Information processing method and device

Abstract

The present invention discloses an information processing method and device, including obtaining first information, the first information including a first event and demand information for processing the first event; determining at least the intent information, slot information and feedback information corresponding to the first event according to the first information; configuring corresponding annotation information for the intent information, slot information and feedback information respectively, the annotation information is used to characterize the information type; arranging at least the configured intent information, slot information and feedback information according to a set rule to obtain second information; and sending the determined second information to a target object. By extracting intent information, slot information and feedback information from the initial information, configuring corresponding annotation information for the intent information, slot information and feedback information, and arranging the configured information into the second information according to the set rule, the developer can understand the content that the business party wants to express through the annotation information and the second information arranged according to the rule, so that both parties can communicate efficiently even in different industry fields.

Description

Technical Field

[0001] The present invention relates to the field of intelligent customer service technology, and in particular to an information processing method and device. Background Art

[0002] In the process of establishing an intelligent customer service dialogue system, system developers need to continuously analyze needs with business users and define conversation processes. Currently, communication is done in the form of text descriptions or flowcharts, but the language habits and drawing methods used by users in each industry are different, resulting in high communication costs and relatively low efficiency. Summary of the invention

[0003] One aspect of the present invention provides an information processing method, the method comprising:

[0004] Acquire first information, where the first information includes a first event and requirement information for processing the first event;

[0005] Determine at least the intent information, slot information, and feedback information corresponding to the first event according to the first information;

[0006] Configuring corresponding annotation information for the intent information, slot information, and feedback information respectively, wherein the annotation information is used to characterize the information type;

[0007] Arrange at least the configured intention information, slot information, and feedback information according to a set rule to obtain second information;

[0008] The determined second information is sent to the target object.

[0009] In one possible implementation, respectively configuring corresponding annotation information for the intent information, slot information, and feedback information includes:

[0010] Preset multiple different types of annotation information, the multiple different types at least including intent type and slot type;

[0011] The annotation information is mapped to the intent information and slot information according to the type.

[0012] In one possible implementation, after determining at least the intent information, slot information, and feedback information corresponding to the first event according to the first information, the method further includes:

[0013] Determining whether the slot information satisfies necessary information set for the intent information;

[0014] If the judgment is negative, prompt information representing the supplementary information is fed back based on the slot information.

[0015] In one possible implementation manner, the determining whether the slot information satisfies necessary information set for the intent information includes:

[0016] At least inputting the intention information and slot information into the dialogue engine for a trial operation;

[0017] According to whether the trial operation result meets expectations, it is determined whether the slot information meets the necessary information set for the intention information.

[0018] In one embodiment, the method further comprises:

[0019] Determine, based on the first information, operation information representing a clear intention corresponding to the first event;

[0020] Accordingly, corresponding annotation information is configured for the operation information;

[0021] The configured intention information, slot information, operation information and feedback information are arranged according to set rules to obtain second information.

[0022] In one possible implementation, the second information is a code block, and the code block includes an intention module, a requirement module, and a feedback module;

[0023] Wherein, the intention module includes at least one set of the demand module and the feedback module, which is used to map the intention information;

[0024] The requirement module is used to map the slot information and the operation information;

[0025] The feedback module is used to map the feedback information.

[0026] In one possible implementation manner, the second information is a structure flow chart, and the structure flow chart is obtained by converting the code block, specifically:

[0027] Determining a first process node according to the intent module of the code block;

[0028] Determining a second process node according to a requirement module of the code block;

[0029] Determining a third process node according to the feedback module of the code block;

[0030] If the same intention information contains multiple groups of demand modules and feedback modules, then determine the intermediate process node representing the judgment type;

[0031] The first process node, the intermediate process node, the second process node and the third process node are connected to form the structural flow chart.

[0032] In one embodiment, the method further comprises:

[0033] Based on the second information, third information executable by the dialog engine is generated.

[0034] In one possible implementation manner, the step of arranging the configured intention information, slot information, and feedback information according to a set rule to obtain the second information includes:

[0035] According to the logical operation relationship between the intention information, slot information and feedback information, corresponding logical symbols are configured and corresponding sorting processing is performed to generate the second information.

[0036] Another aspect of the present invention provides an information processing device, the device comprising:

[0037] An information acquisition module, configured to acquire first information, wherein the first information includes a first event and requirement information for processing the first event;

[0038] an information determination module, configured to determine at least intent information, slot information, and feedback information corresponding to the first event based on the first information;

[0039] An information configuration module, used to configure corresponding annotation information for the intent information, slot information and feedback information respectively, wherein the annotation information is used to characterize the information type;

[0040] An information arrangement module, used to arrange at least the configured intention information, slot information and feedback information according to a set rule to obtain second information;

[0041] The information sending module is used to send the determined second information to the target object.

