Figure 3 It is a flowchart for determining the user intention disclosed in the embodiment of the present application. Combined with Figure 3 As shown, determining the user intention based on the user behavior data and the associated content may include:

[0085] Step 301: Perform semantic analysis on the associated content to obtain semantic features.

[0086] The first operation is an active interaction during the user's access to the display content, such as word selection, circle selection, etc.; the associated content at least includes the content selected by the user triggering the first operation, and the associated content may also include the context content of the content selected by the first operation.

[0087] In this embodiment, when performing intention inference, first perform semantic analysis on the associated content to obtain semantic features, that is, first analyze the associated content, combine the relevant context content, and try to understand the user intention.

[0088] Step 302: Perform fusion processing on the semantic features and the user behavior data to obtain fusion features.

[0089] Simply determining user intent based on analysis of associated content has a certain error rate. Therefore, in this application, when inferring user intent, not only the semantic features corresponding to the associated content are used, but also comprehensive analysis is required in combination with user behavior data. Specifically, the semantic features and the user behavior data can be fused to obtain fused features.

[0090] Step 303: perform intent recognition on the fused features and the local knowledge base to determine the user intent.

[0091] After obtaining the fusion features, the local knowledge base can be used as a reference to further clarify the direction of the user's intention. For example, for the related content of "deep learning", if there is a large amount of content about deep learning in the local knowledge base (which can be understood as a personal knowledge base), it can be determined that the user has already understood the basic concepts of deep learning, and the real intention of the user to trigger the first operation is more likely to want to have a deeper understanding of "deep learning"; if there is no relevant content about deep learning in the local knowledge base, it can be determined that the user is first exposed to "deep learning" and needs to understand its basic concepts.

[0092] Figure 4 This is a schematic diagram of the overall implementation architecture of the user intent inference disclosed in the embodiment of the present application. Figure 4 The first operation is an active interaction triggered by the user when accessing the currently displayed content. The core of the active interaction is to actively perceive the user's needs based on the user's current task and provide knowledge enhancement in real time.

[0093] In the implementation, based on the local knowledge base (Knowledge base retrieval), when the user triggers the first operation such as word marking (Trigger), the system will actively match the fusion features of the word marking content, the current paragraph or document context (Content and context) and the user behavior data (UserBehaviors) with the local knowledge base (Knowledge base retrieval); the processing results of the above processing are output to the intent inference model for intent inference (Intent Inferring), so as to understand the user's current word marking intention.

[0094] For example, determined user intent may include: further data comparison, deeper understanding of a concept, tracing back to the original source of a quote, etc.

[0095] Combination Figure 4 As shown, the user intention data obtained by intention inference may contain multiple alternative intentions, so the data processing method may also include: obtaining the user's selection operation (corresponding to Figure 4In the User Feedback module, a "√" is selected (checked), and the selection operation is an operation of selecting at least one alternative intention from the at least one alternative intention; based on the selection operation, the reference data corresponding to the selected alternative intention is obtained and the reference data is output.

[0096] In another implementation, it is possible that none of the multiple alternative intentions are the true intentions of the user, so the user can manually input the true intention (corresponding to Figure 4 the dashed box part in the User Feedback module). Therefore, the method may further include: obtaining the intended intention data input by the user; obtaining the reference data corresponding to the intended intention data and outputting the reference data (corresponding to Figure 4 the Adaptive Knowledge Augmentation module).

[0097] After the user inputs the true intention, the true intention can be fed back to the intention generation model to enrich and correct the next intention understanding (corresponding to Figure 4 the part fed back to the Intent Inferring module in the User Feedback module). Then the data processing method may further include: feeding the intended intention data back to the intention generation model for outputting user intention data to enrich the output types of the intention generation model.

[0098] Among them, outputting the reference data may include: determining the output form of the reference data based on the type of the reference data or the relationship between the reference data and the associated content; outputting the reference data based on the output form.

[0099] For example, comparison questions are presented in a table; concept questions are shown through a knowledge tree; traceability questions directly give the results of a web search, etc. Figure 5 This is an example diagram showing a reference data disclosed in an embodiment of the present application. Figure 6 This is another example diagram showing a reference data disclosed in an embodiment of the present application, which can be combined with Figure 5 and Figure 6 to understand the relevant content.

[0100] Determining the output form of the reference data based on the type of the reference data or the relationship between the reference data and the associated content makes the presentation of the reference data more intuitive and easy to read, which helps the user experience and the speed of processing tasks.

