User intention recognition method, interaction method, device, system, equipment and medium

Abstract

The disclosed embodiments relate to a method for identifying user intent, an interaction method, an apparatus, a system, a device, and a medium, wherein the method for identifying user intent may include: obtaining a user request sentence to be identified; obtaining at least one user intent sample that matches the user request sentence according to a text similarity algorithm; and determining the user intent implied in the user request sentence from at least one user intent sample according to a sentence implication model. The technical solution provided by the disclosed embodiments does not require the construction of a large number of user sentence samples and the annotation of a large number of user intents during the process of identifying user intent, and the cost of identifying user intent is low. At the same time, the solution has high generalization and can be applied to any intelligent interaction scenario that requires identifying user intent.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a user intention recognition method, interaction method, apparatus, system, device, and medium. Background Art

[0002] As a new interaction mode, intelligent voice interaction is widely used in modern life, such as smart navigation, smart home, smart education, and smart reception. Intelligent voice interaction provides great convenience for people.

[0003] Currently, in voice interaction scenarios, user intent is typically determined within a specific corpus using whitelist matching or regular expression matching of user sentences. However, this approach is labor-intensive and lacks generalizability. Alternatively, a large number of sentence samples can be collected and manually labeled with user intent, which can then be used to train a model capable of identifying user intent. Once a user sentence is captured, user intent can be directly determined. However, this approach requires extensive pre-processing of user intent annotation, and the accuracy of the annotation results directly impacts the accuracy of the model training results, specifically the accuracy of the user intent output by the model.

[0004] Therefore, how to reduce the cost of identifying user intentions, improve the generalization of user intention identification methods, and at the same time ensure the accuracy of user intention identification is still a problem that needs to be solved. Summary of the Invention

[0005] In order to solve the above technical problems or at least partially solve the above technical problems, the embodiments of the present disclosure provide a user intent recognition method, interaction method, apparatus, system, device and medium.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for identifying user intent, comprising:

[0007] Obtain the user request statement to be identified;

[0008] Acquire at least one user intent sample that matches the user request statement according to a text similarity algorithm;

[0009] The user intent implied by the user request sentence is determined from the at least one user intent sample according to the sentence implication model.

[0010] In a second aspect, the embodiments of the present disclosure further provide an interaction method based on user intent recognition, which is applied to a user interaction device, including:

[0011] Obtaining a user request sentence input by a user, and sending the user request sentence to an intent analysis device, wherein the intent analysis device is configured to obtain at least one user intent sample matching the user request sentence based on a text similarity algorithm, and determine the user intent implied by the user request sentence from the at least one user intent sample based on a sentence implication model;

[0012] Output interactive content corresponding to the user's intention to the user.

[0013] In a third aspect, an embodiment of the present disclosure further provides an interaction method based on user intent recognition, which is applied to an intent analysis device, including:

[0014] Obtaining a user request statement sent by a user interaction device;

[0015] Acquire at least one user intent sample that matches the user request statement according to a text similarity algorithm;

[0016] Determining the user intent implied by the user request sentence from the at least one user intent sample according to the sentence implication model;

[0017] sending the user intention to the user interaction device, so that the user interaction device determines interaction content corresponding to the user intention and outputs the interaction content to the user; or

[0018] Determine interaction content corresponding to the user intention, and send the interaction content to the user interaction device, so that the user interaction device outputs the interaction content to the user.

[0019] In a fourth aspect, an embodiment of the present disclosure further provides a user intention recognition device, comprising:

[0020] A user request statement acquisition module is used to acquire the user request statement to be identified;

[0021] A user intention sample determination module is used to obtain at least one user intention sample that matches the user request statement according to a text similarity algorithm;

[0022] A user intention determination module is used to determine the user intention implied in the user request sentence from the at least one user intention sample according to a sentence implication model.

[0023] In a fifth aspect, an embodiment of the present disclosure further provides an interaction apparatus based on user intent recognition, which is configured on a user interaction device and includes:

[0024] A user request sentence acquisition module is used to acquire a user request sentence input by a user and send the user request sentence to an intention analysis device. The intention analysis device is used to acquire at least one user intention sample that matches the user request sentence based on a text similarity algorithm, and determine the user intention implied by the user request sentence from the at least one user intention sample based on a sentence implication model.

[0025] The interactive content output module is used to output interactive content corresponding to the user's intention to the user.

[0026] In a sixth aspect, an embodiment of the present disclosure further provides an interactive device based on user intent recognition, configured in an intent analysis device, comprising:

[0027] A user request statement acquisition module is used to acquire the user request statement sent by the user interaction device;

[0028] A user intention sample determination module is used to obtain at least one user intention sample that matches the user request statement according to a text similarity algorithm;

[0029] A user intention recognition module is configured to determine the user intention implied in the user request sentence from the at least one user intention sample according to a sentence implication model;

[0030] A user intention sending module is configured to send the user intention to the user interaction device, so that the user interaction device determines interaction content corresponding to the user intention and outputs the interaction content to the user; or

[0031] The interactive content sending module is configured to determine interactive content corresponding to the user intention and send the interactive content to the user interactive device so that the user interactive device outputs the interactive content to the user.

[0032] In a seventh aspect, the embodiments of the present disclosure further provide an interactive system based on user intent recognition, comprising:

[0033] A user interaction device is used to obtain a user request statement input by a user and send the user request statement to the intention analysis device;

[0034] The intention analysis device is configured to obtain at least one user intention sample matching the user request sentence according to a text similarity algorithm, and determine the user intention implied by the user request sentence from the at least one user intention sample according to a sentence implication model;

[0035] The intention analysis device is further configured to send the user intention to the user interaction device, so that the user interaction device determines interaction content corresponding to the user intention and outputs the interaction content to the user; or

[0036] The intention analysis device is further used to determine the interaction content corresponding to the user intention and send the interaction content to the user interaction device so that the user interaction device outputs the interaction content to the user.

[0037] In the eighth aspect, an embodiment of the present disclosure further provides an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes any user intention recognition method provided by the embodiment of the present disclosure, or executes any interaction method based on user intention recognition provided by the embodiment of the present disclosure.

[0038] In the ninth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the processor executes any user intent recognition method provided by an embodiment of the present disclosure, or executes any interaction method based on user intent recognition provided by an embodiment of the present disclosure.

[0039] In the tenth aspect, an embodiment of the present disclosure also provides a computer program product, which includes: a computer program, which is stored in a computer-readable storage medium, and when the computer program is read and executed by a processor, enables the processor to execute any user intention recognition method provided by the embodiment of the present disclosure, or execute any interaction method based on user intention recognition provided by the embodiment of the present disclosure.

[0040] Compared with the prior art, the technical solution provided by the embodiments of the present disclosure has at least the following advantages: in the embodiments of the present disclosure, after obtaining the user request sentence, a text similarity algorithm is first used to obtain at least one user intent sample that matches the user request sentence, which provides a matching intent selection range for the subsequent accurate determination of the user intent, and also helps to reduce the amount of calculation for the subsequent determination of the user intent using the sentence implication model. Then, the sentence implication model is used to determine the user intent implied in the user request sentence from at least one user intent sample, thereby ensuring the accuracy of the determination of the user intent. In addition, during the implementation of the entire solution, there is no need to construct a large number of user sentence samples and perform a large amount of user intent annotation, and the cost of identifying the user intent is low. At the same time, the technical solution of the embodiments of the present disclosure has high generalization and can be applied to any intelligent interaction scenario that requires identifying user intent. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0042] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0043] Figure 1 is a flow chart of a method for identifying user intent provided by an embodiment of the present disclosure;

[0044] Figure 2 Schematic diagram of a processing architecture for determining user intent based on the ALBERT-NLI model provided by an embodiment of the present disclosure;

[0045] Figure 3 is a flowchart of another method for identifying user intent provided by an embodiment of the present disclosure;

