Marketing verbal skill identification method and identification system, and electronic equipment

Abstract

The embodiment of the invention provides a marketing verbal skill recognition method and recognition system and electronic equipment, and belongs to the field of data analysis. The recognition method comprises the following steps: constructing a keyword dictionary and standard dictionary set of a specific marketing field; correcting error expressions in the original dialogue text by using the keyword dictionary of the specific marketing field to obtain a standard dialogue text; performing word segmentation processing on the standard dialogue text by utilizing the standard dictionary set and the dictionary tree to obtain a dictionary corresponding to the standard dialogue text; according to the various verbal skill category templates in the specific marketing field, generating a dictionary corresponding to the various verbal skill keywords; and matching a dictionary corresponding to a standard dialogue text with dictionaries corresponding to the various verbal skill keywords, and identifying whether a target verbal skill keyword exists in the standard dialogue text. According to the method, the accuracy of language recognition can be improved, and errors caused by recognition errors are reduced.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a marketing speech recognition method, a marketing speech recognition system, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the widespread adoption of digital and intelligent technologies across various industries, sales management is also seeking to leverage advanced technologies to improve sales efficiency and service quality. In the gas station retail industry, in particular, facing increasingly fierce market competition, the question of how to scientifically and accurately identify the quality of employees' marketing pitches during work, thereby optimizing sales strategies and enhancing their sales capabilities, has become a key industry concern. Summary of the Invention

[0003] The purpose of the embodiments of the present invention is to provide a sales talk recognition method, a marketing talk recognition system, an electronic device, a computer-readable storage medium and a computer program product. The recognition method can partially or completely solve the problem of low accuracy of existing talk recognition.

[0004] In order to achieve the above-mentioned purpose, an embodiment of the present invention provides a method for identifying marketing speech, which includes: constructing a keyword dictionary and a standard dictionary set for a specific marketing field; using the keyword dictionary for the specific marketing field to correct incorrect expressions in the original dialogue text to obtain a standard dialogue text; using the standard dictionary set and the dictionary tree to perform word segmentation processing on the standard dialogue text to obtain a dictionary corresponding to the standard dialogue text; generating a dictionary corresponding to the multiple speech keywords based on multiple speech category templates in the specific marketing field; and matching the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords to identify whether the standard dialogue text contains target speech keywords.

[0005] Optionally, constructing a keyword dictionary for a specific marketing field includes: constructing a keyword dictionary for the specific marketing field according to keywords for the specific marketing field and incorrect expressions corresponding to the keywords.

[0006] Optionally, the use of the keyword dictionary of the specific marketing field to correct erroneous expressions in the original dialogue text to obtain a standard dialogue text includes: matching the original dialogue text with the keyword dictionary of the specific marketing field, and when an erroneous expression corresponding to the original dialogue text is matched in the keyword dictionary, correcting the original dialogue text to obtain a standard dialogue text.

[0007] Optionally, the multiple speech keywords include greeting keywords, promotional speech keywords, blacklist speech keywords and custom keywords; matching the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords, and identifying whether there are target speech keywords in the standard dialogue text includes: when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any of the multiple speech keywords are not matched, determining that there are no target speech keywords in the standard dialogue text; or, when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple speech keywords is matched, comparing the matched speech keywords with the target speech keywords; and, if the matched keywords are consistent with the target speech keywords, determining that the standard dialogue text contains the target speech keywords.

[0008] Optionally, before executing the step of constructing a keyword dictionary and a standard dictionary set for a specific marketing field, the recognition method also includes: obtaining the original conversation voice between customer service and customers in the target store within a set time period; and performing voice recognition on the original conversation voice to obtain the original conversation text.

[0009] On the other hand, the present invention also provides a marketing speech recognition system, which includes: a construction module for constructing a keyword dictionary and a standard dictionary set for a specific marketing field; a first acquisition module for correcting incorrect expressions in the original dialogue text using the keyword dictionary for the specific marketing field to obtain a standard dialogue text; a second acquisition module for performing word segmentation processing on the standard dialogue text using the standard dictionary set and the dictionary tree to obtain a dictionary corresponding to the standard dialogue text; a generation module for generating a dictionary corresponding to the multiple speech keywords based on the multiple speech category templates for the specific marketing field; and an identification module for matching the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords to identify whether the standard dialogue text contains target speech keywords.

[0010] Optionally, the construction module for constructing a keyword dictionary in a specific marketing field includes: constructing a keyword dictionary in the specific marketing field according to keywords in the specific marketing field and incorrect expressions corresponding to the keywords.