[0101] Since different users have different user behavior habits for accessing display content, in other implementations, it may further include content for personalized adaptation of the data processing solution. Figure 7 This is a flowchart of the personalized adaptation of the data processing solution disclosed in an embodiment of the present application. SeeFigure 7 As shown, the process of personalized adaptation may include:

[0102] Step 701: Construct a context feature vector, which is comprehensive information for expressing the semantic features of user behavior and associated content. The context feature vector includes multiple cumulative terms, including cumulative terms for various types of user behavior, and the cumulative terms for different user behaviors have different weights and / or adjustment coefficients.

[0103] The calculation formula of the context feature vector X can be:

[0104] X = β1*(γ1*T_rel_view)+β2*(γ2*T_rel_input)+β3*S_select+β4*T_view+β5*T_input+β6*H_hover+β7*R_scroll+β8*(γ3*C_cursor)+β9*E_eye+Semantic_Context+ε

[0105] Where:

[0106] X: The context feature vector, which is used to express the comprehensive information of user behavior and semantics. Semantic_Context: The semantic context feature, which is the semantic embedding extracted based on the user's current interaction content.

[0107] ε: The error term, which represents the uncaught randomness or model noise.

[0108] β and γ: The weights and adjustment coefficients, which are dynamically optimized according to online learning.

[0109] Among them, the constraint conditions are:

[0110] β3 > β1, β2 > β4, β5: The highest for mouse word highlighting, followed by the relative browsing and input duration of unit characters.

[0111] β8 > β6, β7: The cursor position switching and page scrolling are more important.

[0112] γ1, γ2, γ3 > 1: High-weight processing of the context differences in relative browsing duration, relative input duration, and cursor switching.

[0113] Step 702: Train and optimize the weights and / or adjustment parameters of the cumulative terms of various user behaviors in the context feature vector based on the user's feedback data.

[0114] Through multiple user feedbacks, using the online reinforcement learning method (contextual Bandit), train the parameters in the formula under the above constraint conditions to make the efficiency of joint intention understanding of behavior and semantics optimal.

[0115] The design is as follows:

[0116] Since the user interacts with the system in real time, the Contextual Bandit method is selected for training:

[0117] 1. Contextual features: `X` includes user behavior and semantic features.

[0118] 2. Intentional actions: `A = {A_1, A_2,..., A_k}`, representing all possible intent categories.

[0119] 3. Reward function:

[0120]

[0121] 4. Goal: Select the intent category that maximizes the reward:

[0122] A* = argmaxA P(A|X)

[0123] The intents involved can be but are not limited to the following:

[0124] Basic intents: Information browsing, content location, definition and explanation;

[0125] Content understanding: In-depth reading, context association, contradiction or conflict recognition;

[0126] Information processing: Information filtering, comparison and contrast, summary and refinement;

[0127] Information expansion: Background information query, knowledge graph construction, data verification;

[0128] Task-oriented: Problem solving, task planning, document proofreading;

[0129] Advanced reasoning: Inference and hypothesis, cross-document association, content reorganization;

[0130] In implementation, intent types can be dynamically added according to the user's Query (which can be understood as the real intent feedback by the user).

[0131] In the specific implementation, the following can be carried out in sequence:

[0132] 1. Define the problem (as in the previous part):

[0133] - Contextual features: User behavior features (quantitative data such as unit character browsing duration, mouse word selection, page scrolling, etc.) and semantic features (paragraph semantic embedding).

[0134] - Action space: Intent categories, such as query, operation, task completion.

[0135] - Rewards: Explicit and implicit feedback after user interaction to measure the correctness of intent prediction.

[0136] 2. Model framework:

[0137] -Use the Contextual Bandit framework, including context features, action space, policy function, and reward function.

[0138] -The strategy function can use Epsilon-Greedy, UCB or Softmax strategy.

[0139] 3. Algorithm implementation steps:

[0140] -Step 1: Context feature extraction:

[0141] -User behavior characteristics: relative browsing time per character, mouse word selection, cursor switching, etc.

[0142] -Semantic features: Calculate the semantic similarity between the current content and the marked word content.

[0143] -Build context feature vector.

[0144] -Step 2: Action selection strategy:

[0145] -Initialize the policy function and select the best action under the current context.

[0146] -Action-value function: Estimate Q(A|X) based on context features X and parameters θ.

[0147] -Step 3: Reward function definition:

[0148] - Explicit Reward: Reward R=1 when the user confirms the intention.

[0149] -Implicit rewards: user's subsequent behavior and recommendation intention Figure 1 The reward at that time is R=0.8.