[0046] Figure 4 This is a schematic diagram of a user intent recognition architecture provided by an embodiment of the present disclosure;

[0047] Figure 5 This is a schematic diagram of user intention recognition for a travel navigation scenario provided by an embodiment of the present disclosure;

[0048] Figure 6 is a flow chart of an interaction method based on user intent recognition provided by an embodiment of the present disclosure;

[0049] Figure 7 is a flowchart of another interaction method based on user intent recognition provided by an embodiment of the present disclosure;

[0050] Figure 8 is a structural diagram of a user intention recognition device provided by an embodiment of the present disclosure;

[0051] Figure 9 is a structural diagram of an interactive device based on user intent recognition provided by an embodiment of the present disclosure;

[0052] Figure 10 is a structural diagram of another interactive device based on user intent recognition provided by an embodiment of the present disclosure;

[0053] Figure 11 is a structural diagram of an interactive system based on user intent recognition provided by an embodiment of the present disclosure;

[0054] Figure 12 It is a structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0055] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features therein can be combined with each other in the absence of conflict.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0057] Figure 1 This is a flowchart of a user intent recognition method provided by an embodiment of the present disclosure, which can be applied to the situation of how to recognize user intent based on a user request statement. The user intent recognition method can be executed by a user intent recognition device, which can be implemented in software and / or hardware and can be integrated on any electronic device with computing capabilities, such as a terminal device and / or a server device, that is, the user intent recognition method provided by an embodiment of the present disclosure can be implemented on a terminal device and / or a server device. The terminal device may include, but is not limited to: a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device, or a smart speaker.

[0058] like Figure 1 As shown, the user intention recognition method provided by the embodiment of the present disclosure may include:

[0059] S101: Obtain a user request statement to be identified.

[0060] Exemplarily, the user's voice data can be collected by a voice collection device (such as a microphone, etc.) on the terminal device, and then the user's voice data can be recognized using voice recognition technology to obtain the user's request statement (query), wherein the voice recognition technology can be implemented with reference to the existing technology. Alternatively, based on the user's interaction needs, the terminal device can be used to display an online interaction page to the user, and then the user's request statement can be determined based on the user's text input operation on the online interaction page; or some guide words can be displayed on the online interaction page, and the user enters the request statement according to the guidance of the guide words. Of course, for the interaction scenario between the terminal device and the server device, the user's voice data can be collected by a voice collection device on the terminal device, or the user's input request statement can be obtained by using the online interaction page displayed by the terminal device and then sent to the server device. That is, in the embodiment of the present disclosure, obtaining the user request statement to be identified may include: obtaining the voice data input by the user, performing voice recognition on the voice data to obtain the user request statement; or obtaining the user request statement entered by the user through the online interaction page.

[0061] S102: Obtain at least one user intent sample that matches the user request statement according to a text similarity algorithm.

[0062] Among them, the text similarity algorithms that can be used include but are not limited to vector-based text retrieval methods, BM25 algorithms, or term frequency-inverse document frequency index algorithms (Term Frequency-Inverse Document Frequency, TF-IDF), etc. The specific implementation principles of the algorithms can refer to existing technologies. Among them, the BM25 algorithm is a classic method for calculating text similarity. Its main implementation principle is: first, the user's request statement (query) is segmented to obtain multiple words, and then the similarity value between each word and the intent sample to be selected is calculated separately. Finally, based on the similarity value corresponding to each word and the weight of each word in the user's request statement, a weighted sum calculation is performed to obtain the similarity value between the user's request statement and the intent sample to be selected.

[0063] Exemplarily, at least one user intent sample that matches the user request statement can be extracted from the intent sample pool based on a vector-based text retrieval method, a BM25 algorithm, or a word frequency-inverse text frequency index algorithm. The intent sample pool is a pre-constructed set of intent samples that can be used to roughly recall user intent based on the user request statement, i.e., determine a matching intent selection range for the user request statement. The intent samples included in the intent sample pool can largely cover the user's possible intentions in different interaction scenarios (although there are diverse user expressions, different user expressions can correspond to the same user intent, i.e., user intent can still be quantified), and support dynamic updates based on the user intent recognition results, such as adding new intent samples to the intent sample pool based on the user intent recognition results or replacing existing intent samples with new intent samples, etc., to ensure that without the need to construct a large number of user intent samples, the limited intent samples included in the intent sample pool can meet the user's intent recognition needs. By using the intent sample pool to determine at least one user intent sample that matches the user request statement, the effect of determining the user intent using limited intent samples can be achieved, thereby reducing the cost of identifying the user intent.

[0064] Optionally, obtaining at least one user intent sample that matches the user request statement according to a text similarity algorithm includes:

[0065] Obtain the similarity value between each user intent sample and the user request statement in the intent sample pool based on the text similarity algorithm;

[0066] A user intention sample whose similarity value reaches a first similarity threshold (or called a first threshold) is obtained as at least one user intention sample that matches the user request statement, wherein the value of the first similarity threshold can be flexibly set according to actual needs, for example, according to the recall rate and recognition accuracy of the user intention, and the embodiment of the present disclosure does not make specific limitations.

[0067] Optionally, the user intent recognition method provided by the embodiment of the present disclosure may further include:

[0068] A user intent sample whose similarity value reaches a second similarity threshold is obtained as the user intent implied by the user request sentence.

[0069] Among them, the value of the second similarity threshold is greater than the first similarity threshold, and the specific value can also be set flexibly. When the similarity value between a user intent sample and a user request statement reaches the second similarity threshold, it can be considered that the user request statement and the user intent sample are almost completely matched, and the user intent sample can be directly used as the user intent. At this time, there is no need to perform the operation of determining the user intent using the sentence implication model. For example, the value of the second similarity threshold is set to 0.95, which is greater than the value of the first similarity threshold of 0.8. When there is a user intent sample A and a user request statement with a similarity value of 0.98, the semantics of the user intent sample A can be directly used as the user intent.

[0070] S103. Determine the user intent implied in the user request sentence from at least one user intent sample according to the sentence implication model.

[0071] The sentence implication model has the function of determining the user intention based on the user request sentence. The specific training logic of the model is not limited in the embodiment of this disclosure and can be implemented with reference to the prior art. For example, the user intention implied in the user request sentence can be determined from at least one user intention sample based on one of other available deep learning models such as a convolutional neural network model, a recurrent neural network model, a BERT (Bidirectional Encoder Representations from Transformers) model or an ALBERT-NLI model (an improved model based on the BERT model). Among them, the BERT model is a bidirectional language model. During the model processing, the context information of each word in the sentence to be processed can be used at the same time to obtain accurate semantic processing results; the ALBERT-NLI model refers to a model that can be applied to natural language inference (NLI). It is a lightweight BERT model with fewer model parameters than the BERT model and better model performance than the BERT model.

[0072] Optionally, determining the user intent implied by the user request sentence from at least one user intent sample according to the sentence implication model includes:

[0073] Inputting a user request sentence and at least one user intent sample into a sentence entailment model to obtain an entailment value corresponding to each user intent sample;

[0074] The user intent samples whose entailment values ​​reach the entailment threshold are determined as the user intent implied by the user request sentence.

[0075] Among them, the implication value is used to represent the intention matching degree between the user intent sample and the user request sentence; the implication threshold can be flexibly set according to the user intent recognition accuracy. For example, by reasonably selecting the implication threshold, the user intent sample with the largest implication value can be determined as the user intent implied by the user request sentence.