[0042] In an embodiment of the present invention, by extracting intent information, slot information and feedback information from originally unprocessed information, and configuring corresponding annotation information for the intent information, slot information and feedback information, the configured intent information, slot information and feedback information are arranged into second information according to set rules. The developer can understand what the business party wants to express through the annotation information and the second information after rule sorting, so that the business party and the developer can communicate efficiently even in different industry fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0044] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0045] Figure 1A schematic diagram of an implementation flow of an information processing method according to an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of a method for processing information in accordance with an embodiment of the present invention, in which the second information is a code block;

[0047] Figure 3 A schematic diagram of a structural flow chart in which the second information in an information processing method according to an embodiment of the present invention is a flowchart;

[0048] Figure 4 The figure is a schematic diagram of the structure of an information processing device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0050] like Figure 1 As shown, the present invention provides an information processing method on one hand, the method comprising:

[0051] Step 101, obtaining first information, where the first information includes a first event and requirement information for processing the first event;

[0052] Step 102, determining at least the intent information, slot information, and feedback information corresponding to the first event according to the first information;

[0053] Step 103, respectively configuring corresponding annotation information for the intent information, slot information, and feedback information, where the annotation information is used to characterize the information type;

[0054] Step 104, at least arranging the configured intention information, slot information and feedback information according to a set rule to obtain second information;

[0055] Step 105: Send the determined second information to the target object.

[0056] In this embodiment, in step 101, the first information is generally the content of the interactive dialogue between the user and the intelligent customer service. The interactive dialogue content can be manually written based on a simulated interactive scenario, or can be directly extracted from the historical dialogue log between the intelligent customer service and the user.

[0057] The first event is represented by the interaction type or interaction theme corresponding to the first information. Assuming that the first information is "Purchase a children's ticket to Beijing, name xxx, ID number xxx, user confirms payment, and the feedback result is that the purchase of the children's ticket is successful", then the interaction theme of the first information is buying and selling tickets. The determination method of the first event can be obtained by identifying the keywords in the first information. For example, if the keywords related to tickets such as "children's ticket" or "ticket" are identified in the first information, the interaction theme can be determined to be buying and selling tickets. Of course, the first event can also be directly reflected in the first information, such as "buying and selling ticket types, purchase a children's ticket to Beijing, name xxx, ID number xxx, user confirms payment, and the feedback result is that the purchase of the children's ticket is successful". The determination method of this first event can also be obtained by identifying the keywords in the first information.

[0058] The demand information for processing the first event includes the information that the user needs to provide for the first event and the corresponding feedback results. For example, if the first event is the ticket purchase and sale event in the above example, the demand information may include name, personal ID number, age, destination, etc., and the feedback results include successful purchase, failed purchase, etc.

[0059] In step 102, after obtaining the first information, at least the intent information, slot information and feedback information of the first event need to be determined, wherein the intent information is the purpose of the user's needs. For example, for the first information about ticket sales in the above example, the intent information is to buy tickets. The intent information can be determined by inputting the first information into an existing intent recognition model (such as a rule-based model, a traditional machine learning algorithm (SVM), a deep learning algorithm (CNN, LSTM, RCNN, C-LSTM, FastText) model, etc.) for intent recognition, and outputting the intent information; the slot information is the keyword information required to achieve the intent information. For example, if the intent information is to buy tickets, the slot information includes name, personal ID number, age, destination, etc. The slot information can be extracted in a manner such as: The first information is subjected to named entity recognition, label recognition, or obtained based on user portrait data, wherein named entity recognition refers to identifying entities with specific meanings in the text, such as names of people, names of organizations, names of places, etc., and can be specifically based on a deep learning model based on LSTM+CRF or a deep learning model based on Dilated-CNN. When applied specifically, the slot information in the first information can be extracted by performing named entity recognition on the first information using the above-mentioned deep learning model; label recognition requires setting corresponding labels for each word in the first information in advance. When applied, the slot information can be extracted by identifying the set labels; user portrait data is a virtual representation of real users, and is a target user model based on a series of real data. When applied, the slot information for the first event can be extracted from the user portrait data set in advance.

[0060] The feedback information is specifically the processing result made for the identified intent information and slot information, such as the successful purchase in the ticket purchase event mentioned above.

[0061] In step 103, after the intent information, slot information and feedback information are obtained, the intent information, slot information and feedback information are respectively configured with annotation information corresponding to the information type. The specific configuration method may be to configure the corresponding annotation information according to the information type determined in step 102. For example, taking the above-mentioned ticket buying and selling event as an example, if the intent information is identified as "want to buy tickets", the corresponding type of annotation information such as "INTENT" is configured for "want to buy tickets", and the result after configuration is "INTENT (want to buy tickets)". The configuration method of slot information and feedback information is the same as that of intent information. The final configuration result of slot information may be "SLOT (name xxx)", "SLOT (ID number xxx)", etc., and the final configuration result of feedback information may be RESULT (purchase successful).

[0062] In step 104, at least the configured intent information, slot information, and feedback information are arranged according to the set rules to obtain the second information; wherein the set rules can be specifically arranged according to the logical operation relationship between the configured intent information, slot information, and feedback information. The arrangement of the logical operation relationship can refer to Figure 2 As shown, Figure 2 The structured information shown in the figure is composed of the intent information INTENT (want_to_buy_ticket), slot information SLOT (user_name), and feedback information RESULT (child_ticket_success). Figure 2 In the code block, the intent information is located in the header, and multiple slot information is located in the header code block and is combined into a code block EXPECT, which represents the keyword information required under the same intent information. It should be noted that multiple EXPECT code blocks can exist under the same intent information at the same time, and each EXPECT code block contains different slot information. The feedback information represents the processing result information under the same intent information and slot information. Figure 2 The structured information shown can clearly indicate that the first information includes an intent information, under which one or more groups of slot information are stored, and feedback information for each group of slot information.

