Information processing method and electronic equipment

Abstract

The invention provides an information processing method, which comprises the following steps: in response to input of target information, performing intention recognition on partial information which is not subjected to intention recognition in the target information to obtain first sub-information which represents at least one intention; and adding an intention identifier for indicating the first sub-information into the target information, and inputting the target information again.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and more particularly, to an information processing method and an electronic device. Background Art

[0002] Intention recognition is a key task in natural language processing and is widely used in fields such as intelligent assistants, task automation, and human-computer interaction. Ideal intention recognition should be able to accurately extract all the intentions expressed by the user and maintain their logical order to ensure that task execution meets expectations. However, there are still many problems in the existing technologies in practical applications. For example, the number of intention extractions is often inaccurate, and intention omission or redundancy may occur, making the system unable to stably recognize the complete user requirements. On the other hand, the order of intentions is uncontrollable, resulting in the task execution process may not match the user's expectations, affecting the coherence and correctness of the operation. Summary of the Invention

[0003] In view of this, the present disclosure provides an information processing method and an electronic device.

[0004] One aspect of the present disclosure provides an information processing method, including: in response to the input of target information, performing intention recognition on the part of the target information that has not been subjected to intention recognition to obtain first sub-information, where the first sub-information represents at least one intention; adding an intention identifier for indicating the first sub-information to the target information, and inputting the target information again.

[0005] According to an embodiment of the present disclosure, performing intention recognition on the part of the target information that has not been subjected to intention recognition to obtain first sub-information includes: in response to the existence of at least one intention identifier in the target information, obtaining the recognized information, where the recognized information is at least one second sub-information indicated by at least one intention identifier in the target information; according to the recognized information, performing intention recognition on the part of the information that has not been subjected to intention recognition, and determining the first sub-information from the part of the information, where at least part of the intentions represented by the first sub-information has a logical relationship with at least part of the intentions represented by the recognized information; generating an intention identifier for the first sub-information.

[0006] According to an embodiment of the present disclosure, performing intention recognition on the part of the target information that has not been subjected to intention recognition to obtain first sub-information includes: in response to the non-existence of an intention identifier in the target information, performing intention recognition on the target information, and determining the first sub-information from the target information, where the execution priority of the intention represented by the first sub-information meets the target condition; generating an intention identifier for the first sub-information.

[0007] According to an embodiment of the present disclosure, the target information also includes a recursive identifier, the recursive identifier is a first identification identifier or a second identification identifier, and intent identification is performed on part of the target information that has not been subjected to intent identification, including: in response to the recursive identifier being the first identification identifier, intent identification is performed on part of the target information that has not been subjected to intent identification; wherein the first identification identifier at least represents the continuation of intent identification on the part of the information that has not been subjected to intent identification.

[0008] According to an embodiment of the present disclosure, the information processing method also includes: in response to the absence of information representing the intent in the partial information for which intent recognition has not been performed, updating the recursive identifier to a second identification identifier; wherein the second identification identifier at least represents the cessation of intent recognition in the partial information for which intent recognition has not been performed.

[0009] According to an embodiment of the present disclosure, after adding an intent identifier for indicating the first sub-information in the target information, the information processing method further includes: adjusting the position of the first sub-information in the target information so that in the target information, the first sub-information is located before the portion of information on which intent recognition is not performed; the information processing method further includes: inputting the first sub-information indicated by each intent identifier in the target information into the intent extraction module in sequence to obtain multiple intents.

[0010] According to an embodiment of the present disclosure, intent recognition is performed on part of the target information that has not been subjected to intent recognition to obtain first sub-information, including: based on an intent splitting module, intent recognition is performed on part of the target information that has not been subjected to intent recognition to obtain first sub-information.

[0011] According to an embodiment of the present disclosure, the acquisition process of the intent splitting module includes: acquiring sample data, each group of sample data includes sample target information and multiple sample output information, the sample output information includes at least one sample sub-information, and the sample sub-information represents at least one intent; based on the intention splitting module, multiple intent splittings are performed on part of the sample target information that has not been subjected to intent recognition to obtain multiple output information, wherein the output information includes at least one sub-output information, and the sub-output information represents at least one intent identified in the sample target information. In the multiple intent splittings, the input information of the latter intent splitting module is the output information generated by the previous intent splitting module; multiple first losses are determined based on the sub-output information of the multiple output information and the sample sub-information in the sample output information that satisfies the corresponding relationship with each output information; and based on the first loss, the parameters of the intent splitting module are adjusted until the first loss is less than or equal to the loss threshold.

[0012] According to an embodiment of the present disclosure, the sample output information also includes a sample recursive identification, and the output information also includes identification information, and the identification information represents whether intent recognition is performed on part of the information for which intent recognition has not been performed; the information processing method also includes: determining multiple second losses based on the identification information in the multiple output information, and the sample recursive identification in the sample output information that satisfies the corresponding relationship with each output information; and adjusting the parameters of the intent splitting module based on the second loss until the second loss is less than or equal to the loss threshold.

[0013] Another aspect of the present disclosure provides an information processing device, including: a first identification module, used to, in response to input of target information, perform intent identification on part of the target information that has not been subjected to intent identification, to obtain first sub-information, wherein the first sub-information represents at least one intent; and a first input module, used to add an intent identifier indicating the first sub-information to the target information, and input the target information again.

[0014] Another aspect of the present disclosure provides an electronic device, comprising: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the information processing method of any one of the aforementioned embodiments.

[0015] Another aspect of the present disclosure provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the information processing method according to any one of the aforementioned embodiments.

