Slot keyword identification method and device, equipment, medium and product

Abstract

The invention discloses a slot keyword recognition method and device, equipment, a medium and a product, and relates to the technical field of keyword recognition, and the method comprises the steps: obtaining user request information of a target user; based on a large model cue word, extracting a keyword from the user request information through a large model; judging whether the keyword exists in a target slot lexicon or not; when it is judged that the keyword does not exist in the target slot lexicon, screening out a candidate word set related to the keyword from the target slot lexicon; and determining a target candidate word in the candidate word set, and taking the target candidate word as a slot keyword. According to the method and the device, the slot position keyword in the user request information can be accurately identified.

Description

Technical Field

[0001] The present application relates to the field of keyword recognition technology, and in particular to slot keyword recognition methods, devices, equipment, media, and products. Background Art

[0002] During human-machine conversations, the machine needs to understand the meaning of the conversation. Currently, structured representations of intent and slots are commonly used to represent the semantic information of user conversations. Therefore, accurately extracting slot information from conversations is a key research topic. Related technologies typically use deep learning to identify slot information in conversations. Deep learning models typically require large amounts of annotated data for training to ensure that the model can learn sufficient features and generalize to new conversational scenarios. However, in practice, obtaining large amounts of high-quality annotated data is often costly and time-consuming.

[0003] The above content is only used to assist in understanding the technical solution of this application and does not constitute an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a slot keyword recognition method, device, equipment, medium and product, aiming to solve the technical problem of high cost and time-consuming to identify slot information in human-computer dialogue based on deep learning.

[0005] To achieve the above objectives, the present application proposes a slot keyword recognition method, which includes:

[0006] Obtain user request information of the target user;

[0007] Based on the large model prompt words, extract keywords from the user request information through the large model;

[0008] Determine whether the keyword exists in the target slot vocabulary;

[0009] When it is determined that the keyword does not exist in the target slot vocabulary, a candidate word set related to the keyword is screened out from the target slot vocabulary;

[0010] Determine a target candidate word in the candidate word set, and use the target candidate word as a slot keyword.

[0011] In one embodiment, the step of extracting keywords from the user request information using the large model based on the large model prompt words includes:

[0012] Build large model prompt words based on slot extraction tasks, output requirements, example demonstrations, and context specifications;

[0013] Based on the large model prompt words, keywords are extracted from the user request information through the large model.

[0014] In one embodiment, before the step of determining whether the keyword exists in the target slot vocabulary, the method further includes:

[0015] Determining the target intent corresponding to the user request information;

[0016] Based on the target intent, determining an initial slot vocabulary corresponding to the keyword;

[0017] Determining the slot vocabulary permissions of the target user according to the identity information of the target user;

[0018] Determining whether the target user has permission for the initial slot vocabulary based on the slot vocabulary permission;

[0019] When it is determined that the target user has the authority of the initial slot vocabulary, the initial slot vocabulary is used as the target slot vocabulary.

[0020] In one embodiment, the step of determining a target candidate word in the candidate word set includes:

[0021] Determining semantic similarity and semantic relevance between the keyword and each candidate word in the candidate word set;

[0022] Determining a semantic similarity score based on the semantic similarity and the semantic relevance;

[0023] Determining whether there is a candidate word in the candidate word set whose semantic similarity score is greater than a preset threshold;

[0024] If so, the candidate word in the candidate word set whose semantic similarity score is greater than the preset threshold is taken as the target candidate word;

[0025] If not, update the target slot vocabulary based on the keyword.

[0026] In one embodiment, the step of updating the target slot vocabulary based on the keyword includes:

[0027] Marking the keyword as an unmatched keyword, and determining context information of the unmatched keyword;

[0028] Standardizing the unmatched keywords to generate candidate standard entries;

[0029] Submitting the candidate standard terms and the context information to the review platform for manual verification to obtain manual verification results;

[0030] When the manual verification result is verification passed, the mapping relationship between the unmatched keyword and the candidate standard entry is added to the target slot vocabulary.

