Information processing method, device and computer readable storage medium

Abstract

The present disclosure relates to an information processing method, device and computer-readable storage medium, and relates to the field of computer technology. The method of the present disclosure includes: obtaining text corresponding to user input information; identifying the industry category corresponding to the text according to the text; inputting the text into a natural language processing model corresponding to the industry category to obtain output information corresponding to the text; wherein different industry categories correspond to different natural language processing models.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to an information processing method, device and computer-readable storage medium. Background Art

[0002] With the development of computer technology, the application of artificial intelligence is becoming more and more common in all walks of life. For example, human-computer interaction technologies such as intelligent customer service, voice recognition, intelligent calling, and machine translation are being applied in more and more fields.

[0003] At present, in the field of human-computer interaction, natural language processing technology is usually used to identify user intentions for user input information, and then automatically give corresponding responses. Summary of the invention

[0004] The inventors found that when human-computer interaction technology is applied to cross-industry scenarios, the information input by the user is usually input into a common natural language processing model for processing. However, due to the specialization of the industry and the intersection of words between industries, robots often have understanding accuracy deviations and understanding ambiguity when performing semantic intent recognition.

[0005] A technical problem to be solved by the present disclosure is: how to improve the accuracy of human-computer interaction in cross-industry application scenarios.

[0006] According to some embodiments of the present disclosure, an information processing method is provided, including: obtaining text corresponding to user input information; identifying the industry category corresponding to the text based on the text; inputting the text into a natural language processing model corresponding to the industry category to obtain output information corresponding to the text; wherein different industry categories correspond to different natural language processing models.

[0007] In some embodiments, based on the text, identifying the industry category corresponding to the text includes: after identifying the industry category corresponding to the text, identifying the professional category corresponding to the text under the industry type based on the text; inputting the text into a natural language processing model corresponding to the industry category includes: inputting the text into a natural language processing model corresponding to the industry category and the professional category; wherein, one industry category corresponds to multiple professional categories.

[0008] In some embodiments, based on the text, identifying the industry category corresponding to the text includes: segmenting the text, using a named entity recognition model to determine the nouns and verbs in the text, comparing the nouns and verbs in the text with the words in the dictionary of each industry category, and determining the industry category corresponding to the text based on the comparison results; or, inputting the text into a pre-trained industry classification model to obtain the industry category corresponding to the text.

[0009] In some embodiments, based on the text, identifying the professional category corresponding to the text includes: segmenting the text; determining the nouns and verbs in the text using a named entity recognition model; comparing the nouns and verbs in the text with the words in the dictionary of each professional category; determining the professional category corresponding to the text based on the comparison results; or, inputting the text into a pre-trained professional classification model to obtain the professional category corresponding to the text.

[0010] In some embodiments, determining the professional category corresponding to the text based on the comparison results includes: determining the professional category corresponding to the text based on the number of times the above-mentioned nouns and verbs match the same words in the dictionaries of various professional categories, and the weights of each word in the dictionaries of various professional categories.

[0011] In some embodiments, in the field of human-computer interaction, the output information is a response to user input information; the method further includes: converting the output information into speech.

[0012] In some embodiments, the method further includes: when the industry category corresponding to the text cannot be identified, inputting the text into a preset general natural language processing model.

[0013] According to some other embodiments of the present disclosure, an information processing device is provided, including: an acquisition module for acquiring text corresponding to user input information; an identification module for identifying, based on the text, the industry category corresponding to the text; and a control module for inputting the text into a natural language processing model corresponding to the industry category to obtain output information corresponding to the text; wherein different industry categories correspond to different natural language processing models.

[0014] In some embodiments, the recognition module is used to identify the professional category corresponding to the text under the industry type after identifying the industry category corresponding to the text; the control module is used to input the text into a natural language processing model corresponding to the industry category and the professional category; wherein one industry category corresponds to multiple professional categories.

[0015] In some embodiments, the recognition module is used to segment the text, use a named entity recognition model to determine the nouns and verbs in the text, compare the nouns and verbs in the text with the words in the dictionary of each industry category, and determine the industry category to which the text corresponds based on the comparison results; alternatively, input the text into a pre-trained industry classification model to obtain the industry category to which the text corresponds.