[0150] Step 1-Step 3 is the process of continuing the model connection.

[0151] -Step 4: Online Update:

[0152] -Collect user feedback and update the action value function:

[0153]

[0154] -Adjust the strategy function to dynamically optimize prediction accuracy.

[0155] Step 4 is the process of optimizing X through A (intention).

[0156] - Step 5: Model Evaluation:

[0157] - Speculation Accuracy: The consistency between the predicted intention and the actual intention.

[0158] - Cumulative Reward Value: Records the cumulative reward of the model.

[0159] - User Satisfaction: Based on implicit interaction and explicit scoring.

[0160] 4. Real - time Optimization Mechanism:

[0161] - Context Dynamic Adjustment: Monitor new behavior patterns and expand the feature space.

[0162] - Exploration - Exploitation Balance: Dynamically adjust the Epsilon value (balance value) to enhance the exploration efficiency.

[0163] - Cross - user Transfer Learning: Aggregate multi - user behavior patterns to improve the intention speculation effect for cold - start users.

[0164] Through the above content, the model equipped with the data processing method can better adapt to the user behaviors of different users, providing a more comfortable and accurate intention inference for users.

[0165] For the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0166] In the above - disclosed embodiments of the present application, the method is described in detail. The method of the present application can be implemented by various forms of devices. Therefore, the present application also discloses a device, and specific embodiments are given below for detailed description.

[0167] Figure 8 It is a schematic structural diagram of a data processing device disclosed in an embodiment of the present application. Refer to Figure 8 As shown, the data processing device 80 may include:

[0168] An operation acquisition module 801, configured to acquire a first operation triggered by a user for the currently displayed content, where the first operation has associated content.

[0169] A behavior acquisition module 802, configured to obtain user behavior data for a set period based on the first operation, where the user behavior data is implicit interaction data collected based on control operations of an input device during the process of the user accessing the currently displayed content, and the implicit interaction data can characterize the content that the user is concerned about.

[0170] An intention determination module 803, configured to determine user intention data based on the user behavior data and the associated content.

[0171] In this embodiment, when the user triggers a first operation on the currently displayed content, the data processing device will obtain user behavior data for a set period, and comprehensively determine the user intention by combining the user behavior data and the associated content of the first operation; since the user behavior data is implicit interaction data that can characterize the content that the user is concerned about, using the user behavior data as the basis for determining the user intention can further improve the accuracy of intention inference.

[0172] In one implementation, the first operation is a content selection operation, and the content selection operation includes any one of the following: a box selection operation, a circle selection operation, a swipe selection operation, a line drawing operation; the associated content at least includes the content selected by the content selection operation.

[0173] In one implementation, the behavior acquisition module may specifically be configured to: based on the first operation, obtain user behavior data for a set period before the user triggers the first operation, where the user behavior data includes at least one of the following: access status data of the current content, cursor movement data, page scrolling data, and the user behavior data has a fixed data structure.

[0174] In one implementation, the intention determination module may specifically be configured to: perform semantic analysis on the associated content to obtain semantic features; perform fusion processing on the semantic features and the user behavior data to obtain fusion features; perform intention recognition on the fusion features and a local knowledge base to determine the user intention.

[0175] In one implementation, the user intention data includes at least one alternative intention, and the device may further include: a feedback processing module, configured to obtain a user selection operation, where the selection operation is an operation of selecting at least one alternative intention from the at least one alternative intention; obtain reference data corresponding to the selected alternative intention based on the selection operation and output the reference data.

[0176] In one implementation, the device may further include: a feedback processing module, configured to obtain intended intention data input by the user; obtain reference data corresponding to the intended intention data and output the reference data.

[0177] In one implementation, the feedback processing module may specifically be configured to: determine the output form of the reference data based on the type of the reference data or the relationship between the reference data and the associated content; output the reference data based on the output form.

[0178] In one implementation, the feedback processing module may specifically be configured to: feedback the intended intention data to an intention generation model for outputting user intention data, so as to enrich the output type of the intention generation model.

[0179] In one implementation, the apparatus may further include: an adaptation processing module, configured to: construct a context feature vector, where the context feature vector is comprehensive information for expressing semantic features of user behaviors and associated content, the context feature vector includes a plurality of accumulation terms, including accumulation terms of various user behaviors, and the accumulation terms of different user behaviors have different weights and / or adjustment coefficients; train and optimize the weights and / or adjustment parameters of the accumulation terms of various user behaviors in the context feature vector based on the feedback data of the user.