[0031] In one embodiment, after the step of adding the mapping relationship between the unmatched keyword and the candidate standard term to the target slot vocabulary, the method further includes:

[0032] Real-time monitoring of the usage frequency and matching accuracy of the newly added entries in the target slot vocabulary;

[0033] The matching strategy of the newly added term is dynamically adjusted according to the usage frequency of the term and the matching accuracy, so as to perform slot identification on the new user request information according to the matching strategy.

[0034] In addition, to achieve the above-mentioned purpose, the present application also proposes a slot keyword recognition device, which includes:

[0035] The acquisition module is used to obtain the user request information of the target user;

[0036] An extraction module, configured to extract keywords from the user request information using a large model;

[0037] A judgment module, used to judge whether the keyword exists in the target slot vocabulary;

[0038] a screening module, configured to screen out a set of candidate words related to the keyword from the target slot vocabulary when it is determined that the keyword does not exist in the target slot vocabulary;

[0039] The determination module is used to determine a target candidate word in the candidate word set and use the target candidate word as a slot keyword.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a slot keyword recognition device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the slot keyword recognition method as described above.

[0041] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the slot keyword identification method described above are implemented.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the slot keyword identification method as described above.

[0043] One or more technical solutions proposed in this application have at least the following technical effects:

[0044] The slot keyword recognition method, device, equipment, medium and product proposed in this application obtain the user request information of the target user; based on the large model prompt word, extract the keyword from the user request information through the large model; determine whether the keyword exists in the target slot vocabulary; when it is determined that the keyword does not exist in the target slot vocabulary, filter out a set of candidate words related to the keyword from the target slot vocabulary; determine the target candidate word in the candidate word set, and use the target candidate word as the slot keyword. It solves the technical problem of high cost and time-consuming to identify slot information in human-computer dialogue based on deep learning. Compared with the existing technology, this application, based on the large model, introduces the large model prompt word technology to more efficiently extract keywords from user requests, reduces the dependence on large-scale labeled data, reduces the training cost, and further optimizes the slot keyword recognition process by dynamically screening the candidate word set and determining the target candidate word, improves the recognition efficiency and accuracy, and can quickly respond to user requests in a short time. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0047] Figure 1 A flowchart of the first embodiment of the slot keyword recognition method of this application is provided;

[0048] Figure 2 A flowchart of the second embodiment of the slot keyword identification method of this application is provided;

[0049] Figure 3 This is a schematic diagram of the module structure of the slot keyword recognition device according to an embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of the device structure of the hardware operating environment involved in the slot keyword identification method in the embodiment of the present application.

[0051] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0053] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0054] The main solution of the embodiment of the present application is: obtaining user request information of the target user; extracting keywords from the user request information through the big model based on the big model prompt word; judging whether the keyword exists in the target slot vocabulary; when it is determined that the keyword does not exist in the target slot vocabulary, screening out a set of candidate words related to the keyword from the target slot vocabulary; determining the target candidate word in the candidate word set, and using the target candidate word as the slot keyword.

[0055] As can be seen from the above embodiments, the present application obtains the user request information of the target user; based on the large model prompt word, extracts keywords from the user request information through the large model; determines whether the keyword exists in the target slot vocabulary; when it is determined that the keyword does not exist in the target slot vocabulary, filters out a set of candidate words related to the keyword from the target slot vocabulary; determines the target candidate word in the candidate word set, and uses the target candidate word as the slot keyword. This solves the technical problem of high cost and time-consuming to identify slot information in human-computer dialogue based on deep learning. Compared with the existing technology, this application, based on the large model, introduces the large model prompt word technology to more efficiently extract keywords from user requests, reduces dependence on large-scale labeled data, reduces training costs, and further optimizes the slot keyword recognition process by dynamically screening the candidate word set and determining the target candidate word, improves recognition efficiency and accuracy, and can quickly respond to user requests in a short time.

[0056] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device capable of performing the above functions, a slot keyword recognition device, etc. The following uses the slot keyword recognition device as an example to illustrate this embodiment and the following embodiments.