[0016] In some embodiments, the recognition module is used to segment the text; use a named entity recognition model to determine the nouns and verbs in the text; compare the nouns and verbs in the text with the words in the dictionary of each professional category; determine the professional category corresponding to the text based on the comparison results; or, input the text into a pre-trained professional classification model to obtain the professional category corresponding to the text.

[0017] In some embodiments, the recognition module is used to determine the professional category corresponding to the text based on the number of times the nouns and verbs are matched to the same words in the dictionaries of various professional categories and the weights of each word in the dictionaries of various professional categories.

[0018] In some embodiments, in the field of human-computer interaction, the output information is a response to the user's input information; the device also includes: a conversion module, which is used to convert the output information into speech.

[0019] In some embodiments, the control module is further configured to input the text into a preset general natural language processing model when the industry category corresponding to the text cannot be identified.

[0020] According to some further embodiments of the present disclosure, an information processing device is provided, comprising: a processor; and a memory coupled to the processor, for storing instructions, wherein when the instructions are executed by the processor, the processor executes the information processing method as described in any of the aforementioned embodiments.

[0021] According to some further embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, wherein the program, when executed by a processor, implements the steps of the information processing method of any of the aforementioned embodiments.

[0022] The present disclosure obtains the corresponding text for the information input by the user, identifies the industry category corresponding to the text according to the text, and then inputs the text into the natural language processing model corresponding to the industry category to obtain the output information corresponding to the text. The present disclosure can be applied to cross-industry human-computer interaction scenarios, and different natural language processing models can be set for different industry categories to improve the accuracy of user intent recognition, enhance the accuracy of semantic recognition, reduce the ambiguity of semantic understanding between industries, and facilitate the access of semantic understanding in multiple industries to improve application efficiency.

[0023] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A flowchart illustrating an information processing method according to some embodiments of the present disclosure is provided.

[0026] Figure 2 A schematic flowchart showing information processing methods according to other embodiments of the present disclosure is provided.

[0027] Figure 3 A schematic diagram showing the overall architecture of some embodiments of the present disclosure is shown.

[0028] Figure 4 A schematic diagram showing the structure of an information processing device according to some embodiments of the present disclosure.

[0029] Figure 5 A schematic diagram showing the structure of an information processing device according to some other embodiments of the present disclosure.

[0030] Figure 6 A schematic diagram showing the structure of an information processing device according to some further embodiments of the present disclosure. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present disclosure and its application or use. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0032] The present disclosure provides an information processing method, Figure 1 Give a description.

[0033] Figure 1 Flowcharts of some embodiments of the information processing method disclosed in the present invention. Figure 1 As shown, the method of this embodiment includes: steps S102 to S106.

[0034] In step S102, the text corresponding to the user input information is obtained.

[0035] The information input by the user can be in the form of text, voice, etc. In the case where the information input by the user is text, the user's voice can be converted into corresponding text. For example, the user's voice is first preprocessed, for example, VAD (voice activity detection) is performed, the silent parts at the beginning and end are cut off, noise is removed, etc. Further, the preprocessed voice is framed, for example, a moving window function is used for framing. After that, the feature vector of each frame is extracted, for example, the linear prediction cepstral coefficient (LPCC) or Mel cepstral coefficient (MFCC) of each frame of voice is extracted. The feature vector of each frame is input into the acoustic model to obtain the output phoneme information, for example, the hidden Markov model (HMM). According to the phoneme information, the word or word corresponding to the factor information is searched in the dictionary. The word or word corresponding to each frame is input into the language model to obtain the output text.

[0036] In step S104, the industry category corresponding to the text is identified based on the text.

[0037] Industry categories include, for example, telecommunications industry, medical industry, e-commerce industry, etc., which can be divided according to actual needs and are not limited to the examples given. Unique identifiers can be set for different industry categories. After identifying the industry category corresponding to the text, the text can be associated with the identifier of the industry category.

[0038] In some embodiments, corresponding dictionaries may be pre-set for different industry categories and stored in a database, and the dictionaries include characteristic words for the industry categories. The words in the text are matched with the words in the dictionaries corresponding to the various industry categories, and the corresponding industry category of the text is determined based on the matching results. For example, the text is segmented; each segmented word in the text is compared with the words in the dictionaries corresponding to the various industry categories. The industry category corresponding to the text is determined based on the number of times the segmented words in the text match the same words in the dictionaries corresponding to the various industry categories, and the weight values ​​of each word in the dictionaries corresponding to the various industry categories.