[0076] Figure 2 This is a schematic diagram of a processing architecture for determining user intent based on the ALBERT-NLI model provided by an embodiment of the present disclosure, which is used to exemplify the embodiment of the present disclosure and should not be understood as a specific limitation of the embodiment of the present disclosure. Figure 2 As shown in the figure, the user request sentence and each user intent sample are concatenated as the input of the ALBERT-NLI model, wherein a classifier [CLS] is added at the beginning of the user request sentence, indicating that the representation vector obtained by the ALBERT-NLI model can be used for subsequent classification tasks, and a separator [SEP] is added between the user request sentence and the user intent sample, and a separator [SEP] is also added at the end of the user intent sample; finally, after passing through the classification layer, for example, based on the softmax function, it is output whether each user intent sample is the user intent corresponding to the user request sentence.

[0077] In the embodiment of the present disclosure, at least one user intent sample that matches the user request sentence is first determined to achieve a rough recall of the user intent, and then the user intent is accurately determined using a sentence implication model. This reduces the amount of computation in the process of identifying the user intent and lowers the cost of identifying the user intent, while also ensuring the accuracy of identifying the user intent.

[0078] Optionally, the user intent recognition method provided by the embodiment of the present disclosure may further include:

[0079] If no user intent sample with an entailment value that meets the entailment threshold exists, a message is provided to the user stating that the user intent implied by the user's request cannot be determined. Furthermore, guidance information can be provided to the user to encourage them to re-enter the request. Furthermore, at least one user intent sample can be selected based on the entailment value as an estimated intent for the user's request, and this information is provided to the user for selection and confirmation, thereby improving interaction friendliness.

[0080] In an embodiment of the present disclosure, after obtaining a user request sentence, a text similarity algorithm is first used to obtain at least one user intent sample that matches the user request sentence, which provides a matching intent selection range for subsequent accurate determination of the user intent, and also helps to reduce the amount of computation required for subsequent determination of the user intent using a sentence implication model. Then, the sentence implication model is used to determine the user intent implied in the user request sentence from at least one user intent sample, thereby ensuring the accuracy of determining the user intent. In addition, during the implementation of the entire solution, there is no need to construct a large number of user sentence samples and perform a large amount of user intent annotation, and the cost of identifying user intent is low. At the same time, the technical solution of the embodiment of the present disclosure has high generalization and can be applied to any intelligent interaction scenario that requires identifying user intent.

[0081] Figure 3 This is a flow chart of another method for identifying user intentions provided by an embodiment of the present disclosure, which is further optimized and expanded based on the above technical solution and can be combined with the above optional implementations. Figure 3 As shown, the user intention recognition method provided by the embodiment of the present disclosure may include:

[0082] S201. Build an intent sample pool.

[0083] In the disclosed embodiments, the pre-built intent sample pool includes a limited number of user intent samples that largely cover different interaction scenarios. Although different users express their intent in diverse ways, different expressions can correspond to the same user intent, so user intent can still be quantified.

[0084] The intent sample pool may include positive sample sets and negative sample sets. The positive sample set may include an operational intent sample set, and the negative sample set may include at least one of an intra-domain intent sample set, an automatically constructed obfuscated sample set, and a manually constructed obfuscated sample set. The intent sample pool may include multiple operational intent sample sets, each of which may correspond to different interaction scenarios, such as an operational intent sample set A for an online route navigation scenario, an operational intent sample set B for an online customer service question consultation scenario (specifically, an order inquiry scenario, etc.), etc., which may be determined based on the required interactive service functions. An intra-domain intent sample set refers to a sample set that corresponds to a different interaction scenario from the operational intent sample set, but has certain similarities in the corresponding interaction scenarios. For example, for navigation services, the operational intent sample set in the intent sample pool may be a sample set for the online route navigation scenario in the navigation service, while the intra-domain intent sample set may be a sample set for other functional scenarios in the navigation service other than the online route navigation scenario, such as a sample set for the scenario of online search for nearby points of interest, etc. In other words, the operational intent sample set and the intra-domain intent sample set correspond to different sub-scenarios under the large category of travel navigation. The domain intent sample set includes sentences whose hit frequency in the corresponding interaction scenario exceeds the frequency threshold (the value can be flexibly set) and whose sentence quality meets the sample screening criteria. The sample screening criteria may include but are not limited to sentence structure conforming to standard sentence structure and clear meaning. The operational intent sample set and the domain intent sample set can be manually constructed.

[0085] For example, at the interactive service level, intent sample pools can be categorized and constructed for different interaction scenarios. For example, they can include any one of the following: travel navigation operation intent sample pool, knowledge question and answer operation intent sample pool, question search operation intent sample pool, guidance service operation intent sample pool, multimedia data playback operation intent sample pool, etc. Correspondingly, the operation intent sample set can include an operation intent sample set for travel navigation, an operation intent sample set for smart car interaction, an operation intent sample set for knowledge question and answer, an operation intent sample set for question search, an operation intent sample set for guidance service, and an operation intent sample set for multimedia data playback, thereby laying the foundation for accurately determining user intent in each scenario. The intent sample pool for each interaction scenario can be constructed by statistically analyzing the user demand statements in that interaction scenario.

[0086] An automatically constructed or manually constructed obfuscated sample set is a sample set consisting of similar intent samples to each intent sample in the operational intent sample set. For example, a text similarity algorithm, such as the BM25 algorithm, can be used to determine similar intent samples to each intent sample in the operational intent sample set. For example, domain intent samples whose similarity values ​​to intent samples in the operational intent sample set exceed a screening threshold can be identified as similar intent samples. Based on these similar intent samples, an automatically constructed obfuscated sample set can then be obtained. Of course, a manually constructed obfuscated sample set can also be obtained based on these similar intent samples by manually writing a sample set.

[0087] That is, in the embodiment of the present disclosure, for each operational intent sample set, the intent sample pool will contain at least one of an in-domain intent sample set, an automatically constructed confusion sample set, and a manually constructed confusion sample set that is associated with the operational intent sample set. The number of intent samples included in each sample set can be adaptively determined, and the embodiment of the present disclosure does not limit this. By pre-constructing an intent sample pool including a positive sample set and a negative sample set, the recall rate of the user intent sample can be ensured and the false recall rate can be reduced in the process of determining the user intent using a limited or small number of intent samples.

[0088] S202: Obtain a user request statement to be identified.

[0089] S203. Extract at least one user intent sample that matches the user request statement from the intent sample pool according to a text similarity algorithm.

[0090] Exemplarily, at least one user intent sample matching the user request statement can be extracted from the intent sample pool according to a vector-based text retrieval method, a BM25 algorithm, or a word frequency-inverse text frequency index algorithm.

[0091] Exemplarily, after obtaining the user request statement to be identified, based on the current interaction scenario (for example, the user enters the user request statement through the path search box in the navigation application on the terminal device, then the current interaction scenario can be determined to be an online path navigation scenario, or the current interaction scenario can be determined based on the keywords included in the user request statement), the operational intention sample set corresponding to the current interaction scenario in the intent sample pool, as well as the domain intent sample set associated with the operational intention sample set, the automatically constructed confusion sample set and the manually constructed confusion sample set, at least one user intent sample that matches the user request statement can be determined according to the text similarity algorithm.

[0092] S204: Determine the user intent implied in the user request sentence from at least one user intent sample according to the sentence implication model.

[0093] The sentence implication model can accurately analyze the semantics of user intent samples and user request sentences. It uses the sentence implication model to further screen at the semantic level at least one user intent sample of rough recall to determine the user intent implied in the user request sentence, ensuring the accuracy of determining user intent.

[0094] Optionally, the user intent identification method provided in the embodiments of the present disclosure further includes: adding the user request statement as a user intent sample to the intent sample pool. For example, if the user intent implied by the user request statement is successfully determined, it indicates to some extent that the user request statement is a normative expression. In this case, the user request statement can be added as a user intent sample to the intent sample pool to continuously expand the intent sample pool.