[0057] Based on this, the present application embodiment provides a slot keyword identification method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the slot keyword recognition method of the present application.

[0058] In this embodiment, the slot keyword identification method includes steps S10 to S50:

[0059] Step S10, obtaining user request information of the target user;

[0060] It should be noted that the user request information can be text information input by the user, or can be text converted from voice information input by the user through voice recognition.

[0061] Step S20, based on the large model prompt words, extract keywords from the user request information through the large model;

[0062] It should be noted that large models (such as those based on the Transformer architecture, such as BERT and GPT) can be used to extract key information from user input. Specifically, large models have learned a large amount of language patterns and semantic knowledge through pre-training, and can understand user input and extract key information based on the context. For example, through word embeddings and attention mechanisms, the model can identify important words in the input; large model prompts refer to input instructions and contextual information carefully designed to guide the large model to perform specific tasks. It is the core medium for interacting with the large model and is equivalent to a "task manual" for the large model.

[0063] In a feasible implementation, the step of extracting keywords from the user request information through a big model includes: constructing a big model prompt word according to slot extraction tasks, output requirements, example demonstrations, and context specifications; and extracting keywords from the user request information through a big model based on the big model prompt word.

[0064] It should be noted that the slot extraction task is used to clarify the specific tasks that the large model needs to complete, such as "extracting the name of the bank institution from the user query"; the output requirements are used to specify the output format and restrictions, such as "only output the standard full name, do not include any explanatory text"; the example demonstration is used to provide input-output examples; the context specification is used to define the processing scope (clarify the text range that the model should and should not pay attention to), eliminate ambiguity (that is, solve the problem of multiple meanings or vague expressions of a word), standardize the output (that is, ensure the consistency of the output format in different situations) and exception handling (that is, define how to deal with boundary cases).

[0065] In this example, a complete prompt word system is built based on slot extraction tasks, output requirements, example demonstrations, and contextual specifications. This enables the large model to accurately identify target keywords in complex natural language expressions, just like a professionally trained industry expert, while effectively avoiding common ambiguities and edge cases, and ultimately outputting standardized slot values ​​that meet business requirements. The advantage of this approach is that it retains the semantic understanding capabilities of the large language model while ensuring the accuracy and business compliance of the output results through rule constraints.

[0066] Step S30, determining whether the keyword exists in the target slot vocabulary;

[0067] It can be understood that the judgment operation is the process of comparing and verifying the keywords (such as "CMB", "ICBC", etc.) initially extracted from the user request information with the predefined standardized vocabulary (such as a database containing standard names such as "China Merchants Bank Co., Ltd." and "Industrial and Commercial Bank of China" and their aliases).

[0068] In its implementation, this judgment operation first performs an exact match to check whether the keyword exists directly in the vocabulary's standard values ​​or alias lists. If so, the keyword is directly used as the slot keyword. If not, a fuzzy matching mechanism is initiated, searching for similar terms by calculating the edit distance or using a Trie tree prefix match. If a match still cannot be found, the unrecognized keyword is recorded for subsequent vocabulary optimization, and a semantic calibration process is triggered. This process uses a semantic vector model to select the top 20 most similar candidate terms from the vocabulary for secondary matching. This judgment process not only verifies the business validity of the keyword but also automatically converts non-standard expressions to standard values ​​(for example, mapping "CMB" to "China Merchants Bank"). Dynamic vocabulary loading (based on user permission filtering) and cache optimization (using Redis to accelerate queries) balance recognition accuracy and system performance. This mechanism, serving as a quality check for slot filling, prevents invalid data from being stored and enables the vocabulary to evolve through the continuous accumulation of unrecognized cases.

