An intent recognition method, device, apparatus and medium

Abstract

Embodiments of the present application provide an intent recognition method, device, equipment and medium to recognize the intent corresponding to similar text. The method comprises: in response to an instruction of performing intent recognition on a to-be-recognized text, obtaining the to-be-recognized text; classifying the to-be-recognized text by using a first model to determine the similar text classification corresponding to the to-be-recognized text, wherein the similar text samples in different similar text classifications are used to train corresponding second models in advance; determining the second model corresponding to the similar text classification corresponding to the to-be-recognized text; performing intent recognition on the to-be-recognized text by using the second model to obtain the target intent corresponding to the to-be-recognized text, thereby improving the accuracy of similar text intent recognition.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and in particular to an intent recognition method, apparatus, device, and medium. Background Technology

[0002] Intent recognition is a commonly used semantic understanding method, often used in human-computer interaction. It can identify the user's intent based on the text input and provide corresponding responses or strategies based on that intent.

[0003] Currently, most existing intent recognition methods train intent recognition models using a large number of intent question texts to achieve intent recognition of intent question texts. This method can achieve good results on datasets. However, in practical applications, due to the existence of many intent questions with similar texts but different intents, this method cannot accurately identify the intent corresponding to the intent question text and is prone to misjudgment. Summary of the Invention

[0004] This invention provides an intent recognition method, apparatus, device, and medium to solve the problem that existing intent recognition methods cannot effectively recognize intent questions with similar text but different intents.

[0005] This application provides an intent recognition method, including:

[0006] In response to an instruction to perform intent recognition on the text to be recognized, the text to be recognized is acquired; the text to be recognized is classified using a first model to determine the similar text categories corresponding to the text to be recognized, wherein a second model is pre-trained using similar text samples from different similar text categories; the second model corresponding to the similar text categories corresponding to the text to be recognized is determined; the intent recognition on the text to be recognized is performed using the second model to acquire the target intent corresponding to the text to be recognized.

[0007] The above method improves the accuracy of intent recognition of similar texts by classifying texts in advance based on the similarity between texts, using a first model to classify the text to be identified, determining the similar text category corresponding to the text to be identified, and using a second model corresponding to the similar text category to perform intent recognition on the text to be identified.

[0008] An optional implementation method is to classify the different similar texts according to the following method: obtain multiple text samples with different intentions and calculate the similarity between different text samples; determine the text samples with similarity greater than a first preset threshold as similar text samples in the same group.

[0009] An optional implementation method is to train the first model as follows: label similar text samples in the same group with corresponding group labels; input similar text samples in different groups into the first model respectively, and train the first model with the goal of outputting the corresponding group labels.

[0010] An optional implementation method is that, when using the second model to perform intent recognition on the text to be recognized, the method further includes: using the second model to extract information from the text to be recognized to obtain valid information of the text to be recognized.

[0011] An optional implementation further includes, before responding to the instruction to perform intent recognition on the text to be recognized, the following steps: acquiring similar text samples from different groups, and intent labels and valid information labels corresponding to each similar text sample; inputting similar text samples from the same group, and intent labels and valid information labels corresponding to each similar text sample, into a corresponding second model; determining a first loss function based on the intent recognition result output by the second model and the corresponding intent label, and determining a second loss function based on the information extraction result output by the second model and the corresponding valid information label; training the second model with the goal of minimizing the sum of the first loss function and the second loss function, wherein when the change value of the sum of the first loss function and the second loss function determined for a consecutive preset number of times is less than a second preset threshold, the sum of the first loss function and the second loss function is determined to be minimized.

[0012] An optional implementation includes key information associated with intent in the aforementioned effective information. After obtaining the target intent corresponding to the text to be identified, the implementation further includes: matching the key information of the text to be identified with reference texts corresponding to different domains that are pre-stored in the database; when it is determined that the domain corresponding to the matched reference text is different from the domain corresponding to the target intent, the intent corresponding to the domain corresponding to the matched reference text is re-determined from among the different intents corresponding to the similar text categories corresponding to the text to be identified, based on the domain corresponding to the matched reference text.

[0013] An optional implementation involves matching the key information of the text to be identified with reference texts corresponding to different domains pre-stored in a database, including: obtaining a first domain corresponding to the target intent and matching the key information with reference texts corresponding to the first domain in the database; when it is determined that no reference text corresponding to the key information is matched, matching the key information with reference texts corresponding to other domains in the database other than the first domain.

[0014] This application also provides an intent recognition device, including:

[0015] The acquisition module is used to acquire the text to be identified in response to the instruction to perform intent recognition on the text to be identified;

[0016] The classification module is used to classify the text to be identified using the first model and determine the similar text categories corresponding to the text to be identified. The second model is trained in advance using similar text samples from different similar text categories.

[0017] The determination module is used to determine the second model corresponding to the similar text classification of the above-mentioned text to be identified;

[0018] The recognition module is used to perform intent recognition on the text to be recognized using the second model described above, and to obtain the target intent corresponding to the text to be recognized.

[0019] This application also provides an intent recognition device, including a processor and a memory. The memory stores instructions that can be executed by at least one processor. When the computer program is executed by the processor, it implements the steps of any one of the intent recognition methods.

[0020] This application also provides a computer storage medium including computer-readable instructions, which, when read and executed by a computer, cause the computer to perform any of the steps described in the above-described intent recognition method.

[0021] This application also provides a computer program product, including a computer program stored in a computer-readable storage medium; when the processor of the intent recognition device reads the computer program from the computer-readable storage medium, the processor executes the computer program, causing the electronic device to perform any of the steps of the above-described intent recognition method.