[0095] For example, for any of the operational intention sample sets for travel navigation, the operational intention sample set for smart car interaction, or the operational intention sample set for knowledge question and answer, if the user intention can be accurately analyzed by processing a user request statement in each interaction scenario, the user request statement can be added accordingly to the operational intention sample set corresponding to the interaction scenario, thereby realizing dynamic expansion of the intent sample pool.

[0096] Figure 4 This is a user intention recognition architecture diagram provided by the embodiment of the present disclosure, which is used to illustrate the embodiment of the present disclosure, but should not be understood as a specific limitation of the embodiment of the present disclosure. Figure 4 As shown in the figure, a user request sentence is obtained, and then at least one user intent sample matching the user request sentence is extracted from the intent sample pool according to the text similarity algorithm, i.e., the rough recall of the user intent sample; then, the sentence implication model is used to output the implication value corresponding to the at least one user intent sample, and the at least one user intent sample can be sorted from high to low according to the implication value. The sorting result is shown in the figure. Figure 4 The N top-ranked intent samples (or candidate intent samples) shown in the figure can finally determine the user intent sample corresponding to the maximum implication value as the user intent. In addition, if the user request statement completely matches a user intent sample in the intent sample pool, for example, if the similarity between the user request statement and a user intent sample in the intent sample pool reaches a second similarity threshold, then the user intent sample can be directly determined as the user intent implied by the user request statement.

[0097] Figure 5 This is a schematic diagram of a user intention recognition for a travel navigation scenario provided by an embodiment of the present disclosure, which is used to exemplify the embodiment of the present disclosure. Figure 5As shown in FIG, in the travel navigation scenario, the request sentence input by the user is “I want to team up with my family”. According to the text similarity algorithm, multiple user intent samples roughly recalled from the intent sample pool may include “family map|B, I want to team up|A, team up together|A, team talkback|A, team up immediately|A”. Then, the sentence implication model is used to determine the implication values ​​corresponding to the multiple user intent samples, and the multiple user intent samples are sorted from high to low according to the implication values. The sorting result is “I want to team up|A, team up together|A, family map|B, team up immediately|A, team talkback|A”, that is, “I want to team up|A” has the highest implication value, so “I want to team up|A” can be determined as the user intent implied by the user request sentence (that is, Figure 5 Entering A operational intention as shown in ). Figure 5 The characters A, B or O carried in the user intention sample are used to indicate the sample set to which the user intention sample belongs in the intent sample pool. For example, character A can represent an operational intention sample set corresponding to the current travel navigation scenario.

[0098] Continue as Figure 5 As shown, if in the travel navigation scenario, the request sentence input by the user is "I want to go home", the multiple user intent samples roughly recalled from the intent sample pool according to the text similarity algorithm may include "I want to form a team|A, I want to go home|O, form a team together|A, team talkback|A, form a team immediately|A", and then use the sentence implication model to determine the implication values ​​corresponding to the multiple user intent samples, and sort the multiple user intent samples from high to low according to the implication values. The sorting result is "I want to go home|O, I want to form a team|A, form a team together|A, team talkback|A, form a team immediately|A", that is, "I want to go home|O" has the highest implication value, and "I want to go home|O" can be determined as the user intent implied by the user request sentence. Since "I want to go home|O" does not belong to the intent samples in the operational intent sample set corresponding to the current travel navigation scenario, Figure 5 The operation intention of not entering "I want to form a team|A" is shown in the table to compare with the intention recognition result of "I want to form a team with my family". Figure 5 In the example, if the request sentence input by the user is "I want to team up with my family", after the user's intention is confirmed, the interactive information about teaming with other vehicles can be fed back to the user, such as information about other vehicles participating in the team; if the request sentence input by the user is "I want to go home", after the user's intention is confirmed, the path navigation information for the user's place of residence can be fed back to the user as the interactive content corresponding to the user's intention.

[0099] Figure 6This is a flowchart of an interaction method based on user intent recognition provided by an embodiment of the present disclosure. This interaction method can be applied to a user interaction device and executed by an interaction device based on user intent recognition configured on the user interaction device. This device can be implemented using software and / or hardware. The user interaction device can be any terminal device with interaction capabilities.

[0100] like Figure 6 As shown, the interaction method based on user intent recognition provided by the embodiment of the present disclosure may include:

[0101] S301. Obtain a user request sentence input by the user, and send the user request sentence to an intent analysis device. The intent analysis device is used to obtain at least one user intent sample that matches the user request sentence based on a text similarity algorithm, and determine the user intent implied in the user request sentence from the at least one user intent sample based on a sentence implication model.

[0102] Optionally, obtain the user request statement entered by the user, including:

[0103] The user's voice data is collected through the voice module, and the voice data is subjected to voice recognition to obtain the user request statement; or the user request statement input by the user is obtained through the online interactive page.

[0104] For example, a user's voice data can be collected by a voice collection device (e.g., a microphone) on a user interaction device, and then recognized using voice recognition technology to obtain the user's request statement. The voice recognition technology can be implemented with reference to existing technologies. Alternatively, based on the user's interaction needs, an online interaction page can be displayed to the user using the user interaction device, and the user's request statement can be determined based on the user's text input operation on the online interaction page; or some guiding words can be displayed on the online interaction page, and the user can enter the request statement according to the guidance of the guiding words.

[0105] Regarding how the intent analysis device determines the user intent based on the user request statement, you can refer to the description of the user intent identification method provided in the embodiment of the present disclosure, and will not repeat it here.

[0106] S302: Output interactive content corresponding to the user's intention to the user.

[0107] For different user intents, the user interaction device outputs different interactive content to the user. This interactive content can be determined by the user interaction device based on the user intent fed back by the intent analysis device, or it can be determined by the intent analysis device and then sent to the user interaction device. That is, in the embodiments of the present disclosure, outputting interactive content corresponding to the user intent to the user may include: obtaining interactive content corresponding to the user intent sent by the intent analysis device and outputting it to the user; or, based on the user intent sent by the intent analysis device, determining interactive content corresponding to the user intent and outputting it to the user.

[0108] Furthermore, the user interaction device may receive interaction content corresponding to the user's intent from a server and output it to the user. The server may receive the user's intent from an intent analysis device and determine the interaction content corresponding to the user's intent. Alternatively, the user interaction device may transmit the user's intent from the intent analysis device to a server, allowing the server to determine the interaction content corresponding to the user's intent. The server may receive the interaction content corresponding to the user's intent from the server and output it to the user. The server may be a pre-deployed server specifically configured to determine interaction content corresponding to the user's intent.

[0109] By using the server to determine the interactive content corresponding to the user's intention, it helps to reduce the data processing volume of the user interaction device or the intention analysis device, and also reduces the performance requirements of the user interaction device, so that this solution can be implemented on a user interaction device of any performance.

[0110] For example, the interactive content corresponding to the user's intent includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information, or multimedia data. Based on the different interactive content, the user interaction device can display the interactive content to the user using audio, images, and other methods.

[0111] Example 1: In a travel navigation scenario, the user inputs the user request statement "go to xx hotel" using the user interaction device. After user intent recognition, it is determined that the user intent is to obtain route navigation. The user interaction device can display the navigation route from the current location to xx hotel to the user.

[0112] Example 2: In a knowledge question and answer scenario, the user request sentence input by the user using the user interaction device is "What is an atom?". After user intent recognition, it is determined that the user intention is to ask for the definition of an atom. The user interaction device can display the text "Atoms are the smallest units that constitute general matter and are composed of atomic nuclei and electrons orbiting the nucleus" to the user, or output the definition of the atom to the user by voice, and can also display a picture of the atom.

[0113] Example three: In a question search scenario, the user uses the user interaction device to input the user request statement "What is the weather like today?". After user intent recognition, it is determined that the user's intention is to inquire about the weather. The user interaction device can display today's weather forecast information to the user. It can also perform voice broadcast or display dynamic images of the weather forecast information.