[0039] Furthermore, some words in the text are useless for identifying industry categories, such as stop words, place names, etc. Therefore, these words can be removed and the remaining words can be used to determine the industry category corresponding to the text according to the above method. The named entity recognition model can also be used to determine the nouns and verbs in the text, match these nouns and verbs with the words in the dictionary corresponding to various industry categories, and determine the corresponding industry category of the text based on the matching results. For example, the industry category corresponding to the text can be determined based on the number of times these nouns and verbs match the same words in the dictionaries corresponding to various industry categories, and the weight values ​​of each word in the dictionaries corresponding to various industry categories. Furthermore, the following formula can be used to calculate the association between the text and each industry category.

[0040]

[0041] In formula (1), P j represents the correlation between the text and industry category j, 1≤j≤N, j is a positive integer, N is the number of industry categories, 1≤i≤n, i is a positive integer, i represents the number of nouns, n is the number of nouns, K ij represents the weight of the same words that noun i matches in the dictionary of industry category j, N ij G represents the number of times noun i appears in the text, that is, the number of times noun i matches the same word in the dictionary corresponding to industry category j; 1≤k≤m, k is a positive integer, k represents the number of verbs, m is the number of verbs, kj represents the weight of the same words matched by verb k in the dictionary of industry category j, M kj It indicates the number of times verb k appears in the text, that is, the number of times verb k matches the same word in the dictionary corresponding to industry category j. If verbs and nouns are not distinguished, the above formula can be transformed accordingly, which will not be repeated here. According to the above formula, the correlation between the text and each industry category is obtained, and the industry category with the largest correlation is further determined as the professional category corresponding to the text.

[0042] Different words have different weights in the dictionary of industry categories. The greater the degree of distinction between industry categories, the greater the weight of the words. For example, for the telecommunications industry, the weights of words such as bandwidth and delay may be relatively large. In some embodiments, the construction of dictionaries corresponding to different industry categories can be obtained by training based on a large number of texts of the industry category. For an industry category, a large number of texts corresponding to the industry category are used as training texts, and the training texts are segmented. For each segmentation, the TF-IDF (term frequency-inverse document frequency) of the segmentation in all training texts can be determined, that is, all training texts are taken as a whole, the TF-IDF of the segmentation is calculated, and the segmentation with a TF-IDF value greater than the threshold is used as the corresponding word in the dictionary. Further, different position weights are set for the position where the segmentation appears in the text. For the same segmentation, the position weights of the segmentation in each training text are summed, and then the sum of the position weights is multiplied by the TF-IDF value of the segmentation to obtain the weight of the segmentation for the industry category.

[0043] In some embodiments, an industry classification model can be pre-trained, such as a neural network model, and the text can be input into the trained neural network model to obtain the industry category corresponding to the text. Texts of different industry categories can be selected for corresponding industry category annotation to generate training samples. The training samples are input into the industry classification model to be trained, and the loss function value is calculated according to the output results and annotations. The parameters of the industry classification model are adjusted according to the loss function until the preset conditions are met (for example, the loss function value is minimum or lower than a threshold, etc.), and a trained industry classification model is obtained. The industry classification model can adopt an existing classification model, which will not be repeated here.

[0044] Furthermore, the above two implementations can be combined with each other. For example, according to the first implementation, the correlation between the text and each industry category can be obtained, and according to the second implementation, the probability that the text belongs to each industry category can be obtained using the industry classification model. For each industry category, the correlation between the text and the industry category and the probability that the text belongs to the industry category are weighted to obtain the matching value of the text relative to the industry category, and the industry category with the largest matching value is selected as the industry category corresponding to the text.

[0045] In step S106, the text is input into a natural language processing model corresponding to the industry category to obtain output information corresponding to the text.

[0046] According to the aforementioned embodiment, the industry category corresponding to the text can be obtained, and different industry categories correspond to different natural language processing models, that is, different natural language processing models are trained according to training texts of different industry categories. In the natural language processing model, the semantic understanding of the text for this industry is more accurate, precise intent recognition can be performed, and the output information obtained is more accurate. In the field of human-computer interaction, the output information is a response to the user's input information, for example, an answer to an input question, or a corresponding translation for the input language. For example, in a scenario where the input information is a question and the output information is an answer, the natural language processing model can be an existing model such as QANet, DSSM (deep structure semantic model), which will not be repeated here. The output information can also be converted into speech for output.