[0011] Optionally, the first acquisition module is used to correct incorrect expressions in the original dialogue text using the keyword dictionary of the specific marketing field to obtain a standard dialogue text, including: matching the original dialogue text with the keyword dictionary of the specific marketing field, and when an incorrect expression corresponding to the original dialogue text is matched in the keyword dictionary, correcting the original dialogue text to obtain a standard dialogue text.

[0012] Optionally, the multiple speech keywords include greeting keywords, promotional speech keywords, blacklist speech keywords and custom keywords; the identification module is used to match the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords, and identify whether there are target speech keywords in the standard dialogue text, including: when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any of the multiple corresponding speech keywords are not matched, it is determined that there are no target speech keywords in the standard dialogue text; or, when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple corresponding speech keywords is matched, the matched speech keywords are compared with the target speech keywords; and, if the matched keywords are consistent with the target speech keywords, it is determined that the standard dialogue text contains the target speech keywords.

[0013] Optionally, the recognition system also includes: a third acquisition module, used to obtain the original conversation voice between customer service and customers in the target store within a set time period before executing the step of constructing the keyword dictionary and standard dictionary set for the specific marketing field; and a fourth acquisition module, used to perform voice recognition on the original conversation voice to obtain the original conversation text.

[0014] On the other hand, the present invention also includes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the marketing speech recognition method as described.

[0015] On the other hand, the present invention also includes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the marketing speech recognition method as described above.

[0016] In another aspect, the present invention also includes a computer program product, including a computer program, which implements the marketing speech recognition method as described above when executed by a processor.

[0017] Through the above technical solution, a keyword dictionary for a specific marketing field is constructed, and the keyword dictionary for this specific field is used to identify the original dialogue text. If the original dialogue text contains incorrect expressions, the corresponding incorrect expressions are corrected to obtain a standard dialogue text; then the standard dialogue text is segmented using the created standard dictionary set to obtain a dictionary corresponding to the standard dialogue text; finally, the dictionary corresponding to the standard dialogue text is matched with the dictionaries corresponding to multiple speech keywords to identify whether the standard dialogue text contains the target speech keywords.

[0018] In the technical solution of the present invention, texts that may contain erroneous expressions in the speech text are corrected, the accuracy of language recognition is improved, and errors caused by recognition errors are reduced.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings:

[0021] Figure 1 This is a flowchart of a method for identifying marketing tactics provided by a first embodiment of the present invention;

[0022] Figure 2 This is a flowchart of performing word segmentation processing on a text provided by the first embodiment of the present invention;

[0023] Figure 3 This is a structural diagram of a marketing speech recognition system provided by a second embodiment of the present invention;

[0024] Figure 4 This is a diagram of a speech recognition system provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0026] Figure 1 The first embodiment of the present invention provides a flowchart of a method for identifying marketing language. The method includes the following steps S10-S14.

[0027] In step S10, a keyword dictionary and a standard dictionary set for a specific marketing field are constructed.

[0028] Before executing the step of constructing a keyword dictionary and a standard dictionary set for a specific marketing field, the recognition method also includes: obtaining the original conversation voice between customer service and customers in the target store within a set time period; and performing voice recognition on the original conversation voice to obtain the original conversation text.

[0029] In some embodiments, the specific marketing area may specifically be the convenience store retail sector, or more specifically, the gas station convenience store sector and similar business models. The present invention uses the gas station convenience store as an example. First, voice acquisition equipment is used to effectively capture the interactive conversations between employees and customers at the gas station store in real time. A series of technical processing is then performed to ensure that the subsequent speech recognition model can accurately convert speech into text information.

[0030] The hardware configuration of the voice acquisition equipment is as follows: The voice acquisition equipment deployed at gas stations serves as the front-end sensing element of the entire system, designed to balance portability, durability, and high-performance voice acquisition capabilities. The equipment typically utilizes an embedded design, including components such as a directional microphone array, a signal processor, a storage unit, a wireless communication module, and a power supply. The directional microphone array utilizes multiple high-sensitivity microphones, focusing on the target voice signal through spatial filtering technology, effectively suppressing noise interference from non-target directions. This makes it particularly suitable for use in complex noise environments such as gas stations. This can be achieved using existing technologies, so we will not elaborate on this further.

[0031] To improve voice capture clarity, the voice capture equipment should be installed and positioned according to the following requirements: Select appropriate locations based on the gas station's layout and workflow, such as key interaction points like convenience store checkout counters and customer service areas. Ensure coverage of key conversation areas while avoiding direct exposure to strong noise sources (such as gas pumps and passing vehicles). The equipment's height and angle should be adjustable to optimize the microphone's ability to capture the target voice.