[0114] Example 4: In a welcoming scenario at a hotel or conference, the user request statement input by the user using the user interaction device is "attend and sign in". After user intent recognition, it is determined that the user intention is to attend and sign in. The user interaction device can display guidance service information such as "Please collect xx items after signing in" to the user, or broadcast the guidance service information to the user by voice.

[0115] Example 5: In a multimedia data playback scenario, the user request statement input by the user using the user interaction device is "I want to listen to a song named xx". After user intent recognition, it is determined that the user intention is to play a song on demand. The user interaction device can play the song named xx to the user, and can also display lyrics information or singer images related to the song.

[0116] Moreover, the user request statements in the above examples can be respectively used as new intent sample statements and added by the intent analysis device to the operational intent sample set for travel navigation, the operational intent sample set for knowledge question and answer, the operational intent sample set for question search, the operational intent sample set for guidance services, or the operational intent sample set for multimedia data playback, to achieve dynamic expansion of the operational intent sample set in different interaction scenarios.

[0117] It should be noted that the above examples are merely illustrative descriptions of the interactive content corresponding to user intent in the embodiments of the present disclosure and should not be construed as specific limitations on the embodiments of the present disclosure. The interactive content for different interactive scenarios can be stored as an interactive content library in an intent analysis device, server, or interactive device. After the user intent is identified by the user interactive device, it can be called from the corresponding interactive content library based on the user intent. Alternatively, a dedicated server can dynamically determine the interactive content based on the user intent.

[0118] Based on the above technical solution, in an optional implementation, the user interaction device and the intent analysis device are the same terminal device. In this case, the user interaction device and the intent analysis device are equivalent to functionally differentiating the same terminal device from the perspective of implementing different functions. Optionally, the terminal device may include a smartphone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device, or a smart speaker.

[0119] like Figure 11As shown, the interactive system 1100 formed by the user interaction device 1101 and the intent analysis device 1102 also refers to the terminal device. The interaction between the user interaction device 1101 and the intent analysis device 1102 is equivalent to the internal interaction of the same terminal device.

[0120] Specifically, for the same terminal device, from the perspective of the user interaction device, the terminal device's interactive function is implemented by an internally integrated functional module with interactive functions, while from the perspective of the intent analysis device, the terminal's intent recognition function is implemented by an internally integrated functional module with intent recognition functions. In other words, the user interaction device can refer to a terminal device that integrates a functional module with interactive functions, and the intent analysis device can refer to a terminal device that integrates a functional module with intent recognition functions.

[0121] Taking a smartphone as an example, a navigation application (Application, App) is installed in the smartphone. The navigation application integrates functional modules with interactive functions and intention recognition functions. After the user activates the navigation application, the functional module with interactive functions provides the user with an application interaction page. The application interaction page displays a request statement input box, which is used to obtain user input from the request statement and then send it to the functional module with intent recognition function to identify the user's intention.

[0122] When the user interaction device and the intent analysis device are the same terminal device, it helps to improve the integration of interactive functions based on user intent recognition, reduces the amount of traffic consumption when the user interaction device and the intent analysis device are used as independent devices, and also helps to improve interaction efficiency.

[0123] Taking a smartphone as an example, the user enters the current request sentence "What's the weather like today?" The smartphone recognizes the user's request sentence, determines that the user's current intention is to ask about today's weather, and obtains today's weather forecast information and displays it to the user as interactive content.

[0124] In an optional embodiment, the user interaction device and the intent analysis device are different devices, the user interaction device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

[0125] Continue as Figure 11 As shown, user interaction device 1101 and intent analysis device 1102 each refer to a different independent device, and network communication can be established between user interaction device 1101 and intent analysis device 1102 to achieve data transmission. Specifically, intent analysis device 1102 can be any server device that has an integrated functional module with intent recognition capabilities.

[0126] When the user interaction device and the intent analysis device are different independent devices, each device implements its own function, and the dedicated intent analysis device identifies the user's intent, which can reduce the data processing pressure of the user interaction device and reduce the performance requirements of the user interaction device, so that the interaction method based on user intent recognition can be implemented on user interaction devices of any performance.

[0127] Taking a smart speaker as an example, when the microphone collects the user's voice input "I want to listen to xx song", the user's voice can be recognized, the user's request statement "I want to listen to xx song" can be determined, and then sent to the intent analysis device (such as the control server of the smart speaker); the intent analysis device recognizes the user's intention and sends the xx song information to the smart speaker; after receiving the xx song information, the smart speaker plays it to the user.

[0128] In the embodiment of the present disclosure, the intent analysis device can accurately identify the user intent implied in the user request statement based on a limited or small number of intent samples. The user interaction device is used to obtain the user request statement and display the interactive content. Through the interaction between the two, the effect of accurately outputting the interactive content corresponding to the user intent to the user can be achieved, and it helps to reduce the cost of identifying the user intent during the interaction process and improve the generalization of user intent identification during the interaction process.

[0129] Figure 7 This is a flowchart of another interactive method based on user intent recognition provided by an embodiment of the present disclosure. This interactive method can be applied to an intent analysis device and can be executed by an interactive device based on user intent recognition configured on the intent analysis device, which can be implemented using software and / or hardware.

[0130] The interaction method applied to the intent analysis device provided in the embodiment of the present disclosure and the interaction method applied to the user interaction device provided in the above embodiment belong to the same inventive concept, and the two are executed in coordination. For the content not explained in detail below, please refer to the description in the above embodiment.

[0131] In the embodiments of the present disclosure, the intent analysis device and the user interaction device can be the same terminal device or different devices, and the embodiments of the present disclosure do not specifically limit this. The following is an exemplary description of the interaction method based on user intent recognition from the perspective of how to implement user intent recognition.

[0132] like Figure 7 As shown, the interaction method based on user intent recognition provided by the embodiment of the present disclosure may include:

[0133] S401: Obtain a user request statement sent by a user interaction device.

[0134] S402: Obtain at least one user intent sample that matches the user request statement according to a text similarity algorithm.

[0135] S403: Determine the user intent implied in the user request sentence from at least one user intent sample according to the sentence implication model.

[0136] S404: Send the user intention to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intention and outputs the interaction content to the user. Or

[0137] S405: Determine interactive content corresponding to the user intention, and send the interactive content to the user interactive device, so that the user interactive device outputs the interactive content to the user.

[0138] Optionally, the interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

[0139] Optionally, determining interaction content corresponding to the user intent and sending the interaction content to the user interaction device includes:

[0140] The user intention is sent to the server, so that the server determines the interaction content corresponding to the user intention and sends the interaction content to the user interaction device.

[0141] In an optional implementation, the user interaction device and the intent analysis device are the same terminal device.

[0142] Optionally, the terminal device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

[0143] In an optional embodiment, the user interaction device and the intent analysis device are different devices, the user interaction device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

[0144] In the embodiment of the present disclosure, the intent analysis device can accurately identify the user intent implied in the user request statement based on a limited or small number of intent samples. The user interaction device is used to obtain the user request statement and display the interactive content. Through the interaction between the two, the effect of accurately outputting the interactive content corresponding to the user intent to the user can be achieved, and it helps to reduce the cost of identifying the user intent during the interaction process and improve the generalization of user intent identification during the interaction process.

[0145] Figure 8This is a structural diagram of a user intention recognition device provided by an embodiment of the present disclosure. The device can be implemented using software and / or hardware and can be integrated on any electronic device with computing capabilities, such as a terminal device and / or server device.

[0146] like Figure 8 As shown, the user intention recognition device 800 provided by the embodiment of the present disclosure may include a user request statement acquisition module 801, a user intention sample determination module 802 and a user intention determination module 803, wherein:

[0147] The user request statement acquisition module 801 is used to acquire the user request statement to be identified;

[0148] A user intention sample determination module 802 is configured to obtain at least one user intention sample that matches the user request statement based on a text similarity algorithm;

[0149] The user intention determination module 803 is used to determine the user intention implied in the user request sentence from at least one user intention sample according to the sentence implication model.