[0022] Furthermore, the technical effects of any of the above-mentioned implementations of the intent recognition device, equipment, storage medium, and computer program product provided in the embodiments of this application can be found in the technical effects of the above-mentioned implementations of the intent recognition method, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic diagram illustrating an application scenario provided in an embodiment of this application;

[0025] Figure 2A flowchart illustrating an intent recognition method provided in an embodiment of this application;

[0026] Figure 3 A flowchart illustrating a similar text classification method provided in this application embodiment;

[0027] Figure 4 A schematic diagram illustrating different intentions in the field of people's livelihood, provided for the embodiments of this application;

[0028] Figure 5 A schematic diagram illustrating different intentions in the ecological field, provided for embodiments of this application;

[0029] Figure 6 A schematic diagram illustrating an intent grouping result provided in an embodiment of this application;

[0030] Figure 7 A schematic diagram illustrating another intentional grouping result provided in an embodiment of this application;

[0031] Figure 8 A schematic diagram illustrating an intent grouping process provided in an embodiment of this application;

[0032] Figure 9 This is a schematic diagram of the structure of a first model provided in an embodiment of this application;

[0033] Figure 10 A flowchart illustrating a first model training method provided in an embodiment of this application;

[0034] Figure 11 A schematic diagram illustrating the function of a second model provided in an embodiment of this application;

[0035] Figure 12 This is a schematic diagram of the structure of a second model provided in an embodiment of this application;

[0036] Figure 13 A flowchart illustrating a second model training method provided in an embodiment of this application;

[0037] Figure 14 A flowchart illustrating an intent recognition result verification method provided in an embodiment of this application;

[0038] Figure 15 A flowchart illustrating another intent recognition result verification method provided in this application embodiment;

[0039] Figure 16 A schematic diagram illustrating the overall process of intent recognition provided in an embodiment of this application;

[0040] Figure 17 A schematic diagram of an intent recognition device provided in an embodiment of this application;

[0041] Figure 18 A schematic diagram of another intent recognition device provided in the embodiments of this application;

[0042] Figure 19 This is a schematic diagram of the structure of an intent recognition device provided in an embodiment of this application;

[0043] Figure 20 This is a schematic diagram of another intent recognition device provided in an embodiment of this application. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Although the embodiments of this application provide method operation steps as shown in the following embodiments or drawings, the method may include more or fewer operation steps based on conventional or non-creative labor. In steps where there is no logically necessary causal relationship, the execution order of these steps is not limited to the execution order provided in the embodiments of this application. In actual processing or when the device executes the method, it may be executed in the order shown in the embodiments or drawings, or in combination. Obviously, the embodiments described in this application are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of this application.

[0045] It should be noted that the application scenarios described in the following embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0046] Intent recognition is a commonly used semantic understanding method, often used in human-computer interaction. It can identify the user's intent based on the text input and provide corresponding responses or strategies based on that intent.

[0047] Existing intent recognition methods involve training a domain identification model using a large number of intent question texts when dealing with multiple domains. This allows the model to identify the domain corresponding to the intent of the intent question text, and then train the intent recognition model to determine which intent within that domain the intent question text corresponds to. While this method performs well on datasets, in practical applications, it struggles because semantic understanding often reveals texts with similar sentence structures across different domains, differing only in keywords. In such cases, this approach fails to accurately identify the intent corresponding to the intent question text and is prone to misjudgment. For example, the intent question text corresponding to an economic indicator might be "What is the GDP of City A?", while the intent question text corresponding to a livelihood indicator might be "What is the per capita income of residents in City A?". Although these belong to different domains, the texts are very similar, differing only in the keywords of the indicators. Since keywords across different domains cannot be enumerated, and new keywords may be added during the actual intent recognition process, this method easily leads to incorrect domain identification, resulting in incorrect intent recognition results.

[0048] This application addresses the problem that existing intent recognition methods cannot effectively identify intent questions with similar text but different intents. It provides an intent recognition method, apparatus, device, and medium. By pre-classifying texts based on their similarity, and using a first model to classify the text to be identified, the similar text category corresponding to the text to be identified is determined. Then, using a second model corresponding to the similar text category, the intent of the text to be identified is recognized, thereby improving the accuracy of intent recognition for similar texts.

[0049] like Figure 1 The diagram shown is an application scenario illustration of an embodiment of this application. The application scenario diagram includes a terminal device 110 and a server 120.

[0050] In this embodiment, the terminal device 110 includes, but is not limited to, mobile phones, tablets, laptops, desktop computers, e-book readers, smart voice interaction devices, smart home appliances, and in-vehicle terminals. The terminal device may have an intent recognition-related client installed; this client can be software, a webpage, a mini-program, etc. The server 120 is the backend server corresponding to the software, webpage, mini-program, etc., or a server specifically used for intent recognition; this application does not impose specific limitations. The server 120 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0051] It should be noted that the intent recognition method in this application embodiment can be executed by an electronic device, which can be a server 120 or a terminal device 110. That is, the method can be executed by the server 120 or the terminal device 110 alone, or by the server 120 and the terminal device 110 together. For example, when executed by server 120, a user can upload the text to be recognized and generate an instruction to perform intent recognition on the text by logging into the client on terminal device 110. Terminal device 110 sends the text to be recognized and the instruction to perform intent recognition on the text to be recognized to server 120 through a communication network. After receiving the instruction to perform intent recognition on the text to be recognized, server 120 retrieves the text to be recognized from terminal device 110 in response to the instruction, classifies the text to be recognized using a first model, and determines the similar text category corresponding to the text to be recognized. In this process, a second model is pre-trained using similar text samples from different similar text categories; the second model corresponding to the similar text category corresponding to the text to be recognized is determined; the second model is used to perform intent recognition on the text to be recognized, obtain the target intent corresponding to the text to be recognized, and send the target intent to terminal device 110. After receiving the target intent corresponding to the text to be recognized sent by server 120, terminal device 110 provides feedback on the target intent to the user.