[0063] In step 105, the generated second information is sent to the target object, wherein the target object may be a terminal device or a server, and the system developer may view the second information through the terminal device or the server.

[0064] Therefore, by extracting the intent information, slot information and feedback information from the original unprocessed information, and configuring corresponding annotation information for the intent information, slot information and feedback information, the configured intent information, slot information and feedback information are arranged into the second information according to the set rules. The developer can understand what the business party wants to express through the annotation information and the second information after the rule sorting, so that the business party and the developer can communicate efficiently even in different industry fields.

[0065] In one possible implementation, corresponding annotation information is configured for the intent information, slot information, and feedback information, respectively, including:

[0066] Preset multiple different types of annotation information, where the multiple different types at least include intent type and slot type;

[0067] Map annotation information to intent information and slot information according to type.

[0068] In this embodiment, the specific process of step 102 is:

[0069] Preset multiple different types of annotation information, where the preset annotation information types include intent types (refer to Figure 2 INTENT in), slot type (refer to Figure 2 SLOT in), feedback type (reference Figure 2 RESULT in , etc.

[0070] After determining the intent information, slot information, and feedback information in the first information, the intent information is matched with the annotation information of the intent type and mapped to each other (refer to Figure 2 INTENT(want_to_buy_ticket) in the slot information, matching the slot information with the slot type annotation information and mapping them to each other, slot information (reference Figure 2 SLOT(user_name) in the feedback information), matching the feedback information with the annotation information of the feedback type and mapping them to each other (refer to Figure 2 RESULT(child_ticket_success) in .

[0071] In one possible implementation, after determining at least the intent information, slot information, and feedback information corresponding to the first event according to the first information, the method further includes:

[0072] Determine whether the slot information meets the necessary information set for the intent information;

[0073] If the judgment is negative, prompt information representing the supplementary information is fed back based on the slot information.

[0074] In this embodiment, for each first event type, corresponding slot information can be set in advance. For example, for ticket buying and selling events, the required slot information can be set to name, age, ID number, destination, travel time, etc.

[0075] After step 102, it is determined whether the slot information satisfies the necessary information set for the intent information, wherein the necessary information is the pre-set slot information mentioned above. The determination of whether it is satisfied can be determined whether the determined slot information contains the set necessary information, or whether the difference between the determined slot information quantity and the necessary information quantity is within a preset threshold.

[0076] If it is determined that the slot information does not meet the necessary information set for the intent information, it means that the amount of information provided by the business direction developer is insufficient, then prompt information representing the supplementary information is fed back based on the slot information. For example, for the ticket purchase and sale event, if the name is missing in the slot information determined in step 102, feedback can be given based on the missing slot information to prompt the need to supplement the slot information, such as "You are missing key name information, please supplement it."

[0077] In one possible implementation, determining whether the slot information satisfies necessary information set for the intent information includes:

[0078] At least the intent information and the slot information are input into the dialogue engine for a trial operation;

[0079] Depending on whether the trial run results meet expectations, determine whether the slot information meets the necessary information set for the intent information.

[0080] In this embodiment, the specific process of determining whether the slot information meets the requirements is as follows:

[0081] The currently determined intent information and slot information are input into the existing dialogue engine for trial operation. If the dialogue engine can feedback accurate answers based on the current intent information and slot information, it means that the currently determined slot information is sufficient to meet the query conditions, and the slot information is determined to meet the necessary information set for the intent information. On the contrary, if the dialogue engine feedbacks multiple answers or cannot feedback an answer based on the current intent information and slot information, it means that the currently determined slot information is insufficient to meet the query conditions, and the slot information is determined to not meet the necessary information set for the intent information.

[0082] In one embodiment, the method further comprises:

[0083] Determining, based on the first information, operation information representing a clear intention corresponding to the first event;

[0084] Accordingly, corresponding annotation information is configured for the operation information;

[0085] The configured intention information, slot information, operation information and feedback information are arranged according to set rules to obtain second information.

[0086] In this embodiment, in addition to the above-mentioned determination of intent information, slot information and feedback information, operation information representing a clear intent is also extracted from the first information. The operation information may represent interactive actions such as user clicking buttons, icons, links, etc. The first information "Purchase a child ticket to Beijing, name xxx, ID number xxx, user confirms payment, feedback result is successful purchase of child ticket" is still taken as an example, where the intent information of the first information is to purchase a ticket, and the corresponding operation information is "user confirms payment" in the first information. The extraction method of operation information can also be label recognition or user portrait data recognition. These two recognition methods are consistent with the method of extracting slot information, and will not be repeated here.

[0087] After extracting the operation information, also configure the corresponding annotation information for the operation information. The annotation information refers to Figure 2 The configured operation information can be expressed as ACTION(btn_confirm_pay). The configuration method of operation information is the same as that of slot information and intent information, and will not be repeated here.