[0150] Optionally, the user request statement acquisition module 801 includes:

[0151] A first acquiring unit is configured to acquire voice data input by a user and perform voice recognition on the voice data to obtain a user request statement; or

[0152] The second acquiring unit is configured to acquire a user request statement input by the user through the online interactive page.

[0153] Optionally, the user intention sample determination module 802 is specifically configured to:

[0154] According to the vector-based text retrieval method, BM25 algorithm or word frequency-inverse text frequency index algorithm, at least one user intent sample matching the user request statement is extracted from the intent sample pool.

[0155] Optionally, the user intention recognition device 800 provided in the embodiment of the present disclosure further includes:

[0156] The intent sample pool construction module is used to construct the intent sample pool, which includes an operational intent sample set, and at least one of a high-frequency and high-quality intent sample set, an automatically constructed high-confusion sample set, and a manually constructed high-confusion sample set.

[0157] Optionally, the user intention recognition device 800 provided in the embodiment of the present disclosure further includes:

[0158] The intent sample adding module is used to add user request sentences as user intent samples to the intent sample pool.

[0159] Optionally, the user intention sample determination module 802 includes:

[0160] A similarity determination unit is used to obtain a similarity value between each user intent sample in the intent sample pool and the user request sentence according to a text similarity algorithm;

[0161] The intention sample determination unit is used to obtain a user intention sample whose similarity value reaches a first similarity threshold as at least one user intention sample matching the user request statement.

[0162] Optionally, the user intention sample determination module 802 further includes:

[0163] The user intention determination unit is configured to obtain a user intention sample whose similarity value reaches a second similarity threshold as the user intention implied in the user request sentence.

[0164] Optionally, the intent sample pool includes any one of a travel navigation operation intent sample pool, a knowledge question and answer operation intent sample pool, a question search operation intent sample pool, a guidance service operation intent sample pool, and a multimedia data playback operation intent sample pool.

[0165] Optionally, the user intention determination module 803 includes:

[0166] an entailment value determining unit, configured to input a user request sentence and at least one user intent sample into a sentence entailment model to obtain an entailment value corresponding to each user intent sample;

[0167] The user intention determination unit is configured to determine a user intention sample whose implication value reaches an implication threshold as the user intention implied by the user request sentence.

[0168] Optionally, the user intention recognition device 800 provided in the embodiment of the present disclosure further includes:

[0169] The message feedback module is used to feedback to the user a message that the user intent implied in the user request sentence cannot be determined if there is no user intent sample whose implication value reaches the implication threshold.

[0170] Optionally, the user intention determination module 803 is specifically configured to:

[0171] Determine the user intent implied in the user request statement from at least one user intent sample based on one of a convolutional neural network model, a recurrent neural network model, a BERT model, or an ALBERT-NLI model.

[0172] Optionally, the user intention recognition device 800 provided in the embodiment of the present disclosure can be integrated into a terminal device and / or a server device for implementation.

[0173] Optionally, the terminal device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

[0174] The user intention recognition device provided in the embodiments of the present disclosure can execute any user intention recognition method provided in the embodiments of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For content not fully described in the embodiments of the device of the present disclosure, reference can be made to the description of any method embodiment of the present disclosure.

[0175] Figure 9 This is a schematic diagram of the structure of an interactive device based on user intention recognition provided by an embodiment of the present disclosure. The interactive device can be configured in a user interaction device and can be implemented using software and / or hardware. Figure 9 As shown, the interactive device 900 based on user intent recognition provided by the embodiment of the present disclosure may include a user request statement acquisition module 901 and an interactive content output module 902, wherein:

[0176] A user request sentence acquisition module 901 is configured to acquire a user request sentence input by a user and send the user request sentence to an intent analysis device. The intent analysis device is configured to acquire at least one user intent sample that matches the user request sentence based on a text similarity algorithm and determine the user intent implied in the user request sentence from the at least one user intent sample based on a sentence implication model.

[0177] The interactive content output module 902 is configured to output interactive content corresponding to the user's intention to the user.

[0178] Optionally, the user request statement acquisition module 901 includes:

[0179] The first acquisition unit is configured to collect user voice data through a voice module and perform voice recognition on the voice data to obtain a user request statement; or

[0180] The second acquiring unit is configured to acquire a user request statement input by the user through the online interactive page.

[0181] Optionally, the interactive content output module 902 includes:

[0182] A first output unit is configured to obtain interactive content corresponding to the user's intention sent by the intention analysis device and output it to the user; or

[0183] The second output unit is used to determine the interaction content corresponding to the user intent based on the user intent sent by the intent analysis device, and output it to the user.

[0184] Optionally, the first output unit is specifically configured to:

[0185] Receive interactive content corresponding to the user's intention sent by the server and output it to the user, where the server is used to receive the user's intention sent by the intention analysis device and determine the interactive content corresponding to the user's intention; or

[0186] The second output unit includes:

[0187] a sending subunit, configured to send the user intent sent by the intent analysis device to the server, so that the server determines the interaction content corresponding to the user intent;

[0188] The receiving subunit is used to receive the interactive content corresponding to the user's intention sent by the server and output it to the user.

[0189] In an optional implementation, the user interaction device and the intent analysis device are the same terminal device.

[0190] Optionally, the terminal device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

[0191] In an optional embodiment, the user interaction device and the intent analysis device are different devices, the user interaction device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

[0192] Optionally, the interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

[0193] The user intent recognition-based interaction device provided in the embodiments of the present disclosure can execute any of the user intent recognition-based interaction methods provided in the embodiments of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For any content not fully described in the embodiments of the device of the present disclosure, reference can be made to the description of any method embodiment of the present disclosure.

[0194] Figure 10 This is a structural diagram of another interactive device based on user intention recognition provided by an embodiment of the present disclosure. The interactive device can be configured in an intention analysis device and can be implemented using software and / or hardware. Figure 10 As shown, the interactive device 1000 based on user intent recognition provided by the embodiment of the present disclosure may include a user request statement acquisition module 1001, a user intent sample determination module 1002, and a user intent recognition module 1003, and may also include at least one of a user intent sending module 1004 or an interactive content sending module 1005, wherein:

[0195] A user request statement acquisition module 1001 is used to acquire a user request statement sent by a user interaction device;

[0196] A user intention sample determination module 1002 is configured to obtain at least one user intention sample that matches the user request statement based on a text similarity algorithm;

[0197] A user intention identification module 1003 is configured to determine the user intention implied in the user request sentence from at least one user intention sample according to a sentence implication model;

[0198] The user intention sending module 1004 is configured to send the user intention to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intention and outputs the interaction content to the user; or

[0199] The interactive content sending module 1005 is configured to determine interactive content corresponding to the user's intention and send the interactive content to the user interactive device so that the user interactive device outputs the interactive content to the user.

[0200] Optionally, the interactive content sending module 1005 is specifically configured to:

[0201] The user intention is sent to the server, so that the server determines the interaction content corresponding to the user intention and sends the interaction content to the user interaction device.

[0202] In an optional implementation, the user interaction device and the intent analysis device are the same terminal device.

[0203] Optionally, the terminal device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

[0204] In an optional embodiment, the user interaction device and the intent analysis device are different devices, the user interaction device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

[0205] Optionally, the interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

[0206] The user intent recognition-based interaction device provided in the embodiments of the present disclosure can execute any of the user intent recognition-based interaction methods provided in the embodiments of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For any content not fully described in the embodiments of the device of the present disclosure, reference can be made to the description of any method embodiment of the present disclosure.