[0052] In one optional implementation, the terminal device 110 and the server 120 can communicate via a communication network. In another optional implementation, the communication network is a wired network or a wireless network.

[0053] It should be noted that, Figure 1The examples shown are merely illustrative; in reality, the number of terminal devices and servers is unlimited and not specifically limited in this application embodiment. In this application embodiment, when there are multiple servers, these multiple servers can form a blockchain, and the servers are nodes on the blockchain; as disclosed in the intent recognition method of this application embodiment, the intent recognition data involved can be stored on the blockchain. Furthermore, this application embodiment can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, and other scenarios.

[0054] To further illustrate the technical solutions provided by the embodiments of this application, the intent recognition method provided by the exemplary embodiments of this application will be described below in conjunction with the application scenarios described above, with reference to the accompanying drawings and specific implementation methods. It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principles of this application, and the implementation methods of this application are not limited in any way in this respect.

[0055] Figure 2 This is a flowchart illustrating an intent recognition method provided in an embodiment of this application, as shown below. Figure 2 As shown in the figure, this application provides an intent recognition method, which may include the following steps:

[0056] Step 201: In response to the instruction to perform intent recognition on the text to be recognized, obtain the text to be recognized;

[0057] Step 202: Classify the text to be identified using the first model and determine the similar text classifications corresponding to the text to be identified. The second model is trained in advance using similar text samples from different similar text classifications.

[0058] In this embodiment of the application, in order to improve the accuracy of intent recognition of similar texts, the texts are classified in advance according to the similarity between the texts. Texts with high similarity are divided into a similar text category. When performing intent recognition on the text to be recognized, the first model is first used to determine the similar text category corresponding to the text to be recognized, so that the second model can be used for intent recognition in the future.

[0059] Step 203: Determine the second model corresponding to the similar text classification of the text to be identified above;

[0060] In this embodiment, the second model is pre-trained using similar texts from different similar text categories, meaning there is a one-to-one correspondence between the similar text categories and the second model, with each similar text category corresponding to one second model. After determining the similar text category corresponding to the text to be identified using the first model, the corresponding second model is determined based on that similar text category, and the second model is used to perform intent recognition on the text to be identified.

[0061] Step 204: Use the second model described above to perform intent recognition on the text to be recognized, and obtain the target intent corresponding to the text to be recognized.

[0062] In this embodiment, similar texts are pre-classified and grouped together. When recognizing the text to be identified, a first model is used to determine the corresponding similar text category, and a second model corresponding to that category is used for feature extraction. By integrating similar texts, it first determines which similar text category the text to be identified belongs to, and then uses the second model to determine the intent corresponding to the text to be identified from the intents contained in that similar text category, thereby improving the accuracy of text intent recognition.

[0063] In some possible embodiments, the classification method for similar texts in step 202 above is as follows: Figure 3 As shown, it specifically includes:

[0064] Step 301: Obtain multiple text samples with different intentions and calculate the similarity between different text samples;

[0065] Step 302: Text samples with a similarity greater than a first preset threshold are identified as similar text samples in the same group.

[0066] In some embodiments, multiple text samples with identified intentions and different intentions are obtained, and the similarity between the multiple samples is calculated. The text samples are grouped by similarity, that is, text samples with similarity greater than a first preset threshold are divided into similar text samples of the same similar text category. That is, each similar text category includes multiple texts with similarity between them that are higher than the first preset threshold.

[0067] After classifying similar texts based on text similarity, the intent of each text sample in each similar text category is recorded. The same similar text category may contain the same intent or different intents. The intent list corresponding to each similar text category is also recorded so that the intent corresponding to the text to be identified can be determined from the intent list corresponding to the similar text category using the second model corresponding to the similar text category.

[0068] The following example illustrates the process of segmenting similar text samples in detail. It should be noted that, for ease of explanation, only eight different intents are shown in the following example. In the actual intent recognition process, more intents and domains are included.

[0069] Please refer to Figure 4 and Figure 5 Suppose there are multiple intentions belonging to two different areas: people's livelihood and ecology, such as Figure 4As shown, there are four different intentions behind checking livelihood indicators, livelihood statistics, emergency response plans for livelihood incidents, and governance resources in the field of people's livelihood; for example... Figure 5 As shown, there are four different purposes in the ecological field: checking ecological indicators, ecological statistics, ecological incident contingency plans, and law enforcement agencies.

[0070] In this embodiment of the application, text samples of multiple identified intents corresponding to the above 8 different intents are first obtained. In order to ensure the accuracy of the calculation results, multiple text samples can be obtained for each intent.

[0071] The similarity between the above multiple text samples is calculated and compared with a first preset threshold. Texts with a similarity greater than the first preset threshold are identified as belonging to the same similar text category.

[0072] For example, assuming the obtained text samples include text sample 1, text sample 2, and text sample 3, after calculating the similarity between the text samples, it is determined that the similarity between text sample 1 and text sample 2 is 80%, the similarity between text sample 1 and text sample 3 is 40%, the similarity between text sample 2 and text sample 3 is 25%, and the first preset threshold is 70%, then it is determined that text sample 1 and text sample 2 belong to the same similar text category.