[0016] Another aspect of the present disclosure provides a computer program product, including a computer program / instruction, characterized in that when the computer program / instruction is executed by a processor, the operation of the information processing method of any of the aforementioned embodiments is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0018] Figure 1 The flowchart of the information processing method according to the embodiment of the present disclosure is schematically shown;

[0019] Figure 2 A flowchart of generating an intention identifier in an information processing method according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 3 Another flowchart of generating an intention identifier in the information processing method according to an embodiment of the present disclosure is schematically shown;

[0021] Figure 4 Another flowchart of performing intention recognition in the information processing method according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 5 Another flowchart of the information processing method according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 6 Another flowchart of the information processing method according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 7 A flowchart of a process for obtaining an intent splitting module according to an embodiment of the present disclosure is schematically shown;

[0025] Figure 8 Another flowchart schematically shows a process of obtaining an intention splitting module according to an embodiment of the present disclosure;

[0026] Fig. 9 A block diagram schematically shows an information processing device according to an embodiment of the present disclosure; and

[0027] Fig.10 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0028] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0029] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0030] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0031] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0032] In the embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security, and national security.

[0033] An embodiment of the present disclosure provides an information processing method, including: in response to input of target information, performing intent recognition on part of the target information that has not been subjected to intent recognition to obtain first sub-information, the first sub-information representing at least one intent; adding an intent identifier for indicating the first sub-information to the target information, and inputting the target information again.

[0034] Figure 1 The flowchart of the information processing method according to the embodiment of the present disclosure is schematically shown.

[0035] like Figure 1 As shown, the information processing method may at least include operations S110 to S120.

[0036] In operation S110, in response to input of target information, intent recognition is performed on a portion of the target information that has not been subjected to intent recognition to obtain first sub-information, where the first sub-information represents at least one intent.

[0037] The target information refers to the original text information received for intent recognition and decomposition. This information usually comes from the user's natural language input and may contain one or more intents. Intent refers to a specific semantic unit in the target information that represents the operation or instruction that the user expects to perform. Intents are usually expressed in natural language and can correspond to a specific operation behavior, request or instruction. For example, the target information can be "Help me summarize a document and send it to Manager Wang. Remember to translate it into French before sending the summary." Then it can contain the three intents "Summarize the document", "Translate the document into French", and "Send the document to Manager Wang".

[0038] The first sub-information may refer to at least part of the target information, which is extracted from part of the target information that has not been identified for intent and represents at least one intent. The first sub-information may be specific text content extracted from the target information, used to express the identified intent, and is specific content in a section of the target information, while the corresponding intent is a semantic unit in the target information that represents the operation or instruction that the user expects to perform, and is an abstract concept. For example, "help me summarize a document" is the first sub-information, and "summarize the document" is the intent.

[0039] The first sub-information can represent an intention. In each intent recognition process, the first sub-information extracted from the target information can correspond to an intention that can be clearly expressed. For example, "help me summarize a document" corresponds to "summarize the document".

[0040] The first sub-information can represent multiple intents. During each intent recognition process, the first sub-information extracted from the target information can correspond to multiple clearly expressible intents. For example, "Help me translate a document into French and English" corresponds to the two intents "Translate the document into French" and "Translate the document into English".

[0041] In operation S120, an intention identifier for indicating the first sub-information is added to the target information, and the target information is input again.

[0042] The intent identifier refers to the label information used to indicate the first sub-information extracted after the intent identification in the target information. The intent identifier corresponds to the first sub-information one by one. Based on the intent identifier, the part of the target information that has been subjected to the intent identification and the part that has not been subjected to the intent identification can be distinguished.

[0043] For example, the intent identifier is <###>, and the target information is "Help me summarize a document and send it to Manager Wang. Remember to translate it into French before sending it." After the first input, the recognized part, i.e., the first sub-information "Help me summarize a document", is obtained. Then, the intent identifier is added to the first sub-information to indicate the information, such as: "Help me summarize a document <###> and send it to Manager Wang. Remember to translate it into French before sending it."

[0044] Re-entering the target information means re-entering the target information containing the intent identifier after at least one intent recognition is completed, so as to repeat the above steps and continue to perform intent recognition on the unrecognized part. At this time, the target information has been partially parsed after the previous intent recognition and the intent identifier is attached, so the content when re-entered is different from the original target information. Through the intent identifier, the system can identify the parsed part and ensure that the re-entered target information only performs new intent recognition on the unrecognized part.

[0045] For example, the target information is "Help me summarize a document and send it to Manager Wang. Remember to translate it into French before sending the summary." After the first input and execution of the above process, a new target information with an intent identifier is obtained, "Help me summarize a document <###> and send it to Manager Wang. Remember to translate it into French before sending the summary." Then, "Help me summarize a document <###> and send it to Manager Wang. Remember to translate it into French before sending the summary." is re-entered. At this time, "Help me summarize a document" has been marked with the intent identifier, and the part that has never been subjected to intent identification: "Send to Manager Wang. Remember to translate it into French before sending the summary." is subjected to intent identification, and the first sub-information of an information "Translate into French" is extracted and marked, resulting in "Help me summarize a document <###> Translate into French <###> and send it to Manager Wang".

[0046] By repeatedly executing such a recursive process of "input" → "obtain sub-information" → "mark sub-information" → "input again" → "obtain sub-information again", through multiple rounds of intent recognition, the parsing and labeling of intents are gradually completed, so that the structure of the target information is gradually optimized; in addition, when the target information contains multiple intents, directly identifying all intents may lead to information confusion or misjudgment, but through repeated execution, the system can disassemble the information more finely to ensure that each intent is accurately extracted.

[0047] In the example of the intention identifier above, the identifier is only added at the end of the first sub-information, except for the first time the first sub-information is obtained. When adding the identifier, the obtained first sub-information needs to be adjusted to after the first sub-information has been obtained, so that the first sub-information is between the "beginning" and the "intention identifier", or, between the "intention identifier" or the "intention identifier", so that it can be accurately indicated.

[0048] In other embodiments, different intent identifiers may be used to indicate the first sub-information. For example, the first sub-information may be enclosed in brackets, such as "<help me summarize a document><send to Manager Wang>, remember to translate the summary into French before sending it." The first sub-information is "<help me summarize a document>". It should be noted that the given intent identifiers are only examples, and those skilled in the art may select any appropriate intent identifier to indicate the first sub-information.

[0049] Figure 2 The flowchart of generating an intention identification in the information processing method according to an embodiment of the present disclosure is schematically shown.

[0050] like Figure 2 As shown, operation S110 may include operations S210 to S230.