[0207] Figure 11 FIG is a schematic diagram of the structure of an interactive system based on user intention recognition provided by an embodiment of the present disclosure. Figure 11 As shown, the interactive system 1100 based on user intent recognition provided by the embodiment of the present disclosure may include a user interaction device 1101 and an intent analysis device 1102, wherein:

[0208] The user interaction device 1101 is used to obtain a user request statement input by the user and send the user request statement to the intention analysis device 1102;

[0209] Intent analysis device 1102, configured to obtain at least one user intent sample matching the user request sentence based on a text similarity algorithm, and determine the user intent implied in the user request sentence from the at least one user intent sample based on a sentence implication model;

[0210] The intention analysis device 1102 is further configured to send the user intention to the user interaction device 1101, so that the user interaction device 1101 determines the interaction content corresponding to the user intention and outputs the interaction content to the user; or

[0211] The intention analysis device 1102 is further configured to determine interaction content corresponding to the user intention and send the interaction content to the user interaction device 1101 so that the user interaction device 1101 outputs the interaction content to the user.

[0212] Optionally, the user interaction device 1101 is specifically configured to: collect user voice data through a voice module and perform voice recognition on the voice data to obtain a user request statement; or obtain a user request statement input by the user through an online interaction page.

[0213] Optionally, the user interaction device 1101 is also used to: obtain the interactive content corresponding to the user intent sent by the intent analysis device 1102, and output it to the user; or, based on the user intent sent by the intent analysis device 1102, determine the interactive content corresponding to the user intent, and output it to the user.

[0214] Optionally, the user interaction device 1101 is specifically used to: send the user intent sent by the intent analysis device 1102 to the server, so that the server determines the interaction content corresponding to the user intent; receive the interaction content corresponding to the user intent sent by the server, and output it to the user.

[0215] Optionally, the intent analysis device 1102 is specifically configured to send the user intent to the server, so that the server determines the interaction content corresponding to the user intent and sends the interaction content to the user interaction device 1101 .

[0216] In an optional implementation, the user interaction device 1101 and the intent analysis device 1102 are the same terminal device.

[0217] Optionally, the terminal device includes a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

[0218] In an optional embodiment, the user interaction device 1101 and the intent analysis device 1102 are different devices, the user interaction device 1101 includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device 1102 includes a server device.

[0219] Optionally, the interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

[0220] The interactive system based on user intent recognition provided by the embodiments of the present disclosure can implement any interactive method based on user intent recognition provided by the embodiments of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For any content not fully described in the embodiments of the system of the present disclosure, reference can be made to the description of any of the above embodiments of the present disclosure.

[0221] Figure 12 A structural diagram of an electronic device provided in accordance with an embodiment of the present disclosure is used to exemplify an electronic device that implements any user intent recognition method in accordance with an embodiment of the present disclosure, or to exemplify an electronic device that implements any interaction method based on user intent recognition in accordance with an embodiment of the present disclosure, and should not be understood as a specific limitation on the embodiments of the present disclosure.

[0222] like Figure 12 As shown, the electronic device 1200 may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 1201, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1202 or a program loaded from a storage device 1208 into a random access memory (RAM) 1203. Various programs and data required for the operation of the electronic device 1200 are also stored in the RAM 1203. The processor 1201, the ROM 1202, and the RAM 1203 are connected to each other via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.

[0223] Typically, the following devices may be connected to the I / O interface 1205: an input device 1206 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 1207 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1208 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1209. The communication device 1209 may allow the electronic device 1200 to communicate with other devices wirelessly or by wire to exchange data. Although the electronic device 1200 is shown as having various devices, it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may alternatively be implemented or have.

[0224] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart 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 non-transitory computer-readable medium, and the computer program includes 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 device 1209, or installed from the storage device 1208, or installed from the ROM 1202. When the computer program is executed by the processor 1201, the functions defined in any user intent recognition method or interaction method based on user intent recognition provided by the embodiment of the present disclosure can be executed.

[0225] It should be noted that the computer-readable medium mentioned above in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. Computer-readable storage media may be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or components, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or component. In the present disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0226] In some embodiments, the client and server can communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0227] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0228] The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device is enabled to: obtain the user request sentence to be identified; obtain at least one user intent sample that matches the user request sentence based on the text similarity algorithm; and determine the user intent implied in the user request sentence from at least one user intent sample based on the sentence implication model.

[0229] Alternatively, the computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains a user request statement input by the user, and sends the user request statement to an intention analysis device. The intention analysis device is used to obtain at least one user intention sample that matches the user request statement based on a text similarity algorithm, and determines the user intent implied in the user request statement from the at least one user intention sample based on a sentence implication model; and outputs interactive content corresponding to the user intent to the user.

[0230] Alternatively, the computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device: obtains a user request statement sent by the user interaction device; obtains at least one user intent sample that matches the user request statement based on a text similarity algorithm; determines the user intent implied in the user request statement from at least one user intent sample based on a sentence implication model; sends the user intent to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intent and outputs the interaction content to the user; or, determines the interaction content corresponding to the user intent, and sends the interaction content to the user interaction device, so that the user interaction device outputs the interaction content to the user.

[0231] In embodiments of the present disclosure, computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0232] 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 can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than 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 and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0233] The units involved in the embodiments described in this disclosure may be implemented in software or hardware, wherein the name of a unit does not necessarily limit the unit itself.

[0234] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0235] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0236] According to one or more embodiments of the present disclosure, the present disclosure provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor executes any user intent recognition method provided by the embodiments of the present disclosure, or executes any interaction method based on user intent recognition provided by the embodiments of the present disclosure.

[0237] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0238] As used herein, the term "including" and its variations are open-ended, i.e., "including but not limited to." The term "based on" means "based at least in part on." The term "an alternative embodiment" means "at least one alternative embodiment." Concepts such as "first" and "second" mentioned in this disclosure are merely used to distinguish different devices, modules, or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules, or units.

[0239] It should be noted that the modifications of "one" and "multiple" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0240] The above description is merely a preferred embodiment of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also includes other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

[0241] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details have been included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0242] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are merely example forms of implementing the claims.

Claims

1. A method for identifying user intention, characterized in that: include: Obtain the user request statement to be identified; Acquire at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold; Acquire a target user intent sample whose similarity value reaches a second similarity threshold from the user intent sample, and determine that the target user intent sample is the user intent implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold; In the case where the target user intent sample is not obtained, the user intent implied by the user request sentence is determined from the at least one user intent sample according to a sentence implication model.

2. The method according to claim 1, characterized in that The step of obtaining a user request statement to be identified includes: Acquire voice data input by a user, and perform voice recognition on the voice data to obtain the user request statement; or, Obtain the user request statement entered by the user through the online interactive page.

3. The method according to claim 1, characterized in that The acquiring at least one user intent sample matching the user request statement according to the text similarity algorithm includes: According to a vector-based text retrieval method, a BM25 algorithm or a word frequency-inverse text frequency index algorithm, at least one user intent sample matching the user request statement is extracted from an intent sample pool.

4. The method according to claim 3, characterized in that Also includes: An intent sample pool is constructed, wherein the intent sample pool includes an operational intent sample set, and at least one of an in-domain intent sample set, an automatically constructed confusion sample set, and a manually constructed confusion sample set.

5. The method according to claim 4, characterized in that Also includes: The user request sentence is added to the intent sample pool as a user intent sample.

6. The method according to claim 4, characterized in that The intention sample pool includes any one of a travel navigation operation intention sample pool, a knowledge question and answer operation intention sample pool, a question search operation intention sample pool, a guidance service operation intention sample pool, and a multimedia data playback operation intention sample pool.

7. The method according to claim 1, characterized in that The determining the user intent implied by the user request sentence from the at least one user intent sample according to the sentence implied model includes: Inputting the user request sentence and the at least one user intention sample into the sentence implication model to obtain an implication value corresponding to each user intention sample; A user intention sample whose implication value reaches an implication threshold is determined as the user intention implied by the user request sentence.