[0069] Step S40, when it is determined that the keyword does not exist in the target slot vocabulary, a candidate word set related to the keyword is screened out from the target slot vocabulary;

[0070] It is understandable that in the slot keyword recognition process, in order to ensure that the candidate words screened out can effectively replace or supplement the semantic expression of the keywords, the target slot vocabulary can be screened from multiple dimensions. For example, the relevant candidate words can be screened out from the target slot vocabulary based on the semantic similarity, part-of-speech association or pronunciation similarity between the keywords and the words in the target slot vocabulary, and then a candidate word set is generated. Specifically, first, based on semantic similarity, the similarity between the keywords and the entries in the vocabulary is calculated by a semantic vector model, and candidate words with similar semantics are screened out. Secondly, the part-of-speech association is considered to ensure that the candidate words match the keywords in part of speech, such as noun to noun, verb to verb, to ensure semantic coherence. In addition, pronunciation similarity can also be introduced, and phonetic symbols or pinyin matching technology can be used to screen out entries with similar pronunciations, which is particularly suitable for processing transliterated words or colloquial expressions input by users. By comprehensively screening based on semantic similarity, part-of-speech association, and pronunciation similarity, the generated set of candidate words is not only rich and diverse, but also highly matches the keywords in semantics, grammar, and pronunciation, thus providing more accurate and flexible options for subsequent slot identification.

[0071] In a feasible implementation, before the step of determining whether the keyword exists in the target slot vocabulary, it also includes: determining the target intent corresponding to the user request information; determining the initial slot vocabulary corresponding to the keyword based on the target intent; determining the slot vocabulary permissions of the target user according to the identity information of the target user; determining whether the target user has the permissions of the initial slot vocabulary according to the slot vocabulary permissions; when it is determined that the target user has the permissions of the initial slot vocabulary, using the initial slot vocabulary as the target slot vocabulary.

[0072] It is understandable that the initial slot vocabulary will be used as the target slot vocabulary for subsequent keyword matching and recognition only when the target user has the corresponding permissions. This mechanism ensures that the use of the slot vocabulary meets security and privacy requirements, while avoiding users accessing irrelevant or restricted vocabulary content.

[0073] In the specific implementation, we can first analyze the deep target intent in the user request information based on a multimodal intent recognition model (such as BERT-wwm+LSTM+knowledge graph embedding) that integrates text semantics, conversation context and business scenario features, thereby providing a semantic background for subsequent slot keyword recognition. This can better understand the user's real needs and improve recognition accuracy; then dynamically determine the initial slot vocabulary through the three-dimensional vocabulary mapping matrix (intent category × basic vocabulary × extension module); then use the enhanced ABAC permission model (combined with multi-dimensional attributes such as department affiliation, security level and time period control) to perform millisecond-level permission verification.

[0074] Step S50: determining a target candidate word in the candidate word set, and using the target candidate word as a slot keyword.

[0075] In the specific implementation, the top-K (usually K=20) relevant candidate words can be preliminarily screened out from the candidate word set through semantic vector retrieval as the candidate word set; then the bge_reranker model is used to perform refined semantic reranking of the candidate words, and the semantic similarity score of each candidate word with the target keyword is calculated; in the reranking stage, the model will integrate context-aware features (such as conversation history, business scenarios, etc.) and domain knowledge constraints (such as human resources policy terms, job competency models, etc.) to generate a standardized matching score between 0 and 1; finally, the candidate word with the highest semantic similarity score is selected as the target candidate word, and the following optimization conditions are met at the same time: (1) The semantic similarity score of the target candidate word must exceed the dynamically adjusted domain preset threshold (such as the human resources management scenario threshold is set to 0.85); (2) The business compliance of the target candidate word is ensured to meet the compliance requirements through real-time knowledge graph verification; (3) When there are multiple parallel candidates with the highest score, the candidate word with a higher heat weight (i.e., matching frequency) in the target slot vocabulary is given priority.

[0076] This embodiment obtains the user request information of the target user; based on the large model prompt word, extracts the keyword from the user request information through the large model; determines whether the keyword exists in the target slot vocabulary; when it is determined that the keyword does not exist in the target slot vocabulary, filters out a set of candidate words related to the keyword from the target slot vocabulary; determines the target candidate word in the candidate word set, and uses the target candidate word as the slot keyword. This solves the technical problem of high cost and time-consuming to identify slot information in human-computer dialogue based on deep learning. Compared with the existing technology, this application, based on the large model, introduces the large model prompt word technology to more efficiently extract keywords from user requests, reduces the dependence on large-scale annotated data, reduces the training cost, and further optimizes the slot keyword recognition process by dynamically screening the candidate word set and determining the target candidate word, improves the recognition efficiency and accuracy, and can quickly respond to user requests in a short time.