[0051] In operation S210, in response to at least one intent identifier being present in the target information, identified information is acquired, where the identified information is at least one second sub-information indicated by the at least one intent identifier in the target information.

[0052] In the case where the target information already contains at least one intent identifier, based on each intent identifier, the corresponding identified information is extracted, that is, the part of the target information for which intent identification has been completed.

[0053] The identified information specifically refers to at least one second sub-information indicated by at least one intent identifier in the target information. The second sub-information is similar to the first sub-information in terms of representation content, both representing at least one intent. The difference between the two is that the second sub-information can be regarded as the first sub-information marked by the intent identifier after the previous or previous intent identification. To distinguish it from the first sub-information in this operation, it is called the second sub-information.

[0054] The identified information may be a second sub-information, for example, the second sub-information marked by the intent identifier extracted from the previous input. For example, if two intent identifications have been performed, the target information containing the intent identifier "help me summarize a document <###> and translate it into French <###> and send it to Manager Wang" is used as the input for this time, where "translate into French" is the second sub-information indicated by the intent identification identifier extracted from the previous input, then the identified information is "translate into French <###>".

[0055] The identified information may be multiple second sub-information, for example, it may be the second sub-information marked by the intent identifier extracted from the inputs of a specified number of times before the current input. For example, the specified number of times is two, and the intent identification has been performed twice. The target information containing the intent identifier "Help me summarize a certain document <###> and translate it into French <###>, and send it to Manager Wang" is used as the input for this time, where "Translate into French" is the second sub-information indicated by the intent identification identifier extracted from the previous input. Then, the identified information is "Help me summarize a certain document <###> and translate it into French <###>". For another example, it may be all the second sub-information marked by the intent identifier extracted before the current input, and examples will not be repeated here.

[0056] In operation S220, based on the identified information, intent recognition is performed on partial information that has not been subjected to intent recognition, and first sub-information is determined from the partial information, and at least a portion of the intent represented by the first sub-information has a logical relationship with at least a portion of the intent represented by the identified information.

[0057] In operation S230, an intent identifier is generated for the first sub-information.

[0058] In the process of identifying the intent of the target information, on the premise that at least one identified information has been extracted, the identified information is used to assist in parsing the part of the target information that has not been identified, so as to determine the new first sub-information. There is a certain semantic or logical association between the newly extracted first sub-information and the identified information. A new first sub-information is extracted through the identified information and this association, ensuring that the results of each round of intent decomposition are semantically coherent and meet the actual needs or context of the user.

[0059] Logical relationships may include but are not limited to: time sequence relationship, conditional dependency relationship, cause-and-effect relationship, purpose-and-means relationship, etc. Time sequence relationship means that some tasks or operations need to be performed before other tasks can be performed; conditional dependency relationship means that the execution of certain intentions needs to rely on the completion of other intentions, or certain conditions can be met before execution; cause-and-effect relationship means that the execution of a certain intention is the result or premise of the execution of another intention. Purpose-and-means relationship means that a certain intention is proposed to achieve the purpose of another intention.

[0060] For example, the logical relationship is a time sequence relationship, which means that some intentions need to be executed before other intentions can be carried out. For example, the target information is: "Organize the meeting minutes, and then send the minutes to the project team members." The first intention: "Organize the meeting minutes" must be executed before the next intention can be executed. The second intention: "Send to project team members" must be carried out after the meeting minutes are organized. Therefore, there is a time sequence relationship between "organize meeting minutes" and "send to project team members." Then, correspondingly, under the premise that the information "organize meeting minutes" is the identified information, the first sub-information obtained by this identification should be "send to project team members."

[0061] For example, the logical relationship is a conditional dependency, which means that the execution of certain intentions depends on the completion of other intentions, or certain conditions must be met before execution can be performed. For example, the target information is: "After the customer confirms the order, ship and update the inventory." "Update inventory" must be performed after "shipping" is completed, because the update of inventory depends on the actual number of shipments. In this example, the conditional dependency is that "update inventory" depends on the completion of "shipping". Then, correspondingly, under the premise that the information of "shipping" is the identified information, the first sub-information obtained in this identification should be "update inventory".

[0062] For example, the logical relationship is a causal relationship, which means that one intention is the result or premise of another intention. For example, the target information is: "I will go running tomorrow and need to wear sneakers." Going running is the premise, and needing to wear sneakers is the result. Therefore, "wearing sneakers" should come after "running". Correspondingly, under the premise that the information "running" is the recognized information, the first sub-information obtained in this recognition should be "wearing sneakers".

[0063] For example, the logical relationship is a purpose-means relationship, which means that a certain intention is carried out to achieve a specific intention. For example, the target information is: "Go to the supermarket to buy some milk." "Go to the supermarket" is to achieve the intention of "buy milk". Then, correspondingly, under the premise that the information "go to the supermarket" is the identified information, the first sub-information obtained in this identification should be "buy milk".

[0064] At least part of the intent represented by the first sub-information has a logical relationship with one of the intents represented by the identified information. Referring to the aforementioned content, it will not be repeated here. At least part of the intent represented by the first sub-information has a logical relationship with multiple intents represented by the identified information. For example, the target information is "Help me arrange a meeting, determine the time, prepare the agenda, and then notify the participants." Among them, there is no priority between "determine the time" and "prepare the agenda", but "notify the participants" needs to be executed after the first two intentions are completed. Therefore, under the premise that "determine the time" is identified, it is necessary to extract "prepare the agenda" first according to the logical relationship, and then "notify the participants" in the next round of input and recognition.

[0065] By identifying and utilizing the recognized information in the target information, that is, the partial intention previously disassembled, and identifying the unprocessed partial information in each round of recursion, it is ensured that there is a logical relationship between the disassembled first sub-information and the recognized information, so that in the process of intention disassembly, the disassembly priority and logical order of the first sub-information can be coordinated with the recognized intent, avoiding the situation of incorrect or missing disassembly order.

[0066] Figure 3 Another flowchart for generating an intent identifier in an information processing method according to an embodiment of the present disclosure is schematically shown.