[0047] The method of the above embodiment obtains the corresponding text for the information input by the user, identifies the industry category corresponding to the text based on the text, and then inputs the text into the natural language processing model corresponding to the industry category to obtain output information corresponding to the text. The method of the above embodiment can be applied to cross-industry human-computer interaction scenarios, and different natural language processing models can be set for different industry categories to improve the accuracy of user intent recognition, enhance the accuracy of semantic recognition, reduce the ambiguity of semantic understanding between industries, and facilitate the access of semantic understanding in multiple industries to improve application efficiency.

[0048] Combine the following Figure 2 Other embodiments of the information processing method disclosed herein are described.

[0049] Figure 2 Flowcharts of other embodiments of the information processing method disclosed herein. Figure 2 As shown, the method of this embodiment includes: steps S202 to S208.

[0050] In step S202, the text corresponding to the user input information is obtained.

[0051] In step S204, the industry category corresponding to the text is identified based on the text.

[0052] In step S206, the professional category corresponding to the text under the industry type is identified based on the text.

[0053] One industry category corresponds to multiple professional categories. For example, in the telecommunications industry, it can include voice service profession, broadband service profession, etc., or it can include: service activation, complaint, consultation and other professions. The professional categories and division methods can be set according to actual needs, not limited to the examples given. Professional segmentation can further reduce the ambiguity of semantic understanding and improve the accuracy of output information.

[0054] The professional category of the text can be determined by using a keyword recognition method to extract the keywords of the text and match them with the preset keywords (keyword dictionary) of each professional category to determine the professional category of the text. For example, the complaint category and the consultation category have different keywords. Keyword extraction can use existing algorithms, which will not be repeated here.

[0055] The professional category of the text can also be determined by a method similar to that of determining the professional category. For example, corresponding dictionaries can be set up in advance for different professional categories and stored in a database, and the dictionaries include characteristic words of the professional category. The words in the text are matched with the words in the dictionary corresponding to various professional categories, and the corresponding professional category of the text is determined based on the matching results. For example, the text is segmented; each segmented word in the text is compared with the words in the dictionary corresponding to various professional categories. The professional category corresponding to the text is determined based on the number of times the segmented words in the text match the same words in the dictionaries corresponding to various professional categories, and the weight values ​​of each word in the dictionaries corresponding to various professional categories.

[0056] Furthermore, some words in the text have no effect on identifying professional categories, such as stop words, place names, etc. Therefore, these words can be removed and the remaining words can be used according to the above method to determine the professional category corresponding to the text. The named entity recognition model can also be used to determine the nouns and verbs in the text, match these nouns and verbs with the words in the dictionary corresponding to various professional categories, and determine the corresponding professional category of the text based on the matching results. For example, the professional category corresponding to the text can be determined based on the number of times these nouns and verbs match the same words in the dictionaries corresponding to various professional categories, and the weight values ​​of each word in the dictionaries corresponding to various professional categories. Furthermore, the following formula can be used to calculate the correlation between the text and each professional category.

[0057]

[0058] In formula (2), P lrepresents the relevance between the text and the professional category l, 1≤l≤M, l is a positive integer, M is the number of professional categories, 1≤i≤n, i is a positive integer, i represents the number of nouns, n is the number of nouns, K il represents the weight of the same words that noun i matches in the dictionary of professional category l, N il represents the number of times noun i appears in the text, that is, the number of times noun i matches the same word in the dictionary corresponding to professional category l; 1≤k≤m, k is a positive integer, k represents the number of verbs, m is the number of nouns, G kl represents the weight of the same words matched by verb k in the dictionary of professional category l, M kl It indicates the number of times verb k appears in the text, that is, the number of times verb k matches the same word in the dictionary corresponding to professional category l. If verbs and nouns are not distinguished, the above formula can be transformed accordingly, which will not be repeated here. According to the above formula, the correlation between the text and each professional category is obtained, and the professional category with the largest correlation is further determined as the professional category corresponding to the text.

[0059] In some embodiments, the dictionary corresponding to different professional categories can be trained based on a large number of texts in the professional category. For a professional category, a large number of texts corresponding to the professional category are used as training texts, and the training texts are segmented. For each segmentation, the TF-IDF (term frequency-inverse document frequency) of the segmentation in all training texts can be determined, and the segmentation with a TF-IDF value greater than a threshold is used as the corresponding word in the dictionary. Furthermore, different position weights are set for the position where the segmentation appears in the text. For the same segmentation, the position weights of the segmentation in each training text are summed, and then the sum of the position weights is multiplied by the TF-IDF value of the segmentation to obtain the weight of the segmentation for the professional category.