[0073] In some optional embodiments, for ease of calculation, the similarity between each text sample and the same text sample can be calculated to determine whether a text sample belongs to the same similar text category. For example, when the similarity between text sample 1 and text sample 2 is 80%, and the similarity between text sample 1 and text sample 4 is 75%, both of which are greater than the first preset threshold of 70%, it can be determined that text sample 1, text sample 2, and text sample 4 belong to the same similar text category.

[0074] After calculating the similarity between multiple text samples corresponding to the above eight different intentions, the similar text categories are determined based on the similarity, and the intentions included in the similar text categories are determined. One possible implementation is that, based on the similar text categories, if the text similarity between the intentions of searching for livelihood indicators and searching for ecological indicators is greater than a first preset threshold, then the two intentions of searching for livelihood indicators and searching for ecological indicators are grouped into the same group. For example... Figure 6 In another possible embodiment of the query indicator grouping shown, when it is determined, based on similar text classification, that there are no other intents with similarity greater than a first preset threshold corresponding to the intent to query governance resources, the intent to query governance resources is grouped into a separate group, such as... Figure 7 As shown.

[0075] It should be noted that after determining the classification of similar texts based on similarity and identifying the intent corresponding to the texts in each similar text classification, it was found that there was a common intent (for ease of description, let's assume it's the same intent). Figure 1 When multiple texts belong to different similar text categories (let's say similar text category 1 and similar text category 2), based on the meaning included in similar text category 1 and similar text category 2... Figure 1 The number of corresponding texts determines the meaning. Figure 1 The corresponding groups.

[0076] like Figure 8 As shown, according to the above method, the above 8 intentions can be divided into 5 groups: querying indicators, querying statistics, querying contingency plans, querying governance resources, and querying law enforcement units. The intentions corresponding to each group are all intentions whose text similarity is higher than the first preset threshold. That is, the texts corresponding to the intentions in each group belong to the same similar text category.

[0077] In some optional embodiments, the structure of the first model in step 202 above is as follows: Figure 9 As shown, the first model includes an input layer, an embedding layer, a pooling layer, and a fully connected layer. The sample to be identified is input into the embedding layer of the first model through the input layer. The embedding layer extracts the main features of the sample and converts the input text into a vector form. This vector is then input into the pooling layer, where the embedding layer performs average pooling before inputting it into the fully connected layer. The fully connected layer classifies the feature vector of the text to be identified, determining the corresponding similar text category. In some embodiments, the fully connected layer of the first model calculates the probability that the feature vector of the text to be identified belongs to each different similar text category, and determines the corresponding similar text category based on the probability value.

[0078] In some embodiments of this application, the first model described above adopts the following... Figure 10 The following method was used for training:

[0079] Step 1001: Label similar text samples in the same group with corresponding group labels;

[0080] Step 1002: Input similar text samples from different groups into the first model respectively, and train the first model with the goal of outputting the corresponding group labels.

[0081] Specifically, text samples of the same similar text type are labeled, and the text samples are labeled with the corresponding similar text type labels (i.e., group labels). After labeling similar text samples of multiple similar text types in this way, similar text samples corresponding to different similar text types are obtained from them and input into the first model respectively. The similar text types output by the first model are obtained. When it is determined that the similar text type is consistent with the corresponding labeled similar text type label, the training of the first model ends, and the trained first model is obtained.

[0082] In some embodiments of this application, such as Figure 11 As shown, the second model in step 203 above is used not only for intent recognition of the text to be recognized, but also for information extraction of the text to be recognized, to obtain effective information of the text to be recognized.

[0083] In this application embodiment, slots corresponding to each intent are predefined (there is a one-to-one correspondence between slots and intents). The valid information of the text to be identified corresponds to the slot of the target intent of the determined text to be identified, such as time, location, etc. In some embodiments, the above... Figure 4 and Figure 5 The slots corresponding to multiple intents are shown in the table below:

[0084] Table 1 shows the slots corresponding to each intention.

[0085]

[0086] In some embodiments of this application, the structure of the second model in step 203 above is as follows: Figure 12 As shown, the first model mentioned above includes an input layer, an embedding layer, a pooling layer, and a fully connected layer;

[0087] In this process, the sample to be identified is input into the embedding layer of the first model through the input layer. The embedding layer performs feature extraction to obtain a first feature vector for intent recognition and a second feature vector for effective information extraction. The first and second feature vectors are then input into the pooling layer. The pooling layer and the embedding layer perform average pooling operations on the first and second feature vectors respectively, and then input them into the fully connected layer. The fully connected layer performs intent recognition based on the first feature vector to obtain the target intent of the text to be identified, and performs information extraction based on the second feature vector to obtain the effective information of the text to be identified. Finally, the target intent and the effective information are output.

[0088] In some embodiments of this application, such as Figure 13 As shown, the training of the second model mentioned above includes the following steps:

[0089] Step 1301: Obtain similar text samples from different groups, as well as the intent labels and valid information labels corresponding to each similar text sample;

[0090] The intent labels mentioned above refer to the intents corresponding to the labeled similar text samples, while the effective information labels refer to the text information in the labeled similar text samples that corresponds to the slots of the intents, such as words representing time and words representing location in the similar text samples.

[0091] Step 1302: Input the similar text samples in the same group, as well as the intent labels and effective information labels corresponding to each similar text sample, into the corresponding second model;

[0092] The second model mentioned above has a one-to-one correspondence with the similar text classification (i.e., the grouping mentioned above). That is, a second model corresponding to the similar text classification is trained using the similar text samples of each similar text classification.

[0093] Step 1303: Determine the first loss function based on the intent recognition result output by the second model and the corresponding intent label, and determine the second loss function based on the information extraction result output by the second model and the corresponding valid information label;

[0094] Step 1304: Train the second model with the goal of minimizing the sum of the first loss function and the second loss function.