[0067] like Figure 3 As shown, based on the foregoing embodiment, operation S110 may include operations S310 to S320.

[0068] In operation S310, in response to the absence of an intent identifier in the target information, intent recognition is performed on the target information, and first sub-information is determined from the target information, and an execution priority of the intent represented by the first sub-information meets the target condition.

[0069] In operation S320, an intent identifier is generated for the first sub-information.

[0070] The absence of intent identification in the target information means that the target information has not been identified in any intent in the initial state, and it is impossible to disassemble or judge the order based on the existing identification. During the intent recognition process, it will judge and determine which intent should be executed first based on the preset conditions to ensure the rationality of the execution and the consistency of the target conditions.

[0071] For example, the target condition is task priority, which means that some intentions may be more urgent or important than other tasks, and the intentions with higher priority will be executed first. For example, the target information is "handle customer requests and arrange weekly meetings." "Handle customer requests" and "arrange weekly meetings" have no sequential dependency in terms of time logic, but the processing priority of "handle customer requests" may be higher than "arrange weekly meetings", so the first sub-information "handle customer requests" will be determined first.

[0072] For example, the target condition is a sequential relationship, which means that some intentions need to be executed first before subsequent intentions can be executed. For example, the target information is "Check the order information, and ship after confirming the order." The "Confirm order" intention will be recognized and executed first, and the "Shipping" intention will be executed only after the confirmation is completed, ensuring that the execution order conforms to the logical relationship.

[0073] According to the embodiments of the present disclosure, when there is no intent identification in the target information, intent recognition is performed, and the first intent to be executed is determined based on the target conditions (such as task priority, execution logic, time sensitivity, and resource availability, etc.). This allows the most urgent or first task to be executed to be flexibly identified, and the rationality and sequentiality of task execution to be ensured. It also provides an analytical basis for subsequent intent recognition, thereby avoiding confusion or conflict in task execution and improving the accuracy and efficiency of multi-tasking.

[0074] According to the embodiment of the present disclosure, the information processing method may further include operation S130.

[0075] In operation S130, in response to satisfying the target condition, the target information is stopped from being input again. The target condition may be, for example, that the number of inputs satisfies a preset input number threshold, for example, that the first sub-information obtained is the target sub-information, for example, that the part of the target information that has not been recognized for intent does not exist.

[0076] For example, when only a preset number of intentions need to be extracted, assuming 3 times, then the number of recursions only needs to be limited to 3.

[0077] For example, when the user only needs to extract the intention to do a specific thing, for example, the target information is "help me summarize a document and send it to Manager Wang. Remember to translate it into French before sending the summary, and then go to the supermarket to buy milk", and the user only needs the intention before going to the supermarket, then after "go to the supermarket" is recognized, the recursive input can be stopped.

[0078] Figure 4 Another flowchart for performing intent recognition in the information processing method according to an embodiment of the present disclosure is schematically shown.

[0079] like Figure 4 As shown, based on the foregoing embodiment, the target information further includes a recursive identifier, the recursive identifier is a first identification identifier or a second identification identifier, and operation S110 may include operation S410 or S420.

[0080] In operation S410, in response to the recursive identifier being the first identification identifier, the intent identification is performed on the part of the target information that has not been subjected to the intent identification, wherein the first identification identifier at least indicates that the intent identification is continued for the part of the information that has not been subjected to the intent identification.

[0081] In operation S420, in response to the absence of information representing the intent in the partial information for which the intent recognition has not been performed, the recursive identifier is updated to a second identification identifier; wherein the second identification identifier at least represents stopping the intent recognition in the partial information for which the intent recognition has not been performed.

[0082] The recursive flag represents the iconic information used to control the recursive decomposition process during intent recognition, indicating whether to continue to decompose the part of the target information that has not been recognized for intent. The recursive flag helps to determine whether it is necessary to continue to decompose the remaining information in each round of intent decomposition until all intents are recognized. The recursive flag will be updated according to the needs of the current task after each round of decomposition, indicating whether to continue processing the undecomposed part. Among them, the first recognition flag indicates to continue with the next round of intent recognition, that is, the remaining part still needs to be further decomposed. The second recognition flag: indicates to stop decomposition, indicating that all intent recognitions of the target information have been completed.

[0083] For example, the first identification mark is <y>, the second identification mark is <n>, the target information is "Help me summarize document 1 and send it to Manager Wang. Remember to translate it into French before sending it." After the first input, the new target information is obtained, "Help me summarize document 1 <###> and send it to Manager Wang. Remember to translate it into French before sending it < Y >." After entering this target information again, the result is "Help me summarize document 1 <###> and remember to translate it into French <###> and send it to Manager Wang, < Y >." Continue to enter the target information and get "Help me summarize document 1 <###> and remember to translate it into French <###> and send it to Manager Wang <###>." <n>", in this case the recursive identifier in the target information is <n>, indicating that the intent recognition of some information that has not been recognized is stopped

[0084] It should be noted that when the target information is input for the first time, the first identification mark can be automatically added to the target information to indicate that the partial information that has not been identified should continue to be identified. At this time, the partial information that has not been identified is the entire target information. It is also possible to generate a recursive mark for the target information after the target information that does not contain a recursive mark is input for the first time to indicate whether to continue to identify the partial information that has not been identified in the next round of input.

[0085] In other embodiments, the information processing method may only include the above-mentioned operation S420, that is, only when there is no information representing the intent in the partial information for which the intent recognition has not been performed, a second identification identifier is generated indicating that the intent recognition for the partial information for which the intent recognition has not been performed is stopped, and when the existence of the second identification identifier is detected, the identification is stopped, and when the second identification identifier does not exist, the identification identification is continued.

[0086] The target information may contain information that clearly represents the intent, or it may contain information that does not represent the intent. For example, the target information is "The weather is nice today, remind me to go out for a run later", where "The weather is nice today" does not represent any intent. However, in some design frameworks, if the intent extraction is not stopped in some way, the intent recognition of "The weather is nice today" will be forced, resulting in an erroneous intent that is completely inconsistent with the original semantics. Therefore, it is necessary to control the recognition of intent through recursive identification. By controlling the recursive identification, it is possible to determine when to stop the disassembly of the target information, ensuring that each round of disassembly is only performed on valid information that clearly represents the intent, thereby improving the efficiency and accuracy of intent recognition and avoiding unnecessary intent recognition of irrelevant content that leads to misrecognition and erroneous execution.