[0060] Since professional categories are further subdivided into industry categories, the words in the professional category dictionary belong to the dictionary corresponding to the industry category. You can first generate the dictionary corresponding to each professional category according to the above method, and then combine the dictionaries corresponding to each professional category under the same industry category to obtain the dictionary corresponding to the industry category. However, the dictionary corresponding to the general industry category may contain a wider range, not only the words corresponding to each professional category. Therefore, after determining the dictionary corresponding to the industry category, you can judge which professional category each word belongs to, and then generate the dictionary of the professional category.

[0061] Generate a training sample set for each professional category, and determine the professional category to which the word belongs based on the position and number of times each word in the dictionary corresponding to the industry category appears in each training sample set. First, select a word from the dictionary corresponding to the industry category, match the word with a training sample set corresponding to a professional category, determine the position of the word in each training sample and the number of times each position appears. Different positions correspond to different position weights. Weight each position weight and the number of times each position appears to obtain the degree of membership of the word relative to the professional category. Select the professional category with the largest degree of membership or the degree of membership higher than the degree of membership threshold as the professional category corresponding to the word, and add the word to the corresponding dictionary. If the selection is based on the degree of membership higher than the degree of membership threshold, a word may be assigned to multiple dictionaries, which is also normal. There may be overlapping words in different professional categories.

[0062] The above embodiment can also be modified as follows: first, a word is selected from the dictionary corresponding to the industry category, the word is matched with a training sample set corresponding to a professional category, the position where the word appears in each training sample and the TF-IDF value of the word in the training sample set are determined, different positions correspond to different position weights, the weights of each position are summed, the sum is multiplied by the TF-IDF value, and the membership of the word relative to the professional category is obtained. The professional category with the largest membership or a membership higher than the membership threshold is selected as the professional category corresponding to the word, and the word is added to the corresponding dictionary.

[0063] Furthermore, the membership degree in the above embodiment can be used as the weight of the word segmentation for the professional category, or the word segmentations can be sorted from large to small according to the membership degree and different weights can be set for each word segmentation.

[0064] In some embodiments, a professional classification model can be pre-trained, such as a neural network model, and the text can be input into the trained neural network model to obtain the professional category corresponding to the text. Texts of different professional categories can be selected for corresponding professional category annotation to generate training samples. The training samples are input into the professional classification model to be trained, and the loss function value is calculated according to the output results and annotations. The parameters of the professional classification model are adjusted according to the loss function until the preset conditions are met (for example, the loss function value is minimum or lower than a threshold, etc.), and a trained professional classification model is obtained. The professional classification model can adopt an existing classification model, which will not be repeated here.

[0065] Furthermore, the above two implementations can be combined with each other. For example, according to the first implementation, the correlation between the text and each professional category can be obtained, and according to the second implementation, the probability that the text belongs to each professional category can be obtained using the professional classification model. For each professional category, the correlation between the text and the professional category and the probability that the text belongs to the professional category are weighted to obtain the matching value of the text relative to the professional category, and the one with the largest matching value is selected as the professional category corresponding to the text.

[0066] Several embodiments of how to determine the professional category corresponding to the text and several embodiments of how to determine the industry category corresponding to the text can be arbitrarily combined and will not be described in detail here.

[0067] In step S208, the text is input into a natural language processing model corresponding to the industry category and the professional category to obtain output information corresponding to the text.

[0068] Different industry categories and different professional categories correspond to different natural language processing models. After identifying the industry category and professional category of the text, the text can be associated with the industry identifier and professional identifier, and the corresponding natural language processing model can be further found based on the industry identifier and professional identifier.

[0069] In some embodiments, when the industry category corresponding to the text cannot be identified, the text is input into a preset general natural language processing model. Alternatively, when the industry category and professional category corresponding to the text cannot be identified, the text is input into a preset general natural language processing model.

[0070] Based on the above embodiments, the main structure of the present disclosure is as follows: Figure 3 As shown. First, the user inputs information and converts it into corresponding text. Then the industry category and professional category are divided. Furthermore, NLP access control is performed according to the professional category and industry category, and different NLPs are accessed. If the industry category and professional category are not identified, the general NLP can be accessed, and finally the NLP outputs the result.