[0095] Specifically, when the change in the sum of the first loss function and the second loss function determined after a preset number of consecutive training iterations is less than the second preset threshold, that is, when the sum of the first loss function and the second loss function determined after a preset number of consecutive training iterations no longer changes significantly, the sum of the first loss function and the second loss function is determined to be the minimum, and the training of the second model is completed.

[0096] Specifically, the change value (i.e. the difference) of the sum of the first loss function and the second loss function determined by the above-mentioned consecutive preset number of times is less than the second preset threshold. This can be either the change value of the sum of the first loss function and the second loss function determined in every two adjacent consecutive preset number of times is less than the second preset threshold, or the largest change value among the sum of the first loss function and the second loss function determined by the consecutive preset number of times is less than the second preset threshold.

[0097] The aforementioned consecutive preset number of times and the second preset threshold can be set according to actual needs, and are not limited in this embodiment;

[0098] By utilizing the intent recognition and information extraction results output by the second model, and combining them with the labels corresponding to similar text samples, two loss functions (i.e., the first loss function and the second loss function) are determined. The second model is then jointly trained for intent recognition and information extraction based on the sum of the two loss functions, ensuring the accuracy of the second model's output.

[0099] In some embodiments of this application, such as Figure 14 As shown, after obtaining the target intent corresponding to the text to be identified, the intent recognition result is verified, as follows: Figure 14 As shown, verifying the intent recognition result includes the following steps:

[0100] Step 1401: Match the key information of the text to be identified with reference texts corresponding to different fields that are pre-stored in the database;

[0101] In some embodiments of this application, the aforementioned valid information includes key information associated with the intent, such as the livelihood indicator information corresponding to the intent to query livelihood indicators in Table 1 above, and the livelihood event type information corresponding to the intent to conduct livelihood statistics.

[0102] In some embodiments of this application, a database is pre-established, which stores reference text libraries corresponding to various fields. The reference text libraries store reference texts (vocabularies) corresponding to each field. For example, the reference text library for the field of people's livelihood stores reference texts such as per capita income of residents, while the reference text library for the field of ecology stores reference texts such as air quality and temperature.

[0103] Step 1402: When it is determined that the domain corresponding to the matched reference text is different from the domain corresponding to the target intent, the intent corresponding to the domain corresponding to the matched reference text is re-determined from among the different intents corresponding to the similar text categories of the identified text.

[0104] By determining whether the domain corresponding to the matched reference text is the same as the domain corresponding to the target intent, it can be further determined whether the identified target domain is correct. If they are different, based on the domain corresponding to the matched reference text, the intent corresponding to the domain corresponding to the matched reference text is re-determined among the different intents corresponding to the similar text categories of the text to be identified, and this intent is taken as the intent corresponding to the text to be identified.

[0105] For example, when the target intent of the text to be identified is to search for livelihood indicators and the key information to be extracted is temperature, the temperature is matched with the reference texts corresponding to various fields in the database. When the field corresponding to the reference text is determined to be ecology, the intent corresponding to ecology is determined from the similar text classification (i.e., searching for indicators) corresponding to the text to be identified, i.e., searching for ecological indicators, and this intent is taken as the intent corresponding to the text to be identified.

[0106] In some embodiments of this application, to reduce the computational load of matching the key information of the text to be identified with reference texts corresponding to different fields pre-stored in the database, and to improve the matching efficiency, the embodiments of this application match the key information of the text to be identified with reference texts corresponding to different fields pre-stored in the database, specifically including, as follows: Figure 15 The following steps are shown:

[0107] Step 1501: Obtain the first domain corresponding to the target intent, and match the key information with the reference text corresponding to the first domain in the database.

[0108] By determining whether the key information of the text to be identified matches the reference text in the database corresponding to the target intent, it is determined whether the identified target intent is correct. If a reference text corresponding to the key information is matched, it means that the identified target intent is correct, and step 1502 is not executed. If a reference text corresponding to the key information is not matched, it means that the identified target intent is incorrect, and step 1502 is executed.

[0109] Step 1502: If no matching reference text is found for the key information, match the key information with the reference texts for other domains in the database, excluding the first domain.

[0110] Once it is determined that the identified target intent is incorrect, the key information is matched with reference texts in other domains in the database besides the first domain to obtain the domains corresponding to the matched reference texts. This facilitates subsequent modification of the intent recognition results based on the matched domains.

[0111] The following is based on Figure 16 The flowchart shown illustrates the intent recognition process in this embodiment of the application in detail. Please refer to it. Figure 16 The intent recognition method in this application includes the following steps:

[0112] Step 1601: Obtain the text to be recognized;

[0113] Step 1602: Classify the text to be identified using the first model to obtain the corresponding similar text classification;

[0114] Step 1603: Use the second model corresponding to the similar text classification to perform intent recognition on the text to be identified, and obtain the target intent corresponding to the text to be identified;

[0115] Step 1604: Match the key information of the text to be identified with the reference text in the database that corresponds to the first domain to which the target intent belongs;

[0116] Step 1605: Determine whether a reference text is matched. If yes, proceed to step 1606; otherwise, proceed to step 1607.

[0117] Step 1606: Confirm that the intent recognition result is correct;

[0118] Step 1607: Match the key information with the reference texts corresponding to other domains in the database, excluding the first domain;

[0119] Step 1608: Determine the second domain corresponding to the matched reference text;

[0120] Step 1609: Determine the intent corresponding to the second domain from the similar text classification, and identify it as the intent corresponding to the text to be identified.