[0077] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction and will not be described in detail later. Figure 2 , step S50 further includes steps S501 to S505:

[0078] Step S501, determining the semantic similarity and semantic relevance between the keyword and each candidate word in the candidate word set;

[0079] It should be noted that the deep semantic similarity between candidate words and target keywords can be calculated based on a domain-optimized semantic matching model (such as HR-BERT), and then the semantic correlation between terms can be captured through a multi-level attention mechanism (for example, the semantic correlation between "talent development" and "echelon building" reaches 0.91).

[0080] Step S502, determining a semantic similarity score according to the semantic similarity and the semantic relevance;

[0081] Understandably, in human resource management, different business scenarios may have different requirements for the weights of semantic similarity and semantic relevance. Therefore, similarity and relevance weights can be set according to the actual scenario. For example, in a talent inventory scenario, the direct semantic similarity between the candidate word and the keyword (weight 0.7) may be more important, while the requirement for relevance (weight 0.3) is relatively low. For example, if the keyword is "senior management talent," the candidate words "senior manager" and "senior executive" may be highly semantically similar, but have lower relevance to "middle management talent." For another example, in a training needs analysis scenario, semantic relevance (weight 0.4) may be more important because the training content may need to be linked to the employee's long-term career development path. For example, if the keyword is "leadership improvement," the candidate words "leadership training" and "management capacity improvement" may score higher in semantic relevance.

[0082] Step S503, determining whether there is a candidate word in the candidate word set whose semantic similarity score is greater than a preset threshold;

[0083] It should be noted that the preset threshold is a criterion for determining whether a candidate word is sufficiently close to a keyword, and this threshold can be adjusted dynamically based on the specific business scenario. For example, in human resources management, if the business scenario is "talent inventory," a higher semantic similarity threshold (such as 0.85) may be required to more accurately match keywords; while in the "training needs analysis" scenario, the threshold may be slightly lower (such as 0.7) to allow for broader semantic associations.

[0084] Step S504: If so, the candidate word in the candidate word set with a semantic similarity score greater than the preset threshold is used as the target candidate word;

[0085] In a specific implementation, if only one candidate word in the candidate word set has a semantic similarity score greater than a preset threshold, the candidate word is directly selected as the target candidate word. If multiple candidate words in the candidate word set have semantic similarity scores greater than the preset threshold, the final target candidate word is determined based on the candidate word's popularity weight, contextual information, domain knowledge, and semantic relevance. The popularity weight refers to the frequency of use or matching of the candidate word in the target slot vocabulary. Frequently used terms are generally more consistent with user needs and are therefore preferred. Contextual information includes other words in the user's request information, conversation history, or business scenario descriptions. This contextual information can help further determine the semantic background and specific meaning of the candidate word. Domain knowledge constraints: In certain specific fields (such as human resources, healthcare, and finance), there may be domain-specific knowledge rules or constraints that can help further filter candidate words. Semantic relevance reflects the degree of semantic relevance between the candidate word and the keyword. Multiple highly similar candidate words can also be presented to the user, allowing them to select the one that best meets their needs. User feedback can also be used as a basis for further optimizing the matching strategy. Through the above-mentioned strategies, the system can accurately screen out the target candidate words that best meet user needs and business scenarios when there are multiple high-similarity candidate words. This comprehensive selection mechanism not only significantly improves the accuracy and reliability of slot keyword recognition, but also provides users with more accurate and efficient services.

[0086] Step S505: If the target slot word does not exist, update the target slot word library based on the keyword.