[0071] The disclosed solution performs industry understanding of the input information in advance, selects NLP of the appropriate industry for semantic understanding, improves the accuracy of semantic understanding, and reduces ambiguity. Pre-judgment of professional categories, thereby predicting industry subdivisions, facilitates the next step of detailed judgment, and improves the accuracy of semantic understanding. The disclosed solution effectively improves the accuracy and efficiency of semantic intent recognition in various complex scenarios in reality. Compared with other similar technologies, it first uses industry understanding and professional understanding to determine the industry and subdivision of the content, and then uses the corresponding NLP for recognition and understanding to improve the accuracy of recognition. It applies unified NLP access and control to facilitate semantic understanding access in multiple industries to improve application efficiency.

[0072] The present disclosure also provides an information processing device, Figure 4 Give a description.

[0073] Figure 4 Figure 1 is a structural diagram of some embodiments of the information processing device disclosed in the present invention. Figure 4 As shown, the device 40 of this embodiment includes: an acquisition module 410 , an identification module 420 , and a control module 430 .

[0074] The acquisition module 410 is used to acquire text corresponding to the user input information.

[0075] The identification module 420 is used to identify the industry category corresponding to the text based on the text.

[0076] In some embodiments, the recognition module 420 is used to identify the professional category corresponding to the text under the industry type after identifying the industry category corresponding to the text; the control module is used to input the text into a natural language processing model corresponding to the industry category and the professional category; wherein one industry category corresponds to multiple professional categories.

[0077] In some embodiments, the recognition module 420 is used to segment the text, use a named entity recognition model to determine the nouns and verbs in the text, compare the nouns and verbs in the text with the words in the dictionary of each industry category, and determine the industry category to which the text corresponds based on the comparison results; alternatively, input the text into a pre-trained industry classification model to obtain the industry category to which the text corresponds.

[0078] In some embodiments, the recognition module 420 is used to segment the text; use a named entity recognition model to determine the nouns and verbs in the text; compare the nouns and verbs in the text with the words in the dictionary of each professional category; determine the professional category corresponding to the text based on the comparison results; or, input the text into a pre-trained professional classification model to obtain the professional category corresponding to the text.

[0079] In some embodiments, the recognition module 420 is used to determine the professional category corresponding to the text based on the number of times the nouns and verbs are matched to the same words in the dictionaries of various professional categories and the weights of each word in the dictionaries of various professional categories.

[0080] The control module 430 is used to input the text into a natural language processing model corresponding to the industry category to obtain output information corresponding to the text.

[0081] In some embodiments, the control module 430 is further configured to input the text into a preset general natural language processing model when the industry category corresponding to the text cannot be identified.

[0082] In some embodiments, different industry categories correspond to different natural language processing models.

[0083] In some embodiments, in the field of human-computer interaction, the output information is a response to the user input information; the device also includes: a conversion module 440, which is used to convert the output information into speech.

[0084] The information processing apparatus in the embodiments of the present disclosure may be implemented by various computing devices or computer systems. Figure 5 as well as Figure 6 Give a description.

[0085] Figure 5 Figure 1 is a structural diagram of some embodiments of the information processing device disclosed in the present invention. Figure 5 As shown, the device 50 of this embodiment includes: a memory 510 and a processor 520 coupled to the memory 510, and the processor 520 is configured to execute the information processing method in any of some embodiments of the present disclosure based on the instructions stored in the memory 510.

[0086] The memory 510 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, a database, and other programs.

[0087] Figure 6 FIG. 1 is a structural diagram of some other embodiments of the information processing device disclosed in the present invention. Figure 6 As shown, the device 60 of this embodiment includes: a memory 610 and a processor 620, which are similar to the memory 510 and the processor 520 respectively. It can also include an input and output interface 630, a network interface 640, a storage interface 650, etc. These interfaces 630, 640, 650 and the memory 610 and the processor 620 can be connected, for example, through a bus 660. Among them, the input and output interface 630 provides a connection interface for input and output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 640 provides a connection interface for various networked devices, for example, it can be connected to a database server or a cloud storage server. The storage interface 650 provides a connection interface for external storage devices such as SD cards and USB flash drives.