[0088] When generating the second information, operation information is added on the basis of the original intention information, slot information and feedback information, and the above four information are arranged according to the set rules to obtain the second information.

[0089] In one possible implementation, the second information is a code block, and the code block includes an intention module, a requirement module, and a feedback module;

[0090] Wherein, the intention module includes at least one set of demand modules and feedback modules for mapping intention information;

[0091] The requirement module is used to map slot information and operation information;

[0092] The feedback module is used to map feedback information.

[0093] In this embodiment, the second information may be presented in the form of a code block, specifically, a SCFL (Structure Chatbot Flow Language) language code, see Figure 2As shown, the intent module is specifically a code block formed by INTENT(xxx){}, which is used to map intent information, the requirement module is specifically a code block formed by EXPECT{}, which is used to map slot information and operation information, and the feedback module is specifically a code block formed by RESULT(), which is used to map feedback information. The requirement module and the feedback module are both located in the intent module to represent the requirement information and processing result information required under the intent.

[0094] Furthermore, the requirement module may include operation information in addition to the mapping slot information and feedback information.

[0095] Furthermore, the same intent information can include multiple sets of demand modules and feedback modules at the same time. Figure 2 It shows that the same intention information (buy tickets) includes two kinds of needs. The first one is the demand information and corresponding feedback information when the age is less than 12 years old, and the second one is the demand information and corresponding feedback information when the age is greater than 12 years old.

[0096] In one possible implementation, the second information is a structure flow chart, which is obtained by converting the code block, specifically:

[0097] Determine the first process node according to the intent module of the code block;

[0098] Determine the second process node according to the requirement module of the code block;

[0099] Determine the third process node according to the feedback module of the code block;

[0100] If the same intent information contains multiple sets of demand modules and feedback modules, determine the intermediate process node that represents the judgment type;

[0101] The first process node, the middle process node, the second process node and the third process node are connected to form a structural flow chart.

[0102] In this embodiment, the structural flow chart can refer to Figure 3 As shown, the confirmation method of the process node can be to determine whether there is an intention module, a requirement module or a feedback module in the code block. If so, the corresponding process node is obtained. Each different process node corresponds to a different style, such as Figure 3 In the figure, the first process node is represented by a rectangle, the second process node is represented by a rounded rectangle, and the third process node is represented by a trapezoid. It should be noted that the style of the process node is not limited to Figure 3 shown.

[0103] In addition to the above-mentioned first process node, second process node, third process node and intermediate process node, it further includes a confirmation process node, which corresponds to the above-mentioned operation information, and may also include multiple identical process nodes if necessary, such as Figure 3 Two second process nodes are included.

[0104] There are many ways to connect the above process nodes. Figure 3 For example, the connection method can be through arrows from top to bottom, the first process node, the second process node, the intermediate process node, the second process node, the confirmation process node and the third process node, Figure 3 The process logic is: first determine the user's intention information (want_to_buy_ticket), determine the necessary slot information (user_name, user_age) based on the intention information, make multiple judgments based on the slot information with non-fixed attributes (user_age<=12, user_age>12), and execute the third process node in each case when the user confirms the intention. It should be noted that the connection method of the process nodes is not limited to Figure 3 The connection method of each process node can be interchanged.

[0105] In one embodiment, the method further comprises:

[0106] Based on the second information, third information executable by the dialog engine is generated.

[0107] In this embodiment, the third information is specifically information that can be directly used by the dialogue engine, such as slot information, intent information, etc. The third information can be extracted through code blocks or through structural flowcharts, wherein the extraction of code blocks can be to parse the annotation information in the code blocks, such as INTENT, SLOT, etc., and then obtain the corresponding information based on the annotation information; the extraction of structural flowcharts can be to determine the pattern of each node in the flowchart or the pre-set annotation, and obtain the corresponding information based on the determination result. Through this step, data that can be directly run by the dialogue engine can be quickly obtained to quickly generate a prototype system.

[0108] In one possible implementation mode, corresponding annotation information is configured for the intent information, the slot information, and the feedback information, respectively, and the annotation information is used to characterize the information type; the configured intent information, the slot information, and the feedback information are arranged according to a set rule to obtain the second information, including:

[0109] According to the logical operation relationship between the intention information, the slot information and the feedback information, the corresponding logical symbols are configured and the corresponding sorting processing is performed to generate the second information.

[0110] In this embodiment, the specific process of generating the second information is as follows:

[0111] Taking the finally generated second information as a code block as an example, it is generally understood that the logical running order of intent information, slot information and feedback information is intent information, slot information and feedback information, to indicate that a corresponding group or multiple groups of slot information and feedback information are included under a total intent information.