[0121] In this embodiment, for similar texts with different intentions but similar texts, the texts are classified in advance according to the similarity between the texts. That is, different intentions are regrouped according to the similarity between the texts, thereby improving the accuracy of text intention grouping (which similar text category the text belongs to); and by verifying the intention recognition results using key information extracted from the text to be recognized, the accuracy of intention recognition is improved.

[0122] Based on the same inventive concept, embodiments of this application also provide an intent recognition device. For example... Figure 17 and Figure 18 As shown, this is a schematic diagram of the structure of an intent recognition device, which may include:

[0123] The acquisition module 1701 is used to acquire the text to be recognized in response to an instruction to perform intent recognition on the text to be recognized;

[0124] The classification module 1702 is used to classify the text to be identified using the first model and determine the similar text categories corresponding to the text to be identified. The second model is trained in advance using similar text samples from different similar text categories.

[0125] The determination module 1703 is used to determine the second model corresponding to the similar text classification of the text to be identified;

[0126] The recognition module 1704 is used to perform intent recognition on the text to be recognized using the second model, and to obtain the target intent corresponding to the text to be recognized.

[0127] Optionally, the classification module 1702 is further used to classify different similar texts according to the following method: obtain multiple text samples with different intentions and calculate the similarity between different text samples; determine the text samples with similarity greater than a first preset threshold as similar text samples in the same group.

[0128] Optionally, the above-mentioned intent recognition device further includes a first model training module 1801, which is used to train the first model in the following manner: labeling similar text samples in the same group with corresponding group labels; inputting similar text samples in different groups into the first model respectively, and training the first model with the goal of outputting the corresponding group labels.

[0129] Optionally, when the recognition module 1704 is used to perform intent recognition on the text to be recognized using the second model, it further includes: using the second model to extract information from the text to be recognized and obtaining valid information of the text to be recognized.

[0130] Optionally, the aforementioned intent recognition model further includes a second model training module 1802. Before the acquisition module 1701 responds to the instruction to perform intent recognition on the text to be recognized, the second model training module 1802 is used to: acquire similar text samples from different groups, and intent labels and effective information labels corresponding to each similar text sample; input similar text samples from the same group, and intent labels and effective information labels corresponding to each similar text sample into the corresponding second model; determine a first loss function based on the intent recognition result output by the second model and the corresponding intent label, and determine a second loss function based on the information extraction result output by the second model and the corresponding effective information label; train the second model with the goal of minimizing the sum of the first loss function and the second loss function, wherein when the change value of the sum of the first loss function and the second loss function determined for a consecutive preset number of times is less than a second preset threshold, the sum of the first loss function and the second loss function is determined to be minimized.

[0131] Optionally, the aforementioned valid information includes key information associated with the intent. After obtaining the target intent corresponding to the text to be identified, the aforementioned identification module 1704 is further configured to: match the key information of the text to be identified with reference texts corresponding to different domains that are pre-stored in the database; when it is determined that the domain corresponding to the matched reference text is different from the domain corresponding to the target intent, based on the domain corresponding to the matched reference text, redetermine the intent corresponding to the domain corresponding to the matched reference text among the different intents corresponding to the similar text categories corresponding to the text to be identified.

[0132] Optionally, the aforementioned recognition module 1704 is used to match the key information of the text to be recognized with reference texts corresponding to different fields that are pre-stored in the database, including: obtaining the first field corresponding to the target intent and matching the key information with the reference texts corresponding to the first field in the database; when it is determined that no reference text corresponding to the key information is matched, matching the key information with reference texts corresponding to other fields in the database other than the first field.

[0133] For ease of description, the above sections are divided into modules (or units) according to their functions and described separately. Of course, in implementing this application, the functions of each module (or unit) can be implemented in one or more software or hardware components.

[0134] Having introduced the intention recognition method and apparatus according to exemplary embodiments of this application, we will now introduce an apparatus for intention recognition according to another exemplary embodiment of this application.

[0135] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0136] Based on the same inventive concept as the above-described method embodiments, this application also provides an electronic device. In one embodiment, the electronic device may be a server, such as... Figure 1 The server 120 is shown. In this embodiment, the structure of the electronic device can be as follows: Figure 19 As shown, it includes a memory 1901, a communication module 1903, and one or more processors 1902.

[0137] The memory 1901 is used to store computer programs executed by the processor 1902. The memory 1901 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0138] Memory 1901 may be volatile memory, such as random-access memory (RAM); memory 1901 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 1901 may be any other medium capable of carrying or storing a desired computer program having the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 1901 may be a combination of the above-described memories.

[0139] Processor 1902 may include one or more central processing units (CPUs) or digital processing units, etc. Processor 1902 is used to implement the above-described intent recognition method when calling computer programs stored in memory 1901.

[0140] The communication module 1903 is used to communicate with terminal devices and other servers.

[0141] This application embodiment does not limit the specific connection medium between the memory 1901, communication module 1903, and processor 1902. This application embodiment... Figure 19 The memory 1901 and the processor 1902 are connected via a bus 1904, which is in the middle of the memory. Figure 19 The diagram uses thick lines to describe the connections between other components; these are for illustrative purposes only and should not be considered limiting. The 1904 bus can be divided into address bus, data bus, control bus, etc. For ease of description, Figure 19 It is described using only a thick line, but does not indicate that there is only one bus or one type of bus.

[0142] The memory 1901 stores a computer storage medium, which stores computer-executable instructions for implementing the intent recognition method of this application embodiment. The processor 1902 is used to execute the aforementioned intent recognition method, such as... Figure 2 As shown.