[0180] For the specific implementation of the above data processing apparatus and each module included therein, as well as other possible implementations, reference may be made to the content introduction in the corresponding part of the method embodiment, which will not be repeated here.

[0181] Any one of the data processing apparatuses in the above embodiments includes a processor and a memory. The operation acquisition module, behavior acquisition module, intention determination module, adaptation processing module, feedback processing module, etc. in the above embodiments are all stored in the memory as program modules, and the corresponding functions are implemented by the processor executing the above program modules stored in the memory.

[0182] The processor includes a kernel, and the kernel retrieves the corresponding program module from the memory. One or more kernels may be provided, and the processing of the return visit data is achieved by adjusting the kernel parameters.

[0183] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0184] In an exemplary embodiment, there is also provided a computer-readable storage medium that can be directly loaded into the internal memory of a computer, which contains software code, and the computer program can implement the steps shown in any embodiment of the above data processing method after being loaded and executed by the computer.

[0185] In an exemplary embodiment, a computer program product is also provided, which can be directly loaded into the internal memory of a computer and contains software code. After being loaded and executed by the computer, the computer program can implement the steps shown in any of the above-described data processing methods.

[0186] Furthermore, an embodiment of the present application provides an electronic device. Figure 9 It is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application. Refer to Figure 9 As shown, the electronic device 90 includes at least one processor 901, at least one memory 902 connected to the processor, and a bus 903; wherein, the processor and the memory complete communication with each other through the bus; the processor is used to call program instructions in the memory to execute the above data processing method.

[0187] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0188] It should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or device including the said element.

[0189] The steps of the methods or algorithms described in connection with the embodiments disclosed herein can be implemented directly in hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0190] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data processing method, the method comprising: Obtaining a first operation triggered by a user for current displayed content, the first operation having associated content; Obtaining user behavior data for a set time period based on the first operation, the user behavior data being implicit interaction data collected based on control operations of an input device during the process of the user accessing the current displayed content, and the implicit interaction data being capable of characterizing the content that the user is concerned about; Determining user intention data based on the user behavior data and the associated content.

2. The data processing method according to claim 1, wherein The first operation is a content selection operation, and the content selection operation includes any one of the following: a box selection operation, a circle selection operation, a swipe selection operation, a line drawing operation; The associated content at least includes the content selected by the content selection operation.

3. The data processing method according to claim 1, the obtaining user behavior data for a set time period based on the first operation includes: Based on the first operation, obtaining user behavior data for a set time period before the user triggers the first operation, the user behavior data including at least one of the following: access status data of the current content, cursor movement data, page scrolling data, and the user behavior data having a fixed data structure.

4. The data processing method according to claim 1, the determining the user intention based on the user behavior data and the associated content includes: Performing semantic analysis on the associated content to obtain semantic features; Performing fusion processing on the semantic features and the user behavior data to obtain fusion features; Performing intention recognition on the fusion features and a local knowledge base to determine the user intention.

5. The data processing method according to claim 1, the user intention data includes at least one alternative intention, and the method further includes: Obtaining a selection operation of the user, the selection operation being an operation of selecting at least one alternative intention from the at least one alternative intention; Obtaining reference data corresponding to the selected alternative intention based on the selection operation and outputting the reference data.

6. The data processing method according to claim 1, the method further includes: Obtaining intended intention data input by the user; Obtaining reference data corresponding to the intended intention data and outputting the reference data.

7. The data processing method according to claim 5 or 6, wherein outputting the reference data includes: Determining an output form of the reference data based on the type of the reference data or the relationship between the reference data and the associated content; Outputting the reference data based on the output form.

8. The data processing method according to claim 6, further includes: Feeding back the intended intention data to an intention generation model for outputting user intention data to enrich the output types of the intention generation model.

9. The data processing method according to claim 1, further includes: Construct a context feature vector, which is comprehensive information for expressing the semantic features of user behavior and associated content. The context feature vector includes multiple summation terms, including summation terms for various types of user behavior, and the summation terms for different user behaviors have different weights and / or adjustment coefficients; Train and optimize the weights and / or adjustment parameters of the summation terms for various types of user behavior in the context feature vector based on the user's feedback data.

10. A data processing device, the device comprising: An operation acquisition module, configured to acquire a first operation triggered by a user for the current displayed content, where the first operation has associated content; A behavior acquisition module, configured to acquire user behavior data for a set period based on the first operation, where the user behavior data is implicit interaction data collected based on control operations of an input device during the user's access to the current displayed content, and the implicit interaction data can characterize the user's attention content; An intention determination module, configured to determine user intention data based on the user behavior data and the associated content.

Images