[0112] refer to Figure 2 As shown, during configuration, the intent information is configured first, specifically, a logical symbol indicating that it is an intent is added to the left side of the intent information, such as INTENT{}. According to the logical operation order, one or more groups of slot information and feedback information are then configured. For each group of slot information, a logical symbol indicating one of the required information under the same intent information is added, such as EXPECT[]. A logical symbol indicating that it is slot information is also added to the left side of each slot information, such as SLOT. Furthermore, for slot information with non-fixed attribute values, such as user_age, one of the possibilities of slot information with non-fixed attribute values ​​needs to be written in the SLOT{} code block, such as Figure 2 In the code block, user_age<=12 is used to indicate that the current demand information is the information when user_age<=12. Furthermore, operation information can be added to the demand information code block to indicate the intention to confirm under the current slot information. Similarly, a logical symbol indicating that it is operation information is added to the left side of the operation information, such as ACTTION(). The feedback information is located in the intention information code block and below the demand information to indicate the processing result given for the above demand information under the same intention information. Similarly, a logical symbol indicating that it is feedback information is added to the left side, such as RESULT().

[0113] If there are multiple sets of slot information and feedback information, they can be placed below the previous requirement information code block to represent another type of requirement information. The logic symbol generation method for each piece of information is the same as above. By arranging multiple sets of slot information and feedback information in an up-and-down manner, multiple related branch processes can be split into a single independent process data definition to avoid mutual interference between branch logic judgments, which is convenient for viewing and demonstration.

[0114] like Figure 4 As shown, another aspect of the present invention provides an information processing device, the device comprising:

[0115] The information acquisition module 201 is used to acquire first information, where the first information includes a first event and requirement information for processing the first event;

[0116] An information determination module 202, configured to determine at least intent information, slot information, and feedback information corresponding to the first event according to the first information;

[0117] An information configuration module 203 is used to configure corresponding annotation information for the intent information, slot information and feedback information respectively, where the annotation information is used to characterize the information type;

[0118] An information arrangement module 204, configured to arrange at least the configured intention information, slot information and feedback information according to a set rule to obtain second information;

[0119] The information sending module 205 is used to send the determined second information to the target object.

[0120] In this embodiment, in the information acquisition module 201, the first information is generally the interactive dialogue content between the user and the intelligent customer service. The interactive dialogue content can be manually written based on a simulated interactive scenario, or it can be directly extracted from the historical dialogue log between the intelligent customer service and the user.

[0121] The first event is represented by the interaction type or interaction theme corresponding to the first information. Assuming that the first information is "Purchase a children's ticket to Beijing, name xxx, ID number xxx, user confirms payment, and the feedback result is that the purchase of the children's ticket is successful", then the interaction theme of the first information is buying and selling tickets. The determination method of the first event can be obtained by identifying the keywords in the first information. For example, if the keywords related to tickets such as "children's ticket" or "ticket" are identified in the first information, the interaction theme can be determined to be buying and selling tickets. Of course, the first event can also be directly reflected in the first information, such as "buying and selling ticket types, purchase a children's ticket to Beijing, name xxx, ID number xxx, user confirms payment, and the feedback result is that the purchase of the children's ticket is successful". The determination method of this first event can also be obtained by identifying the keywords in the first information.

[0122] The demand information for processing the first event includes the information that the user needs to provide for the first event and the corresponding feedback results. For example, if the first event is the ticket purchase and sale event in the above example, the demand information may include name, personal ID number, age, destination, etc., and the feedback results include successful purchase, failed purchase, etc.

[0123] In the information determination module 202, after obtaining the first information, at least the intent information, slot information and feedback information of the first event need to be determined, wherein the intent information is the purpose of the user's needs. For example, for the first information about ticket sales in the above example, the intent information is to buy tickets. The intent information can be determined by inputting the first information into an existing intent recognition model (such as a rule-based model, a traditional machine learning algorithm (SVM), a deep learning algorithm (CNN, LSTM, RCNN, C-LSTM, FastText) model, etc.) for intent recognition, and outputting the intent information; the slot information is the keyword information required to achieve the intent information. For example, if the intent information is to buy tickets, the slot information includes name, personal ID number, age, destination, etc. The slot information is extracted in a manner The first information can be subjected to named entity recognition, label recognition, or obtained based on user portrait data, wherein named entity recognition refers to the recognition of entities with specific meanings in the text, such as names of people, names of organizations, names of places, etc., and can specifically be a deep learning model based on LSTM+CRF or a deep learning model based on Dilated-CNN. When applied specifically, the slot information in the first information can be extracted by performing named entity recognition on the first information using the above-mentioned deep learning model; label recognition requires setting corresponding labels for each word in the first information in advance. When applied, the slot information can be extracted by identifying the set labels; user portrait data is a virtual representation of real users, and is a target user model based on a series of real data. When applied, the slot information for the first event can be extracted from the user portrait data set in advance.

[0124] The feedback information is specifically the processing result made for the identified intent information and slot information, such as the successful purchase in the ticket purchase event mentioned above.

[0125] In the information configuration module 203, after the intent information, slot information and feedback information are obtained, the intent information, slot information and feedback information are respectively configured with the label information corresponding to the information type. Among them, the specific configuration method can be to configure the corresponding label information according to the information type determined in the information determination module 202. For example, taking the above-mentioned ticket buying and selling event as an example, if the intent information is identified as "want to buy tickets", then the corresponding type of label information such as "INTENT" is configured for "want to buy tickets", and the result after configuration is "INTENT (want to buy tickets)". The configuration method of slot information and feedback information is the same as that of configuring intent information. The final configuration result of slot information can be "SLOT (name xxx)", "SLOT (ID number xxx)", etc., and the final configuration result of feedback information can be RESULT (purchase successful).