[0087] It should be noted that if there is no candidate word with a semantic similarity score greater than a preset threshold in the candidate word set, the target slot vocabulary needs to be updated to adapt to new business needs or user input requests.

[0088] In one embodiment, the step of updating the target slot vocabulary based on the keyword includes: marking the keyword as an unmatched keyword and determining the context information of the unmatched keyword; standardizing the unmatched keyword to generate a candidate standard entry; submitting the candidate standard entry and the context information to the review platform for manual verification to obtain a manual verification result; when the manual verification result is passed, adding the mapping relationship between the unmatched keyword and the candidate standard entry to the target slot vocabulary.

[0089] It should be noted that when a candidate word with a semantic similarity with the keyword greater than a preset threshold cannot be found from the target slot vocabulary, the keyword is marked as an unmatched keyword; when extracting context information of the unmatched keyword, the context information may include relevant sentences in the user request information, conversation history, business scenario description, etc., which can be used to assist in understanding the semantic background and specific meaning of the unmatched keyword. For example, in a human resources scenario, if the user request information is "looking for talents with digital skills" and the unmatched keyword is "digital talent", the context information can help the verification personnel understand that "digital talent" may refer to "talents with digital technology capabilities", thereby generating more accurate candidate standard entries, such as " Digital technology talents"; standardization of unmatched keywords can ensure that candidate standard entries meet the format and specification requirements of the target slot vocabulary (standardization processing may include but is not limited to spelling correction, synonym replacement, term normalization and other operations); submit the candidate standard entries and context information to the review platform for manual verification by professionals, and the manual verification results may include verification passed, verification failed or further information is required; when the manual verification result is verification passed, the mapping relationship between the unmatched keywords and the candidate standard entries is added to the target slot vocabulary, and the mapping relationship may include fields such as keywords, candidate standard entries, context information, verification time, etc., for subsequent query and management.

[0090] In one embodiment, after the step of adding the mapping relationship between the unmatched keyword and the candidate standard term to the target slot vocabulary, it also includes: real-time monitoring of the term usage frequency and matching accuracy of the newly added terms in the target slot vocabulary; dynamically adjusting the matching strategy of the newly added terms according to the term usage frequency and the matching accuracy, so as to perform slot identification on the new user request information according to the matching strategy.

[0091] It's important to note that term usage frequency refers to the number of times a newly added term is matched during slot identification. It reflects the term's usage in actual business scenarios. In human resources management, certain terms may be associated with popular positions, common skills, or high-frequency business needs. For example, terms such as "data analysis" and "project management" may appear frequently in recruitment, training, or performance evaluations. By monitoring term usage frequency in real time, we can understand which terms are most popular or commonly used in actual business operations. For example, if the term "data analysis" is used frequently, it indicates that this skill is in high demand in the current business.

[0092] It's important to note that matching accuracy refers to the percentage of newly added terms that are correctly matched during the slot identification process. It reflects the accuracy of the term in actual use. In human resources management, matching accuracy directly impacts the reliability of slot identification. For example, if the term "senior manager" is frequently incorrectly matched to the position "middle management," its matching accuracy will be low. By monitoring matching accuracy in real time, we can identify terms that may be ambiguous or inaccurate in actual use. For example, if the matching accuracy for the term "digital technology talent" is low, it may be because the definition of the term is unclear or there is semantic overlap with other terms.

[0093] In specific implementations, the matching strategy of newly added terms can be dynamically adjusted based on the frequency of use and matching accuracy of the terms. The matching strategy includes term priority, matching threshold, context constraints, etc. If a term is used frequently but has a low matching accuracy, it may be necessary to redefine the semantic scope of the term or adjust its matching threshold. For example, if the term "digital technology talent" is often mismatched, its matching threshold can be increased, or context constraints such as "digital technology talent (technical position)" can be added during matching. If a term is used less frequently but has a high matching accuracy, its matching priority can be appropriately lowered to avoid overmatching. For example, the term "senior data scientist" may be used less frequently but has a high matching accuracy, and its priority can be adjusted to moderate.