[0087] Figure 5 Another flowchart of the information processing method according to an embodiment of the present disclosure is schematically shown.

[0088] like Figure 5 As shown, based on the foregoing embodiment, after operation S120, the information processing method may further include operations S510 to S520.

[0089] In operation S510, the position of the first sub-information in the target information is adjusted so that in the target information, the first sub-information is located before the part of the information that has not been subjected to intent recognition. The order is adjusted so that in the process of intent recognition in the target information, the first sub-information can be arranged in a logical order, thereby ensuring the smooth progress of intent recognition. In each round of intent disassembly, when a certain first sub-information is identified, the first sub-information needs to be placed in a suitable position in the target information to avoid confusion or incorrect execution order.

[0090] In operation S520, the sub-information indicated by each intent identifier in the target information is sequentially input into the intent extraction module to obtain multiple intents.

[0091] The intent extraction module is used to convert the expression of natural language into a specific and clear execution intention. In the identified first sub-information, each first sub-information that clearly represents the intention can be accurately extracted and converted into a machine-executable intention. Specifically, the target information is divided into several first sub-information during the intention decomposition process, where each first sub-information corresponds to at least one independent intention. By inputting these sub-information into the intent extraction module according to the order of the first sub-information in the target information, each sub-information can be processed one by one to extract the intention contained therein, and ensure that the extraction order of each intention is consistent with the order in the target information.

[0092] For example, the initial target information is "Help me summarize Document 1 and send it to Manager Wang. Remember to translate it into French before sending the summary." After two rounds of input, the result is "Help me summarize Document 1<###>Remember to translate it into French before sending the summary<###>Send to Manager Wang." "Remember to translate it into French before sending the summary" is arranged after "Help me summarize Document 1" and before the part "Send to Manager Wang" where intent recognition is not performed. The first sub-information is arranged in order of logical relationship, so that when the intent extraction module extracts the machine-executable intent based on the first sub-information, the intent of "summarize the document" is before the intent of "send to Manager Wang".

[0093] In other embodiments, the position of the sub-information indicated by the intent identifier in the target information may not be adjusted, but the intent identifier may be improved, that is, the intent identifier also indicates the number of times the target information is input when the first sub-information is obtained. For example, a numerical intent identifier is used to indicate each first sub-information, and the first sub-information obtained for the first time is <1> As indicated, the first sub-information obtained for the second time is <2> The target information is as follows: For example, the initial target information is "help me summarize a document and send it to Manager Wang. Remember to translate it into French before sending it." The final target information is "help me summarize a document <1> Send to Manager Wang <3> , remember to translate into French before sending your summary <2> ", the order is indicated by the intention identifier. When the first sub-information is subsequently input into the intention recognition module in sequence, it is input in the order indicated by the intention identifier, which can achieve the same effect as the aforementioned embodiment.

[0094] Figure 6 Another flowchart of the information processing method according to an embodiment of the present disclosure is schematically shown.

[0095] like Figure 6 As shown, based on the foregoing embodiment, operation S110 may include operation S610.

[0096] In operation S610, based on the intention splitting module, the intention of the target information is not recognized, and the first sub-information is obtained. Specifically, the intention splitting module is responsible for identifying and extracting the part of the target information that is not recognized, and the first sub-information is obtained by performing intention recognition on the part.

[0097] Figure 7 The flowchart of the process of obtaining the intent splitting module according to the embodiment of the present disclosure is schematically shown.

[0098] like Figure 7 As shown in the aforementioned Figure 6 Based on the illustrated embodiment, the intent splitting module may be a model trained based on deep learning or other machine learning techniques, and the process of obtaining the intent splitting module may include operations S710 to S740.

[0099] In operation S710, sample data is acquired, each set of sample data includes sample target information and a plurality of sample output information, the sample output information includes at least one sample sub-information, and the sample sub-information represents at least one intention.

[0100] The sample target information in the sample data represents a sample of the complete information input by the user, and the sample output information is a sample corresponding to the intermediate or final output result to be generated in the training and disassembly process of the intent splitting module. The sample output information includes at least one sample sub-information, and the sample sub-information represents at least one intent. The sample sub-information is a sample corresponding to the first sub-information to be generated by the intent splitting module, that is, the sample sub-information is the training label of the first sub-information to be generated in the training process of the intent splitting module, and the sample output information is the training label of the complete output information in the training process of the intent splitting module. According to the embodiments of the present disclosure, in each sample output information, the sample sub-information can be marked with an intent identifier to distinguish each sample sub-information in the sample information.

[0101] In operation S720, based on the intent splitting module, multiple intent splittings are performed on part of the sample target information that has not been subjected to intent recognition to obtain multiple output information, wherein the output information includes at least one sub-output information, and the sub-output information represents at least one intent identified in the sample target information. In multiple intent splittings, the input information of the latter intent splitting module is the output information generated by the previous intent splitting module.

[0102] Specifically, the training process is a recursive training process. After the sample target information is input into the intent splitting module, the first output information is obtained. The output information includes at least one sub-output information. The sub-output information represents the partial information representing at least one intent in the sample target information identified by the intent splitting module during the training process. In this embodiment, after obtaining the sub-output information, it can be marked with an intent identifier to indicate the sub-output information in the output information. Then, the obtained output information is used as the input information of the intent splitting module and input into the intent splitting module again. The intent splitting module will perform intent recognition from the part of the input information not indicated by the intent identifier based on the sub-output information that has been indicated in the input information, obtain at least one new sub-output information, and then obtain a new output information. Repeat this process to obtain multiple output information by a recursive method.