8. The method according to claim 7, characterized in that Also includes: If there is no user intent sample whose implication value reaches the implication threshold, a message indicating that the user intent implied by the user request sentence cannot be determined is fed back to the user.

9. The method according to claim 1, characterized in that: The determining the user intent implied by the user request sentence from the at least one user intent sample according to the sentence implied model includes: According to one of the convolutional neural network model, the recurrent neural network model, the BERT model or the ALBER-NLI model, the user intent implied in the user request sentence is determined from the at least one user intent sample.

10. The method according to any one of claims 1 to 9, characterized in that: The method is implemented on a terminal device and / or a server device.

11. The method according to claim 10, characterized in that The terminal device includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

12. An interactive method based on user intention recognition, characterized in that: Applicable to user interaction devices, including: Obtain a user request sentence input by a user, and send the user request sentence to an intention analysis device, wherein the intention analysis device is used to obtain at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold, and obtain a target user intention sample whose similarity value reaches a second similarity threshold from the user intention sample, and determine that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold, and when the target user intention sample is not obtained, determine the user intention implied by the user request sentence from the at least one user intention sample according to a sentence implication model; Output interactive content corresponding to the user's intention to the user.

13. The method according to claim 12, characterized in that The step of obtaining a user request statement input by a user includes: Collecting the user's voice data through a voice module and performing voice recognition on the voice data to obtain the user's request statement; or, The user request statement input by the user is obtained through the online interactive page.

14. The method according to claim 12, characterized in that The outputting the interactive content corresponding to the user intention to the user includes: Acquire the interactive content corresponding to the user's intention sent by the intention analysis device, and output it to the user; or Based on the user intent sent by the intent analysis device, interactive content corresponding to the user intent is determined and output to the user.

15. The method according to claim 14, characterized in that The acquiring the interactive content corresponding to the user intention sent by the intention analysis device and outputting it to the user includes: receiving interactive content corresponding to the user's intention sent by a server, and outputting it to the user, wherein the server is used to receive the user's intention sent by the intention analysis device and determine the interactive content corresponding to the user's intention; or The determining, based on the user intent sent by the intent analysis device, interaction content corresponding to the user intent and outputting it to the user comprises: Sending the user intent sent by the intent analysis device to a server, so that the server determines the interaction content corresponding to the user intent; The interactive content corresponding to the user intention is received from the server and outputted to the user.

16. The method according to any one of claims 12 to 15, characterized in that: The user interaction device and the intent analysis device are the same terminal device.

17. The method according to claim 16, characterized in that The terminal device includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

18. The method according to any one of claims 12 to 15, characterized in that: The user interaction device and the intent analysis device are different devices. The user interaction device includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

19. The method according to any one of claims 12 to 15, characterized in that: The interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

20. An interactive method based on user intention recognition, characterized in that: Applied to intent analysis devices, including: Obtaining a user request statement sent by a user interaction device; Acquire at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold; Obtaining a target user intention sample whose similarity value reaches a second similarity threshold from the user intention sample, and determining that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold; if the target user intention sample is not obtained, determining the user intention implied by the user request sentence from the at least one user intention sample according to a sentence implied model; Sending the user intention to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intention and outputs the interaction content to the user; or Determine interaction content corresponding to the user intention, and send the interaction content to the user interaction device, so that the user interaction device outputs the interaction content to the user.

21. The method according to claim 20, characterized in that The determining the interaction content corresponding to the user intention and sending the interaction content to the user interaction device includes: The user intention is sent to a server, so that the server determines interaction content corresponding to the user intention and sends the interaction content to the user interaction device.

22. The method according to claim 20 or 21, characterized in that The user interaction device and the intent analysis device are the same terminal device.

23. The method according to claim 22, characterized in that The terminal device includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker.

24. The method according to claim 20 or 21, characterized in that The user interaction device and the intent analysis device are different devices. The user interaction device includes one of a smart phone, a smart tablet, a smart robot, a smart car, a computer, a voice navigation device or a smart speaker, and the intent analysis device includes a server device.

25. The method according to claim 20 or 21, characterized in that The interactive content corresponding to the user intention includes at least one of the following: information related to travel navigation, information related to knowledge questions and answers, question search feedback information, guidance service information or multimedia data.

26. A user intention recognition device, characterized in that: include: A user request statement acquisition module, used to acquire the user request statement to be identified; A user intention sample determination module, configured to obtain at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold; A user intention determination module is used to obtain a target user intention sample whose similarity value reaches a second similarity threshold from the user intention samples, and determine that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold; if the target user intention sample is not obtained, determine the user intention implied by the user request sentence from the at least one user intention sample according to the sentence implication model.

27. An interactive device based on user intention recognition, characterized in that: Configured in user interaction equipment, including: A user request sentence acquisition module, used to acquire a user request sentence input by a user, and send the user request sentence to an intention analysis device, the intention analysis device is used to acquire at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold, and acquire a target user intention sample whose similarity value reaches a second similarity threshold from the user intention sample, and determine that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold, and when the target user intention sample is not acquired, determine the user intention implied by the user request sentence from the at least one user intention sample according to a sentence implication model; The interactive content output module is used to output interactive content corresponding to the user's intention to the user.

28. An interactive device based on user intention recognition, characterized in that: Configured on intent analysis equipment, including: A user request statement acquisition module, used to acquire a user request statement sent by a user interaction device; A user intention sample determination module, configured to obtain at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold; A user intention recognition module is used to obtain a target user intention sample whose similarity value reaches a second similarity threshold from the user intention sample, determine that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold, and when the target user intention sample is not obtained, determine the user intention implied by the user request sentence from the at least one user intention sample according to the sentence implication model; A user intention sending module is used to send the user intention to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intention and outputs the interaction content to the user; or The interactive content sending module is used to determine the interactive content corresponding to the user intention and send the interactive content to the user interaction device so that the user interaction device outputs the interactive content to the user.

29. An interactive system based on user intention recognition, characterized in that: include: A user interaction device, used to obtain a user request sentence input by a user, and send the user request sentence to an intention analysis device; The intention analysis device is used to obtain at least one user intention sample matching the user request sentence according to a text similarity algorithm, wherein a similarity value between the user intention sample and the user request sentence reaches a first similarity threshold, and obtain a target user intention sample whose similarity value reaches a second similarity threshold from the user intention sample, and determine that the target user intention sample is the user intention implied by the user request sentence, wherein the second similarity threshold is greater than the first similarity threshold, and when the target user intention sample is not obtained, determine the user intention implied by the user request sentence from the at least one user intention sample according to a sentence implication model; The intention analysis device is further used to send the user intention to the user interaction device, so that the user interaction device determines the interaction content corresponding to the user intention and outputs the interaction content to the user; or The intention analysis device is further used to determine the interaction content corresponding to the user intention and send the interaction content to the user interaction device so that the user interaction device outputs the interaction content to the user.

30. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the user intention recognition method described in any one of claims 1 to 11, or executes the interaction method based on user intention recognition described in any one of claims 12 to 19, or executes the interaction method based on user intention recognition described in any one of claims 20 to 25.

31. A computer-readable storage medium, characterized in that: The storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the user intention recognition method described in any one of claims 1 to 11, or executes the interaction method based on user intention recognition described in any one of claims 12 to 19, or executes the interaction method based on user intention recognition described in any one of claims 20 to 25.

32. A computer program product, characterized in that The computer program product includes: a computer program, which is stored in a computer-readable storage medium. When the computer program is read and executed by a processor, the processor executes the user intent recognition method described in any one of claims 1 to 11, or executes the interaction method based on user intent recognition described in any one of claims 12 to 19, or executes the interaction method based on user intent recognition described in any one of claims 20 to 25.

Images