[0143] In another embodiment, the electronic device may be a terminal, such as Figure 1 The terminal 110 is shown. In this embodiment, the electronic device can be structured as follows: Figure 20 As shown, the terminal device 200 includes components such as a communication component 210, a processor 220, a memory 230, a display 240, an input component 250, an audio circuit 260, a SIM card interface 270, and a sensor 280.

[0144] It should be understood that, Figure 20 The terminal device 200 shown is merely an example, and the terminal device 200 may have more than... Figure 20 The more or fewer components shown can be combined into two or more components, or they can have different component configurations. The various components shown in the figure can be implemented in hardware, software, or a combination of hardware and software, including one or more signal processing and / or application-specific integrated circuits.

[0145] The communication component 210 is used to receive or send call requests, receive and send signals during a call, and connect to a server to upload or download data. The communication component 210 may include an RF (radio frequency) circuit 211 and a Wi-Fi (Wireless Fidelity) module 212.

[0146] RF circuit 211 can be used for receiving and transmitting signals during information transmission or calls. It can receive downlink data from the base station and pass it to processor 220 for processing; it can also send uplink data to the base station. Typically, RF circuit 211 includes, but is not limited to, devices such as an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier, and a duplexer. RF circuit 211 can receive electromagnetic waves via the antenna and perform filtering, amplification, and other processing on the received electromagnetic waves before transmitting them to a modem processor for demodulation. RF circuit 211 can also amplify the signal modulated by the modem processor and radiate it as electromagnetic waves via the antenna. In some embodiments, at least some functional modules of RF circuit 211 can be housed in processor 220. In some embodiments, at least some functional modules of RF circuit 211 and at least some modules of processor 220 can be housed in the same device. The RF circuit 211 and antenna of terminal device 200 are coupled, enabling terminal device 200 to communicate with networks and other devices via wireless communication technology.

[0147] Wi-Fi is a short-range wireless transmission technology. Terminal device 200 can use Wi-Fi module 212 to help users send and receive emails, browse web pages, and access streaming media, providing wireless broadband internet access. Wi-Fi module 212 can connect to a router to connect to an external network. Wi-Fi module 212 can also connect to a server to upload or download data.

[0148] The memory 230 can be used to store data or program code used during the operation of the terminal device. The processor 220 executes various functions and data processing of the terminal device 200 by running the data or program code stored in the memory 230. The memory 230 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. The memory 230 stores the operating system that enables the terminal device 200 to run.

[0149] The display 240 is used to display information input by the user or information provided to the user, as well as various menus of the terminal device 200, forming a graphical user interface (GUI). Specifically, the display 240 may include a screen disposed on the front of the terminal device 200. The display may be configured as a liquid crystal display, a light-emitting diode, or the like. The display 240 can be used to display the interface of the terminal device during operation.

[0150] Input component 250 can be used to receive numeric or character information input by the user, as well as various user operations, and generate signal inputs related to user settings and function control of terminal device 200. Specifically, input component 250 may include buttons and a touch screen. The touch screen may be located on the front of terminal device 200 and can collect touch operations on or near it by the user, such as clicking buttons, dragging scroll bars, etc.

[0151] The touch screen can be placed on top of the display. In some embodiments, the touch screen and the display can be integrated to realize the input and output functions of the terminal device 200. After integration, it can be referred to as a touch display.

[0152] The terminal device 200 may also include a positioning module, such as a satellite positioning module or a mobile communication network positioning module, which can determine the geographical location of the terminal device 200 in real time.

[0153] Audio circuitry 260, speaker 261, and microphone 262 provide an audio interface between the user and terminal device 200. Audio circuitry 260 converts received audio data into electrical signals, which are then transmitted to speaker 261, where they are converted into sound signals for output. Terminal device 200 may also be equipped with volume buttons for adjusting the volume of the sound signal. Conversely, microphone 262 converts collected sound signals into electrical signals, which are received by audio circuitry 260, converted into audio data, and then output to RF circuitry 211 for transmission to, for example, another terminal, or to memory 230 for further processing.

[0154] The SIM card interface 270 is used to connect a SIM card. The SIM card can be inserted into or removed from the SIM card interface 270 to make contact with and separate from the terminal device 200. The terminal device 200 can support one or N SIM card interfaces, where N is a positive integer greater than 1. The SIM card interface 270 can support Nano SIM cards, Micro SIM cards, SIM cards, etc. Multiple cards can be inserted into the same SIM card interface simultaneously. The multiple cards can be of the same or different types. The SIM card interface is also compatible with different types of SIM cards. The SIM card interface is also compatible with external memory cards. The terminal device 200 interacts with the network through the SIM card to realize functions such as calls and data communication. In some embodiments, the terminal device 200 uses an eSIM, i.e., an embedded SIM card. The eSIM card can be embedded in the terminal device 200 and cannot be separated from the terminal device 200. The SIM card is used to identify the user's mobile phone number.

[0155] In addition to the SIM card interface 270, the terminal device 200 may also include a USB (universal serial bus) interface. The USB interface is used to connect a charging cable or other peripherals. For example, the terminal device 200 can connect a charging cable via the USB interface. The various components or modules in the terminal device 200 are connected via a bus.

[0156] The terminal device 200 may also include at least one sensor 280, such as an accelerometer 281, a proximity sensor 282, a fingerprint sensor 283, and a temperature sensor 284. The terminal device 200 may also be equipped with other sensors such as a gyroscope, barometer, hygrometer, thermometer, infrared sensor, light sensor, and motion sensor. For example, the fingerprint sensor 283 can be used to sense when a user clicks an icon on the terminal device 200's user interface.