[0103] The multiple sample output information in each group of sample data has a target arrangement order, and the target arrangement order corresponds to the order in which the output information is generated by the intention splitting module, that is, the output information generated by the intention splitting module for the first time corresponds to the first sample output information in the target arrangement order, and the output information generated by the intention splitting module for the second time corresponds to the second sample output information in the target arrangement order, until the output information finally generated by the intention splitting module corresponds to the final sample output information in the target arrangement order.

[0104] In operation S730, multiple first losses are determined based on the sub-output information of the multiple output information and the sample sub-information in the sample output information that satisfies the corresponding relationship with each output information. The corresponding relationship represents the aforementioned sequential corresponding relationship, that is, the matching relationship between the output order of the output information and the target arrangement order. In this way, multiple groups of comparison data are formed: "the first output information and the first sample output information", "the second output information and the second sample output information", etc. In each group of comparison data, the sub-output information in the output information and the sample sub-information in the sample output information are used to calculate the first loss according to the loss function, and the first loss represents the error between the sub-output information in the output information generated by the intention splitting module and the sample sub-information in the sample output information.

[0105] The calculation of the first loss can use methods such as Cross-Entropy Loss or Mean Squared Error (MSE) to measure the accuracy of the model in generating sub-output information.

[0106] In operation S740, according to the first loss, the parameters of the intention splitting module are adjusted until the first loss is less than or equal to the loss threshold. When the first loss is less than or equal to the loss threshold, it means that each sub-output information in each output information generated by the intention splitting module according to the sample target information is similar or identical to each sample sub-information in the sample output information, and even if there is a difference, it is within an acceptable range. Then it means that the training of the intention splitting module for the sub-output information in the output information has been completed, and the intention splitting module can better extract the sub-output information that meets the requirements from the input information.

[0107] Figure 8 Another flowchart of a process for obtaining an intent splitting module according to an embodiment of the present disclosure is schematically shown.

[0108] According to the embodiment of the present disclosure, Figure 7 Based on the embodiment shown, the sample output information also includes a sample recursive identification, and the output information also includes identification information, which indicates whether to perform intent recognition on the part of the information that has not been recognized. Figure 8 As shown, the process of obtaining the intention splitting module also includes operations S810~S820.

[0109] In operation S810, multiple second losses are determined based on the identification information in the multiple output information and the sample recursive identification in the sample output information that satisfies the corresponding relationship with each of the output information. The output information also includes identification information, and the identification information represents whether to continue to perform intent recognition on the partial information that has not been recognized. In the aforementioned recursive training process, identification information is generated at the same time as the sub-output information, and the identification information indicates whether it is necessary to continue to perform intent recognition on the partial information in the output information that has not been recognized after the current intent splitting module generates output information based on the input information. It should be noted that during the training process, if before the same number of output information as the sample output information is generated, that is, before the recursion ends, that is, the identification information representing the cessation of intent recognition on the partial information that has not been recognized is generated, the recursion can still continue until the number of output information is the same as the number of sample output information.

[0110] Correspondingly, the sample output information also includes a sample recursive identifier. According to the aforementioned target arrangement order, only the sample recursive identifier in the last sample output information is the identifier information that indicates that the intent recognition of the part of the information that has not been recognized has been stopped, while the sample recursive identifiers in the remaining sample output information are the identifier information that indicates that the intent recognition of the part of the information that has not been recognized has been continued. Then, according to the output order and the target arrangement order, multiple groups of comparison data are formed, and the identifier information in the output information in each group of data and the sample recursive identifier in the sample output information are included. The second loss is calculated according to the loss function. The second loss represents the error between the identifier information in the output information generated by the intent splitting module and the sample recursive identifier in the sample output information.

[0111] The second loss can be calculated using methods such as Cross-Entropy Loss or Mean Squared Error (MSE) to measure the accuracy of the model in generating recursive labels.

[0112] In operation S820, according to the second loss, the parameters of the intention splitting module are adjusted until the second loss is less than or equal to the loss threshold. When the second loss is less than or equal to the loss threshold, it means that each identification information in each output information generated by the intention splitting module according to the sample target information is the same as each sample recursive identification in the sample output information. Unlike the first loss, the loss threshold of the second loss needs to be strictly set so that the generated identification information is exactly the same as the sample recursive identification, and it can be determined that the intention splitting module has been able to generate identification information well.

[0113] According to an embodiment of the present disclosure, the multiple sample output information includes at least the target sample output information, and among the multiple sample sub-information included in the target sample output information, at least part of the intents represented by at least two sample sub-information have a logical relationship. The target sample output information includes multiple sample sub-information, wherein the intents represented by at least two sample sub-information have a specific connection in semantics, execution order or dependency, so as to ensure that the model can not only learn how to disassemble intents during training, but also learn the logical relationship between different intents, so that it can output output information in a logical order during the reasoning stage, rather than just disassembling according to a simple segmentation strategy. Accordingly, when calculating the first loss, this logical relationship needs to be taken into consideration.

[0114] In combination with the foregoing embodiment, when the logical relationship is reflected by the order of arrangement of sub-information in the entire information, in the sample output information in the target sample output information, the logically earlier sample sub-information is arranged before the logically later sample sub-information. When the model calculates the first loss, the order of arrangement of each output sub-information and the order of arrangement of each sample sub-information in the sample output information need to be taken into consideration in the calculation.

[0115] In combination with the foregoing embodiments, when the intention flag is used to represent the logical relationship of the sub-information in the entire information, for example, when at least each sub-information is represented by a numerical identifier to reflect the logical relationship, when the model calculates the first loss, it is necessary to include the numerical value of each intention flag in the output information and the numerical value of the intention flag corresponding to each sample sub-information in the sample output information into the calculation.

[0116] Fig. 9 The block diagram schematically shows an information processing device according to an embodiment of the present disclosure.

[0117] like Fig. 9 As shown, the information processing device 900 may include a first recognition module 910 and an input module 920 .

[0118] The first recognition module 910 is used to, in response to the input of the target information, perform intent recognition on the part of the target information that has not been subjected to intent recognition, and obtain first sub-information, where the first sub-information represents at least one intent. In some embodiments, the first recognition module 910 can be used to perform operation S110 in the above information processing method.