[0126] In the information arrangement module 204, at least the configured intent information, slot information and feedback information are arranged according to the set rules to obtain the second information; wherein the set rules can be specifically arranged according to the logical operation relationship between the configured intent information, slot information and feedback information, and the arrangement of the logical operation relationship can refer to Figure 2 As shown, Figure 2 The structured information shown in the figure is composed of the intent information INTENT (want_to_buy_ticket), slot information SLOT (user_name), and feedback information RESULT (child_ticket_success). Figure 2 In the code block, the intent information is located in the header, and multiple slot information is located in the header code block and is combined into a code block EXPECT, which represents the keyword information required under the same intent information. It should be noted that multiple EXPECT code blocks can exist under the same intent information at the same time, and each EXPECT code block contains different slot information. The feedback information represents the processing result information under the same intent information and slot information. Figure 2 The structured information shown can clearly indicate that the first information includes an intent information, under which one or more groups of slot information are stored, and feedback information for each group of slot information.

[0127] In the information sending module 205, the generated second information is sent to the target object, wherein the target object may be a terminal device or a server, and the system developer may view the second information through the terminal device or the server.

[0128] Therefore, by extracting the intent information, slot information and feedback information from the original unprocessed information, and configuring corresponding annotation information for the intent information, slot information and feedback information, the configured intent information, slot information and feedback information are arranged into the second information according to the set rules. The developer can understand what the business party wants to express through the annotation information and the second information after the rule sorting, so that the business party and the developer can communicate efficiently even in different industry fields.

[0129] In one possible implementation, the information configuration module 203 is specifically used to:

[0130] Preset multiple different types of annotation information, where the multiple different types at least include intent type and slot type;

[0131] Map annotation information to intent information and slot information according to type.

[0132] In this embodiment, the specific process of executing the information configuration module 203 is:

[0133] Preset multiple different types of annotation information, where the preset annotation information types include intent types (refer to Figure 2 INTENT in), slot type (refer to Figure 2 SLOT in), feedback type (reference Figure 2 RESULT in , etc.

[0134] After determining the intent information, slot information, and feedback information in the first information, the intent information is matched with the annotation information of the intent type and mapped to each other (refer to Figure 2 INTENT(want_to_buy_ticket) in the slot information, matching the slot information with the slot type annotation information and mapping them to each other, slot information (reference Figure 2 SLOT(user_name) in the feedback information), matching the feedback information with the feedback type annotation information and mapping them to each other (reference Figure 2 RESULT(child_ticket_success) in .

[0135] In one possible implementation, after determining at least the intent information, slot information, and feedback information corresponding to the first event according to the first information, the information determination module 202 is further specifically configured to:

[0136] Determine whether the slot information meets the necessary information set for the intent information;

[0137] If the judgment is negative, prompt information representing the supplementary information is fed back based on the slot information.

[0138] In this embodiment, for each first event type, corresponding slot information can be set in advance. For example, for ticket buying and selling events, the required slot information can be set to name, age, ID number, destination, travel time, etc.

[0139] After executing the information determination module 202, it is determined whether the slot information satisfies the necessary information set for the intent information, wherein the necessary information is the pre-set slot information mentioned above. The determination of whether it is satisfied can be whether the determined slot information contains the set necessary information, or whether the difference between the determined slot information quantity and the necessary information quantity is within a preset threshold.

[0140] If it is determined that the slot information does not meet the necessary information set for the intent information, it means that the amount of information provided by the business direction developer is insufficient, then prompt information representing the supplementary information is fed back based on the slot information. For example, for the ticket buying and selling event, if the name is missing in the slot information determined in the execution information determination module 202, feedback can be given based on the missing slot information to prompt the need to supplement the slot information, such as "You are missing key name information, please supplement it."

[0141] In one possible implementation, the information determination module 202 determines whether the slot information satisfies the necessary information set for the intent information, specifically for:

[0142] At least the intent information and the slot information are input into the dialogue engine for a trial operation;

[0143] Depending on whether the trial run results meet expectations, determine whether the slot information meets the necessary information set for the intent information.

[0144] In this embodiment, the specific process of determining whether the slot information meets the requirements is as follows:

[0145] The currently determined intent information and slot information are input into the existing dialogue engine for trial operation. If the dialogue engine can feedback accurate answers based on the current intent information and slot information, it means that the currently determined slot information is sufficient to meet the query conditions, and the slot information is determined to meet the necessary information set for the intent information. On the contrary, if the dialogue engine feedbacks multiple answers or cannot feedback an answer based on the current intent information and slot information, it means that the currently determined slot information is insufficient to meet the query conditions, and the slot information is determined to not meet the necessary information set for the intent information.

[0146] In one possible implementation, the information determination module 202 is further specifically configured to:

[0147] Determining, based on the first information, operation information representing a clear intention corresponding to the first event;

[0148] Accordingly, when the information configuration module 203 configures the corresponding annotation information for the operation information, it is specifically used for:

[0149] The configured intention information, slot information, operation information and feedback information are arranged according to set rules to obtain second information.