[0157] The terminal device 200 may also include a camera for capturing still images or video. There may be one or more cameras. An object is projected onto a photosensitive element through a lens, generating an optical image. The photosensitive element may be a charge-coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) phototransistor. The photosensitive element converts the light signal into an electrical signal, which is then transmitted to the processor 220 for conversion into a digital image signal.

[0158] The processor 220 is the control center of the terminal device 200, connecting various parts of the terminal through various interfaces and lines. It executes various functions of the terminal device 200 and processes data by running or executing software programs stored in the memory 230 and calling data stored in the memory 230. In some embodiments, the processor 220 may include one or more processing units. In this application, the processor 220 can run an operating system, applications, user interface display and touch response, and the intent recognition method described in the embodiments of this application.

[0159] In some possible implementations, various aspects of the intent recognition method provided in this application can also be implemented in the form of a program product, which includes a computer program. When the program product is run on an electronic device, the computer program causes the electronic device to perform the steps in the intent recognition method according to the various exemplary embodiments of this application described above. For example, the electronic device can perform actions such as... Figure 2 The steps are shown in the figure.

[0160] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0161] The program product of the embodiments of this application may employ a portable compact disc read-only memory (CD-ROM) and include a computer program, and may run on an electronic device. However, the program product of this application is not limited thereto. In this document, the readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with a command execution system, apparatus, or device.

[0162] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a readable computer program. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with a command execution system, apparatus, or device.

[0163] Computer programs contained on readable media may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0164] Computer programs for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The computer program can execute entirely on the user's electronic device, partially on the user's device, as a standalone software package, partially on the user's electronic device and partially on a remote electronic device, or entirely on a remote electronic device or server. In cases involving remote electronic devices, the remote electronic device can be connected to the user's electronic device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external electronic device (e.g., via the Internet using an Internet service provider).

[0165] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0166] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0167] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing a computer-usable computer program.

[0168] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, produce a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0169] These computer program commands may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the commands stored in the computer-readable storage medium produce an article of manufacture including command means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0170] These computer program commands can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing the commands executed on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0171] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0172] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An intent recognition method, characterized in that, include: In response to an instruction to perform intent recognition on the text to be recognized, the text to be recognized is acquired; The first model is used to classify the text to be identified, and the similar text categories corresponding to the text to be identified are determined, wherein the second model is trained in advance using similar text samples from different similar text categories; Determine the second model corresponding to the similar text classification of the text to be identified; The second model is used to perform intent recognition on the text to be recognized, thereby obtaining the target intent corresponding to the text to be recognized; The second model is trained with the goal of minimizing the sum of the first loss function and the second loss function, wherein the first loss function is determined based on the intent recognition result and the intent label, and the second loss function is determined based on the information extraction result and the effective information label. The valid information includes key information associated with the intent; The key information of the text to be identified is matched with reference texts corresponding to different fields that are pre-stored in the database; When it is determined that the domain corresponding to the matched reference text is different from the domain corresponding to the target intent, the intent corresponding to the domain corresponding to the matched reference text is re-determined among the different intents corresponding to the similar text classification, based on the domain corresponding to the matched reference text.

2. The method according to claim 1, characterized in that, The different similar text classifications are based on the following method: Obtain multiple text samples with different intentions and calculate the similarity between the different text samples; Text samples with a similarity greater than a first preset threshold are identified as similar text samples in the same group.

3. The method according to claim 2, characterized in that, The first model was trained in the following manner: Label similar text samples in the same group with the corresponding group labels; Similar text samples from different groups are input into the first model, and the first model is trained with the goal of outputting the corresponding group labels.

4. The method according to any one of claims 1 to 3, characterized in that, When using the second model to perform intent recognition on the text to be recognized, the method further includes: The second model is used to extract information from the text to be identified, thereby obtaining the effective information of the text to be identified.

5. The method according to claim 1, characterized in that, Matching the key information of the text to be identified with reference texts corresponding to different fields pre-stored in the database includes: Obtain the first domain corresponding to the target intent, and match the key information with the reference text corresponding to the first domain in the database; When it is determined that no reference text corresponding to the key information is found, the key information is matched with reference texts corresponding to other fields in the database besides the first field.

6. An intent recognition device, characterized in that, include: The acquisition module is used to acquire the text to be identified in response to an instruction to perform intent recognition on the text to be identified; The classification module is used to classify the text to be identified using the first model and determine the similar text classification corresponding to the text to be identified, wherein the corresponding second model is trained in advance using similar text samples from different similar text classifications; The determination module is used to determine the second model corresponding to the similar text classification of the text to be identified; The recognition module is used to perform intent recognition on the text to be recognized using the second model, and to obtain the target intent corresponding to the text to be recognized; The second model is trained with the goal of minimizing the sum of the first loss function and the second loss function, wherein the first loss function is determined based on the intent recognition result and the intent label, and the second loss function is determined based on the information extraction result and the effective information label. The valid information includes key information associated with the intent; The key information of the text to be identified is matched with reference texts corresponding to different fields that are pre-stored in the database; When it is determined that the domain corresponding to the matched reference text is different from the domain corresponding to the target intent, the intent corresponding to the domain corresponding to the matched reference text is re-determined among the different intents corresponding to the similar text classification, based on the domain corresponding to the matched reference text.

7. An intent recognition device, characterized in that, It includes a processor and a memory, the memory storing instructions executable by the at least one processor, which, when executed by the processor, implement the steps as described in any one of claims 1-5.

8. A computer storage medium, characterized in that, The computer storage medium stores a computer program that causes the computer to perform the steps as described in any one of claims 1-5.

Images