[0119] The input module 920 is used to add an intention identifier for indicating the first sub-information to the target information, and input the target information again. In some embodiments, the input module 920 can be used to perform operation S120 in the above information processing method.

[0120] According to an embodiment of the present disclosure, the first identification module may include a first acquisition module, a first determination module and a first generation module.

[0121] The first acquisition module is used to acquire the identified information in response to at least one intent identifier in the target information, where the identified information is at least one second sub-information indicated by at least one intent identifier in the target information. In some embodiments, the first acquisition module can be used to perform operation S210 in the above information processing method.

[0122] The first determination module is used to identify the intent of the partial information that has not been identified according to the identified information, and determine the first sub-information from the partial information, wherein at least part of the intent represented by the first sub-information has a logical relationship with at least part of the intent represented by the identified information. In some embodiments, the first determination module can be used to perform operation S220 in the above-mentioned information processing method.

[0123] The first generating module is used to generate an intention identifier for the first sub-information. In some embodiments, the first generating module can be used to perform operation S230 in the above information processing method.

[0124] According to an embodiment of the present disclosure, the first identification module may include a second determination module and a second generation module.

[0125] The second determination module is used to identify the intent of the target information in response to the absence of the intent identifier in the target information, determine the first sub-information from the target information, and the execution priority of the intent represented by the first sub-information meets the target condition. In some embodiments, the second determination module can be used to perform operation S310 in the above information processing method.

[0126] The second generating module is used to generate an intention identifier for the first sub-information. In some embodiments, the second generating module can be used to perform operation S320 in the above information processing method.

[0127] According to an embodiment of the present disclosure, the target information further includes a recursive identifier, the recursive identifier is a first identification identifier or a second identification identifier, and the first identification module may include a first sub-identification module and / or a second sub-identification module.

[0128] The first sub-recognition module is configured to, in response to the recursive identification being the first identification identifier, perform intent recognition on a portion of the target information that has not been subjected to intent recognition, wherein the first identification identifier at least indicates that intent recognition will continue to be performed on the portion of the target information that has not been subjected to intent recognition. In some embodiments, the first sub-recognition module can be configured to perform operation S410 in the above-mentioned information processing method.

[0129] A second sub-identification module, configured to update the recursive identifier to a second identification identifier in response to the absence of information representing an intention in the part of the information for which intention recognition has not been performed, where the second identification identifier at least represents stopping intention recognition in the part of the information for which intention recognition has not been performed. In some embodiments, the second sub-identification module may be configured to perform the operation S420 in the above information processing method.

[0130] According to an embodiment of the present disclosure, the information processing apparatus may include an adjustment module and a first input module.

[0131] The adjustment module is configured to adjust the position of the first sub-information in the target information such that, in the target information, the first sub-information is located before the part of the information for which intention recognition has not been performed. In some embodiments, the adjustment module may be configured to perform the operation S510 in the above information processing method.

[0132] The first input module is configured to sequentially input the first sub-information indicated by each intention identifier in the target information into the intention extraction module in sequence to obtain a plurality of intentions. In some embodiments, the first input module may be configured to perform the operation S520 in the above information processing method.

[0133] According to an embodiment of the present disclosure, the first identification module may include a third sub-identification module.

[0134] The third sub-identification module is configured to perform intention recognition on the part of the information for which intention recognition has not been performed in the target information based on the intention splitting module to obtain the first sub-information. In some embodiments, the third sub-identification module may be configured to perform the operation S610 in the above information processing method.

[0135] According to an embodiment of the present disclosure, the information processing apparatus may include a training module. The training module may include a second acquisition module, a second identification module, a third determination module, and a second adjustment module.

[0136] The second acquisition module is configured to acquire sample data. Each group of sample data includes sample target information and a plurality of sample output information. The sample output information includes at least one sample sub-information, and the sample sub-information represents at least one intention. In some embodiments, the second acquisition module may be configured to perform the operation S710 in the above information processing method.

[0137] The second recognition module is used to perform multiple intent splitting on the part of the sample target information that has not been identified, based on the intent splitting module, to obtain multiple output information, wherein the output information includes at least one sub-output information, and the sub-output information represents at least one intent identified in the sample target information. In multiple intent splitting, the input information of the latter intent splitting module is the output information generated by the previous intent splitting module. In some embodiments, the second recognition module can be used to perform operation S720 in the above-mentioned information processing method.

[0138] The third determination module is used to determine multiple first losses according to the sub-output information of the multiple output information and the sample sub-information in the sample output information that satisfies the corresponding relationship with each output information. In some embodiments, the third determination module can be used to perform operation S730 in the above information processing method.

[0139] The second adjustment module is used to adjust the parameters of the intention splitting module according to the first loss until the first loss is less than or equal to the loss threshold. In some embodiments, the second adjustment module can be used to perform operation S740 in the above information processing method.

[0140] According to an embodiment of the present disclosure, the sample output information also includes a sample recursive identifier, and the output information also includes identifier information, which indicates whether to perform intent recognition on part of the information that has not been used for intent recognition. The training module may include a fourth determination module and a third adjustment module.

[0141] The fourth determination module is used to determine multiple second losses according to the identification information in the multiple output information and the sample recursive identification in the sample output information that satisfies the corresponding relationship with each output information. In some embodiments, the fourth determination module can be used to perform operation S810 in the above information processing method.

[0142] The third adjustment module is used to adjust the parameters of the intention splitting module according to the second loss until the second loss is less than or equal to the loss threshold. In some embodiments, the third adjustment module can be used to perform operation S820 in the above information processing method.

[0143] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.

[0144] For example, any multiple of the first identification module 910 and the input module 920 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the first identification module 910 and the input module 920 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in any appropriate combination of any of them. Alternatively, at least one of the first identification module 910 and the input module 920 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed.

[0145] It should be noted that the data processing system part in the embodiments of the present disclosure corresponds to the data processing method part in the embodiments of the present disclosure. The description of the data processing system part specifically refers to the data processing method part, which will not be repeated here.