[0094] In this embodiment, by monitoring the usage frequency and matching accuracy of terms in real time and dynamically adjusting the matching strategy based on this data, the accuracy and efficiency of slot identification can be effectively improved. This has important practical implications for various scenarios in human resource management (such as recruitment, training, and performance evaluation), helping organizations better respond to business changes and improving the intelligence of human resource management.

[0095] This embodiment determines the semantic similarity and semantic relevance between the keyword and each candidate word in the candidate word set; determines a semantic similarity score based on the semantic similarity and the semantic relevance; determines whether there is a candidate word in the candidate word set with a semantic similarity score greater than a preset threshold; if so, uses the candidate word in the candidate word set with a semantic similarity score greater than the preset threshold as the target candidate word; if not, updates the target slot vocabulary based on the keyword. This embodiment accurately screens out the candidate words that are closest in semantics to the keyword by setting a similarity threshold, thereby effectively improving the accuracy and reliability of slot keyword recognition, and can also timely update the target slot vocabulary based on the unmatched candidate word when there is no candidate word that meets the conditions in the candidate word set, ensuring the dynamic adaptability and extensibility of the vocabulary, enabling it to better respond to new keywords and user input, and further improving the robustness and flexibility of the system.

[0096] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the slot keyword identification method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0097] This application also provides a slot keyword recognition device, please refer to Figure 3 , the slot keyword recognition device includes:

[0098] An acquisition module 10 is used to acquire user request information of a target user;

[0099] An extraction module 20 is configured to extract keywords from the user request information based on the large model prompt words through the large model;

[0100] A judgment module 30 is used to judge whether the keyword exists in the target slot vocabulary;

[0101] A screening module 40 is configured to screen a set of candidate words related to the keyword from the target slot vocabulary when it is determined that the keyword does not exist in the target slot vocabulary;

[0102] The determination module 50 is configured to determine a target candidate word in the candidate word set and use the target candidate word as a slot keyword.

[0103] The slot keyword recognition device provided in this application utilizes the slot keyword recognition method described in the aforementioned embodiment, addressing the high cost and time-consuming technical issues associated with deep learning-based slot information recognition in human-computer interaction. Compared to the prior art, the slot keyword recognition device provided in this application offers the same beneficial effects as the slot keyword recognition method described in the aforementioned embodiment. Other technical features of the slot keyword recognition device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0104] The present application provides a slot keyword recognition device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the slot keyword recognition method in the above-mentioned embodiment 1.

[0105] Reference below Figure 4 , which shows a schematic structural diagram of a slot keyword recognition device suitable for implementing the embodiments of the present application. The slot keyword recognition device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The slot keyword recognition device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0106] like Figure 4As shown, the slot keyword recognition device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory 1002 or programs loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the slot keyword recognition device. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and a communication device 1009. Communication device 1009 can allow slot keyword recognition device to carry out wireless or wired communication with other equipment to exchange data.Although the slot keyword recognition device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown.Can implement or have more or less systems instead.

[0107] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are performed.

[0108] The slot keyword recognition device provided in this application, which utilizes the slot keyword recognition method described in the aforementioned embodiment, can address the high cost and time-consuming technical issues associated with deep learning-based slot information recognition in human-computer dialogues. Compared to the prior art, the slot keyword recognition device provided in this application has the same beneficial effects as the slot keyword recognition method described in the aforementioned embodiment. Other technical features of the slot keyword recognition device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0109] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0110] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0111] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer program) stored thereon, and the computer-readable program instructions are used to execute the slot keyword identification method in the above embodiment.

[0112] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0113] The computer-readable storage medium may be included in the slot keyword recognition device; or may exist independently without being assembled into the slot keyword recognition device.

[0114] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the slot keyword recognition device, the slot keyword recognition device: obtains user request information of the target user; extracts keywords from the user request information through the big model based on the big model prompt word; determines whether the keyword exists in the target slot vocabulary; when it is determined that the keyword does not exist in the target slot vocabulary, filters out a set of candidate words related to the keyword from the target slot vocabulary; determines the target candidate word in the candidate word set, and uses the target candidate word as the slot keyword.