[0150] In this embodiment, in addition to the above-mentioned determination of the intention information, slot information and feedback information, operation information representing the clear intention is also extracted from the first information. The operation information can still be taken as the first information "Purchase a child ticket to Beijing, name xxx, ID number xxx, user confirms payment, feedback result is successful purchase of child ticket" as an example, where the intention information of the first information is to purchase a ticket, and the corresponding operation information is "user confirms payment" in the first information. The extraction method of operation information can also be label recognition or user portrait data recognition. These two recognition methods are consistent with the method of extracting slot information, and will not be repeated here.

[0151] After extracting the operation information, also configure the corresponding annotation information for the operation information. The annotation information refers to Figure 2The configured operation information can be expressed as ACTION(btn_confirm_pay). The configuration method of operation information is the same as that of slot information and intent information, and will not be repeated here.

[0152] When generating the second information, operation information is added on the basis of the original intention information, slot information and feedback information, and the above four information are arranged according to the set rules to obtain the second information.

[0153] In one possible implementation, the second information is a code block, and the code block includes an intention module, a requirement module, and a feedback module;

[0154] Wherein, the intention module includes at least one set of demand modules and feedback modules for mapping intention information;

[0155] The requirement module is used to map slot information and operation information;

[0156] The feedback module is used to map feedback information.

[0157] In this embodiment, the second information can be presented in the form of a code block, see Figure 2 As shown, the intent module is specifically a code block formed by INTENT(xxx){}, which is used to map intent information, the requirement module is specifically a code block formed by EXPECT{}, which is used to map slot information and operation information, and the feedback module is specifically a code block formed by RESULT(), which is used to map feedback information. The requirement module and the feedback module are both located in the intent module to represent the requirement information and processing result information required under the intent.

[0158] Furthermore, the requirement module may include operation information in addition to the mapping slot information and feedback information.

[0159] Furthermore, the same intent information can include multiple sets of demand modules and feedback modules at the same time. Figure 2 It shows that the same intention information (buy tickets) includes two kinds of needs. The first one is the demand information and corresponding feedback information when the age is less than 12 years old, and the second one is the demand information and corresponding feedback information when the age is greater than 12 years old.

[0160] In one possible implementation, the second information is a structure flow chart, which is obtained by converting the code block, specifically:

[0161] Determine the first process node according to the intent module of the code block;

[0162] Determine the second process node according to the requirement module of the code block;

[0163] According to the feedback module of the code block, determine the third process node;

[0164] If the same intent information contains multiple sets of demand modules and feedback modules, determine the intermediate process node that represents the judgment type;

[0165] The first process node, the middle process node, the second process node and the third process node are connected to form a structural flow chart.

[0166] In this embodiment, the structural flow chart can refer to Figure 3 As shown, the confirmation method of the process node can be to determine whether there is an intention module, a requirement module or a feedback module in the code block. If so, the corresponding process node is obtained. Each different process node corresponds to a different style, such as Figure 3 In the figure, the first process node is represented by a rectangle, the second process node is represented by a rounded rectangle, and the third process node is represented by a trapezoid. It should be noted that the style of the process node is not limited to Figure 3 shown.

[0167] In addition to the above-mentioned first process node, second process node, third process node and intermediate process node, it further includes a confirmation process node, which corresponds to the above-mentioned operation information, and may also include multiple identical process nodes if necessary, such as Figure 3 Two second process nodes are included.

[0168] There are many ways to connect the above process nodes. Figure 3 For example, the connection method can be through arrows from top to bottom, the first process node, the second process node, the intermediate process node, the second process node, the confirmation process node and the third process node, Figure 3 The process logic is: first determine the user's intention information (want_to_buy_ticket), determine the necessary slot information (user_name, user_age) based on the intention information, make multiple judgments based on the slot information with non-fixed attributes (user_age<=12, user_age>12), and execute the third process node in each case when the user confirms the intention. It should be noted that the connection method of the process nodes is not limited to Figure 3 The connection method of each process node can be interchanged.

[0169] By generating a structural flow chart from the code, developers can understand the information content more intuitively.

[0170] In one possible implementation, the information arrangement module 204 is further specifically configured to:

[0171] Based on the second information, third information executable by the dialog engine is generated.

[0172] In this embodiment, the third information is specifically information that can be directly used by the dialogue engine, such as slot information, intent information, etc. The third information can be extracted through code blocks or through structural flowcharts, wherein the extraction of code blocks can be to parse the annotation information in the code blocks, such as INTENT, SLOT, etc., and then obtain the corresponding information based on the annotation information; the extraction of structural flowcharts can be to determine the pattern of each node in the flowchart or the pre-set annotation, and obtain the corresponding information based on the determination result. Through this step, data that can be directly run by the dialogue engine can be quickly obtained to quickly generate a prototype system.

[0173] In one possible implementation, the information arrangement module 204 is specifically used for:

[0174] According to the logical operation relationship between the intention information, the slot information and the feedback information, the corresponding logical symbols are configured and the corresponding sorting processing is performed to generate the second information.