[0146] Fig.10 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. Fig.10 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0147] like Fig.10 As shown, the electronic device 1000 according to an embodiment of the present disclosure includes a processor 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage part 1008 to a random access memory (RAM) 1003. The processor 1001 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 1001 may also include an onboard memory for caching purposes. The processor 1001 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0148] In RAM 1003, various programs and data required for the operation of electronic device 1000 are stored. Processor 1001, ROM 1002 and RAM 1003 are connected to each other via bus 1004. Processor 1001 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 1002 and / or RAM 1003. It should be noted that the program can also be stored in one or more memories other than ROM 1002 and RAM 1003. Processor 1001 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0149] According to an embodiment of the present disclosure, the electronic device 1000 may further include an input / output (I / O) interface 1005, which is also connected to the bus 1004. The electronic device 1000 may further include one or more of the following components connected to the input / output (I / O) interface 1005: an input portion 1006 including a keyboard, a mouse, etc.; an output portion 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 1008 including a hard disk, etc.; and a communication portion 1009 including a network interface card such as a LAN card, a modem, etc. The communication portion 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the input / output (I / O) interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage portion 1008 as needed.

[0150] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1009, and / or installed from the removable medium 1011. When the computer program is executed by the processor 1001, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0151] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0152] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. For example, it may include, but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0153] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 1002 and / or the RAM 1003 described above and / or one or more memories other than the ROM 1002 and the RAM 1003 .

[0154] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method provided by the embodiment of the present disclosure. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the control method provided by the embodiment of the present disclosure.

[0155] When the computer program is executed by the processor 1001, the above functions defined in the system / device of the embodiment of the present disclosure are executed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0156] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 1009, and / or installed from the removable medium 1011. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0157] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0158] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments of the present disclosure can be combined and / or combined in a variety of ways, even if such a combination or combination is not explicitly recorded in the present disclosure. In particular, without departing from the spirit and teaching of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0159] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.< / n> < / n> < / n> < / y>

Claims

1. An information processing method, comprising: In response to input of target information, performing intent recognition on a portion of the target information that has not been subjected to intent recognition to obtain first sub-information, where the first sub-information represents at least one intent; An intention identifier for indicating the first sub-information is added to the target information, and the target information is input again.

2. According to the method of claim 1, the step of performing intent recognition on a portion of the target information that has not been subjected to intent recognition to obtain the first sub-information comprises: In response to at least one of the intention identifiers being present in the target information, acquiring identified information, where the identified information is at least one second sub-information indicated by at least one of the intention identifiers in the target information; According to the identified information, perform intent recognition on the partial information that has not been subjected to intent recognition, and determine first sub-information from the partial information, wherein at least a portion of the intent represented by the first sub-information has a logical relationship with at least a portion of the intent represented by the identified information; The intent identifier is generated for the first sub-information.

3. According to the method of claim 1, the step of performing intent recognition on a portion of the target information that has not been subjected to intent recognition to obtain the first sub-information comprises: In response to the absence of the intention identifier in the target information, performing intention recognition on the target information, determining first sub-information from the target information, wherein the execution priority of the intention represented by the first sub-information meets the target condition; The intent identifier is generated for the first sub-information.

4. According to the method of claim 1, the target information further includes a recursive identifier, the recursive identifier is a first identification identifier or a second identification identifier, and the performing intent identification on the part of the target information that has not been subjected to intent identification comprises: In response to the recursive identifier being the first identification identifier, performing intent identification on a portion of the target information that has not been subjected to intent identification; The first identification mark at least indicates that the intention recognition is continued for part of the information for which the intention recognition has not been performed.

5. The method according to claim 4, further comprising: In response to the absence of information representing the intent in the partial information for which intent recognition has not been performed, updating the recursive identifier to a second identification identifier; The second identification mark at least indicates stopping the intention recognition for the part of information for which the intention recognition has not been performed.

6. The method according to claim 1, after adding an intention identifier for indicating the first sub-information to the target information, the method further comprises: Adjusting the position of the first sub-information in the target information so that, in the target information, the first sub-information is located before the portion of information for which intent recognition has not been performed; The method further comprises: The first sub-information indicated by each intention identifier in the target information is input into the intention extraction module in sequence to obtain multiple intentions.

7. According to the method of claim 1, the step of performing intent recognition on a portion of the target information that has not been subjected to intent recognition to obtain the first sub-information comprises: Based on the intention splitting module, intention recognition is performed on part of the target information that has not been subjected to intention recognition to obtain first sub-information.

8. According to the method of claim 7, the process of obtaining the intention splitting module comprises: Acquire sample data, each group of the sample data includes sample target information and a plurality of sample output information, the sample output information includes at least one sample sub-information, and the sample sub-information represents at least one intention; Based on the intention splitting module, multiple intention splittings are performed on the part of the sample target information that has not been subjected to intention recognition, to obtain multiple output information, wherein the output information includes at least one sub-output information, and the sub-output information represents at least one intention recognized in the sample target information. In the multiple intention splittings, the input information of the latter intention splitting module is the output information generated by the previous intention splitting module; Determine a plurality of first losses according to the plurality of sub-output information of the output information and the sample sub-information in the sample output information that satisfies the corresponding relationship with each of the output information; According to the first loss, parameters of the intention splitting module are adjusted until the first loss is less than or equal to a loss threshold.

9. The method according to claim 8, wherein the sample output information further includes a sample recursive identifier, and the output information further includes identifier information, wherein the identifier information indicates whether to continue to perform intent recognition on the portion of information for which intent recognition has not been performed; The method further comprises: Determine a plurality of second losses according to identification information in the plurality of output information and sample recursive identifications in the sample output information that satisfy a corresponding relationship with each of the output information; According to the second loss, adjust the parameters of the intention splitting module until the second loss is less than or equal to a loss threshold.

10. An electronic device comprising: at least one processor; and a memory coupled to the at least one processor; The memory stores instructions executable by at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the following operations: in response to input of target information, perform intent recognition on part of the target information that has not been subjected to intent recognition, to obtain first sub-information, wherein the first sub-information represents at least one intent; An intention identifier for indicating the first sub-information is added to the target information, and the target information is input again.

Images