[0115] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0116] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0117] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0118] The computer-readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned slot keyword recognition method. This computer-readable storage medium can address other technical issues related to the high cost and time-consuming nature of deep learning-based slot information recognition in human-computer dialogues. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the slot keyword recognition method provided in the aforementioned embodiment, and are not further elaborated here.

[0119] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned slot keyword identification method when executed by a processor.

[0120] The computer program product provided in this application can solve the technical problem of slot keyword recognition. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the slot keyword recognition method provided in the above embodiment, and will not be repeated here.

[0121] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A slot keyword recognition method, characterized in that: The slot keyword identification method includes: Obtain user request information of the target user; Based on the large model prompt words, extract keywords from the user request information through the large model; Determine whether the keyword exists in the target slot vocabulary; When it is determined that the keyword does not exist in the target slot vocabulary, a candidate word set related to the keyword is screened out from the target slot vocabulary; Determine a target candidate word in the candidate word set, and use the target candidate word as a slot keyword.

2. The method according to claim 1, wherein The step of extracting keywords from the user request information using the large model based on the large model prompt words includes: Build large model prompt words based on slot extraction tasks, output requirements, example demonstrations, and context specifications; Based on the large model prompt words, keywords are extracted from the user request information through the large model.

3. The slot keyword recognition method according to claim 1, wherein: Before the step of determining whether the keyword exists in the target slot vocabulary, the method further includes: Determining the target intent corresponding to the user request information; Based on the target intent, determining an initial slot vocabulary corresponding to the keyword; Determining the slot vocabulary permissions of the target user according to the identity information of the target user; Determining whether the target user has permission for the initial slot vocabulary based on the slot vocabulary permission; When it is determined that the target user has the authority of the initial slot vocabulary, the initial slot vocabulary is used as the target slot vocabulary.

4. The slot keyword recognition method according to claim 1, wherein: The step of determining the target candidate word in the candidate word set includes: Determining semantic similarity and semantic relevance between the keyword and each candidate word in the candidate word set; Determining a semantic similarity score based on the semantic similarity and the semantic relevance; Determining whether there is a candidate word in the candidate word set whose semantic similarity score is greater than a preset threshold; If so, the candidate word in the candidate word set whose semantic similarity score is greater than the preset threshold is taken as the target candidate word; If not, update the target slot vocabulary based on the keyword.

5. The slot keyword recognition method according to claim 4, wherein: The step of updating the target slot vocabulary based on the keyword includes: Marking the keyword as an unmatched keyword, and determining context information of the unmatched keyword; Standardizing the unmatched keywords to generate candidate standard entries; Submitting the candidate standard terms and the context information to the review platform for manual verification to obtain manual verification results; When the manual verification result is verification passed, the mapping relationship between the unmatched keyword and the candidate standard entry is added to the target slot vocabulary.

6. The slot keyword recognition method according to claim 5, wherein: After the step of adding the mapping relationship between the unmatched keyword and the candidate standard term to the target slot vocabulary, the method further includes: Real-time monitoring of the usage frequency and matching accuracy of the newly added entries in the target slot vocabulary; The matching strategy of the newly added term is dynamically adjusted according to the usage frequency of the term and the matching accuracy, so as to perform slot identification on the new user request information according to the matching strategy.

7. A slot keyword recognition device, characterized in that: The slot keyword recognition device includes: The acquisition module is used to obtain the user request information of the target user; An extraction module, configured to extract keywords from the user request information using a large model; A judgment module, used to judge whether the keyword exists in the target slot vocabulary; a screening module, configured to screen out a set of candidate words related to the keyword from the target slot vocabulary when it is determined that the keyword does not exist in the target slot vocabulary; The determination module is used to determine a target candidate word in the candidate word set and use the target candidate word as a slot keyword.

8. A slot keyword recognition device, characterized in that: The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the slot keyword identification method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the slot keyword identification method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the slot keyword identification method according to any one of claims 1 to 6 are implemented.

Images