[0175] In this embodiment, the specific process of generating the second information is as follows:

[0176] Taking the finally generated second information as a code block as an example, it is generally understood that the logical running order of intent information, slot information and feedback information is intent information, slot information and feedback information, to indicate that a corresponding group or multiple groups of slot information and feedback information are included under a total intent information.

[0177] refer to Figure 2 As shown, during configuration, the intent information is configured first, specifically, a logical symbol indicating that it is an intent is added to the left side of the intent information, such as INTENT{}. According to the logical operation order, one or more groups of slot information and feedback information are then configured. For each group of slot information, a logical symbol indicating one of the required information under the same intent information is added, such as EXPECT[]. A logical symbol indicating that it is slot information is also added to the left side of each slot information, such as SLOT. Furthermore, for slot information with non-fixed attribute values, such as user_age, one of the possibilities of slot information with non-fixed attribute values ​​needs to be written in the SLOT{} code block, such as Figure 2 In the code block, user_age<=12 is used to indicate that the current demand information is the information when user_age<=12. Furthermore, operation information can be added to the demand information code block to indicate the intention to confirm under the current slot information. Similarly, a logical symbol indicating that it is operation information is added to the left side of the operation information, such as ACTTION(). The feedback information is located in the intention information code block and below the demand information to indicate the processing result given for the above demand information under the same intention information. Similarly, a logical symbol indicating that it is feedback information is added to the left side, such as RESULT().

[0178] If there are multiple sets of slot information and feedback information, they can be placed below the previous requirement information code block to represent another type of requirement information. The logic symbol generation method for each piece of information is the same as above. By arranging multiple sets of slot information and feedback information in an up-and-down manner, multiple related branch processes can be split into a single independent process data definition to avoid mutual interference between branch logic judgments, which is convenient for viewing and demonstration.

[0179] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0180] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0181] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An information processing method, the method comprising: Acquire first information, where the first information includes a first event and requirement information for processing the first event; Determine at least the intent information, slot information, and feedback information corresponding to the first event according to the first information; Configuring corresponding annotation information for the intent information, slot information, and feedback information respectively, wherein the annotation information is used to characterize the information type; Arrange at least the configured intention information, slot information, and feedback information according to a set rule to obtain second information; The determined second information is sent to a target object, and the target object outputs the content expressed by the first information according to the annotation information and the second information.

2. According to the method of claim 1, the configuring corresponding annotation information for the intent information, slot information and feedback information respectively comprises: Preset multiple different types of annotation information, the multiple different types at least including intent type and slot type; The annotation information is mapped to the intent information and slot information according to the type.

3. The method according to claim 1, after determining at least the intent information, slot information and feedback information corresponding to the first event according to the first information, the method further comprises: Determining whether the slot information satisfies necessary information set for the intent information; If the judgment is negative, prompt information representing the supplementary information is fed back based on the slot information.

4. The method according to claim 3, wherein the determining whether the slot information satisfies the necessary information set for the intent information comprises: At least inputting the intention information and slot information into the dialogue engine for a trial operation; According to whether the trial operation result meets expectations, it is determined whether the slot information meets the necessary information set for the intention information.

5. The method according to claim 1, further comprising: Determine, based on the first information, operation information representing a clear intention corresponding to the first event; Accordingly, corresponding annotation information is configured for the operation information; The configured intention information, slot information, operation information and feedback information are arranged according to set rules to obtain second information.

6. The method according to claim 5, wherein the second information is a code block, and the code block includes an intention module, a requirement module, and a feedback module; in, The intention module includes at least one set of the demand module and the feedback module, which is used to map the intention information; The requirement module is used to map the slot information and the operation information; The feedback module is used to map the feedback information.

7. According to the method of claim 6, the second information is a structure flow chart, and the structure flow chart is obtained by converting the code block, specifically: Determining a first process node according to the intent module of the code block; Determining a second process node according to a requirement module of the code block; Determining a third process node according to the feedback module of the code block; If the same intention information contains multiple groups of demand modules and feedback modules, then determine the intermediate process node representing the judgment type; The first process node, the intermediate process node, the second process node and the third process node are connected to form the structural flow chart.

8. The method according to claim 1, further comprising: Based on the second information, third information executable by the dialog engine is generated.

9. The method according to claim 1, arranging the configured intent information, slot information and feedback information according to a set rule to obtain the second information, comprising: According to the logical operation relationship between the intention information, slot information and feedback information, corresponding logical symbols are configured and corresponding sorting processing is performed to generate the second information.

10. An information processing device, comprising: An information acquisition module, configured to acquire first information, wherein the first information includes a first event and requirement information for processing the first event; an information determination module, configured to determine at least intent information, slot information, and feedback information corresponding to the first event based on the first information; An information configuration module, used to configure corresponding annotation information for the intent information, slot information and feedback information respectively, wherein the annotation information is used to characterize the information type; An information arrangement module, used to arrange at least the configured intention information, slot information and feedback information according to a set rule to obtain second information; The information sending module is used to send the determined second information to a target object, and the target object outputs the content expressed by the first information according to the marked information and the second information.

Images