[0088] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0092] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. An information processing method, comprising: Get the text corresponding to the user input information; According to the text, the industry category corresponding to the text is identified, including: segmenting the text, using a named entity recognition model to determine nouns and verbs in the text, comparing the nouns and verbs in the text with words in dictionaries of various industry categories, determining the association between the text and each industry category according to the number of times the nouns and verbs match the same words in the dictionaries corresponding to various industry categories, and the weight values ​​of each word in the dictionaries corresponding to various industry categories, inputting the text into a pre-trained industry classification model to obtain the probability that the text belongs to each industry category, weighting the association between the text and each industry category and the probability that the text belongs to each industry category to obtain the matching value of the text relative to each industry category, and selecting the matching The largest value is used as the industry category corresponding to the text, and the professional category corresponding to the text under the industry type is identified according to the text, wherein different words have different weights in the dictionary of industry categories, and the weight of the word with greater discrimination for the industry category is greater. One industry category corresponds to multiple professional categories. For one industry category, the text corresponding to the industry category is used as the training text, and the training text is segmented. For each segmented word, the word frequency-inverse document frequency of the segmented word in all training texts is determined, and different position weights are set for the position where the segmented word appears in the text. For the same segmented word, the position weights of the segmented word in each training text are summed, and then the sum of the position weights is multiplied by the word frequency-inverse document frequency value of the segmented word to obtain the weight of the segmented word for the industry category; Inputting the text into a natural language processing model corresponding to the industry category and the professional category to obtain output information corresponding to the text; Among them, different industry categories correspond to different natural language processing models.

2. The information processing method according to claim 1, wherein: The identifying, based on the text, the professional category corresponding to the text comprises: Segmenting the text; determining nouns and verbs in the text using a named entity recognition model; comparing the nouns and verbs in the text with words in dictionaries of various professional categories; and determining the professional category corresponding to the text based on the comparison results; Alternatively, the text is input into a pre-trained professional classification model to obtain the professional category corresponding to the text.

3. The information processing method according to claim 2, wherein: Determining the professional category corresponding to the text according to the comparison result includes: The professional category corresponding to the text is determined according to the number of times the nouns and verbs match the same words in the dictionaries of various professional categories and the weights of each word in the dictionaries of various professional categories.

4. The information processing method according to claim 1, wherein: In the field of human-computer interaction, the output information is a response to the user's input information; The method further comprises: The output information is converted into speech.

5. The information processing method according to claim 1, further comprising: When the industry category corresponding to the text cannot be identified, the text is input into a preset general natural language processing model.

6. An information processing device, comprising: The acquisition module is used to obtain the text corresponding to the user input information; The recognition module is used to identify the industry category corresponding to the text according to the text, including: segmenting the text, using a named entity recognition model to determine the nouns and verbs in the text, comparing the nouns and verbs in the text with the words in the dictionary of each industry category, determining the association between the text and each industry category according to the number of times the nouns and verbs match the same words in the dictionaries corresponding to the various industry categories and the weight values ​​of each word in the dictionaries corresponding to the various industry categories, inputting the text into a pre-trained industry classification model to obtain the probability that the text belongs to each industry category, weighting the association between the text and each industry category and the probability that the text belongs to each industry category to obtain the matching value of the text relative to each industry category, and selecting The largest matching value is taken as the industry category corresponding to the text, and the professional category corresponding to the text under the industry type is identified according to the text, wherein different words have different weights in the dictionary of industry categories, and the weight of the word with greater discrimination for the industry category is greater. One industry category corresponds to multiple professional categories. For one industry category, the text corresponding to the industry category is used as the training text, and the training text is segmented. For each segmented word, the word frequency-inverse document frequency of the segmented word in all training texts is determined, and different position weights are set for the position where the segmented word appears in the text. For the same segmented word, the position weights of the segmented word in each training text are summed, and then the sum of the position weights is multiplied by the word frequency-inverse document frequency value of the segmented word to obtain the weight of the segmented word for the industry category; A control module, used for inputting the text into a natural language processing model corresponding to the industry category and the professional category to obtain output information corresponding to the text; Among them, different industry categories correspond to different natural language processing models.

7. An information processing device, comprising: processor; as well as A memory coupled to the processor, for storing instructions, wherein when the instructions are executed by the processor, the processor executes the information processing method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein: When the program is executed by a processor, the steps of the information processing method described in any one of claims 1 to 5 are implemented.

Images