[0032] The voice capture device has a built-in, stable and reliable wireless communication module (such as Wi-Fi and cellular networks) to enable remote transmission of real-time audio data. Encryption technology is used during data transmission to ensure information security, and Quality of Service (QoS) control mechanisms ensure the continuity and low latency of the audio stream to meet the needs of real-time voice recognition.

[0033] Since the collected speech contains noise, the embodiment of the present invention can also preprocess the speech. The speech preprocessing specifically includes: denoising, segmentation, standardization, volume normalization, sampling rate conversion, channel conversion, audio encoding and compression, data encapsulation and transmission.

[0034] Denoising: Because gas station ambient noise is complex and highly variable in intensity, the primary task in the preprocessing phase is to denoise the collected raw audio. This method utilizes advanced noise suppression algorithms, such as adaptive filters, spectral subtraction, and deep learning noise suppression models. Adaptive filters utilize noise estimation and filter update mechanisms to dynamically eliminate background noise; spectral subtraction reduces the impact of noise components in the frequency domain by comparing the spectral characteristics of noise and speech; and deep learning noise suppression models utilize extensive training data to learn the differences between noise and speech characteristics, achieving more refined noise separation.

[0035] Segmentation is performed as follows: To meet the processing requirements of the speech recognition model, long-time audio data is appropriately segmented into shorter speech segments. Segmentation strategies can be based on energy threshold detection, endpoint detection algorithms (such as VAD, Voice Activity Detection), or specific trigger words (such as opening greetings and closing remarks). The segmentation process should minimize the integrity of the conversation, avoid interrupting key information, and consider the input length limitations of the speech recognition model.

[0036] Standardization: Standardize the segmented speech segments to ensure that all data input into the recognition model has a uniform quality standard. For specific standardization steps, refer to existing technical means.

[0037] Volume normalization: Adjust the peak amplitude of each segment so that the maximum sound pressure level of all segments is roughly the same, preventing fluctuations in recognition results due to volume differences.

[0038] Sampling rate conversion: If the sampling rate of the acquisition device does not match the sampling rate expected by the recognition model, perform sampling rate conversion to ensure that the two are consistent and avoid frequency distortion.

[0039] Channel conversion: If the capture device obtains dual-channel audio, convert it to mono to simplify subsequent processing.

[0040] Audio Coding and Compression: To reduce data transmission and storage requirements, pre-processed voice clips are efficiently encoded and compressed. Widely recognized audio codec standards such as MPEG-1 Layer III (MP3) and Advanced Audio Coding (AAC) are used to balance compression efficiency and preserve sound quality. Compression parameters can be adjusted based on actual network conditions and storage resources, ensuring real-time transmission of audio data within limited bandwidth while reducing storage costs without compromising recognition performance.

[0041] Data encapsulation and transmission: Encapsulate the encoded and compressed voice data according to a predefined communication protocol (such as RTSP or WebSocket), attach necessary metadata (such as timestamp, device ID, and segment sequence number), and send it to the backend server in real time via a wireless network. After receiving the data packet, the server unpacks and decodes it, restoring it to an audio format suitable for processing by the speech recognition model.

[0042] After the original speech is pre-processed, it is recognized using an existing speech recognition algorithm to obtain the original dialogue text. The original speech in the embodiment of the present invention refers to the speech that has been pre-processed.

[0043] In summary, the speech acquisition and preprocessing module of the present invention integrates targeted hardware, efficient noise suppression algorithms, flexible speech segmentation strategies, and coding and compression technologies adapted to network conditions, forming a complete front-end speech data acquisition and preprocessing system. This system can effectively capture the interactive speech between employees and customers in the complex environment of gas stations. After a series of technical processing, it provides high-quality, adaptive audio input for the subsequent speech recognition and error correction module, laying the foundation for the accurate operation of the entire intelligent assessment system.

[0044] For example, to obtain the original conversation text for a specific store within a certain time period, you can define it as follows: store_id: A string that identifies the store's unique ID, such as "STORE001" or "STORE002." effective_time_range: An object containing (specific actual business data, for subsequent optimization). start_date: A string in the "YYYYMMDD" format, indicating the template's effective start date, such as "20240401."

[0045] end_date: string, the format is the same as above, indicating the template's effective end date, such as "20241231".

[0046] Furthermore, constructing a keyword dictionary in a specific marketing field includes: constructing a keyword dictionary in the specific marketing field according to keywords in the specific marketing field and incorrect expressions corresponding to the keywords.

[0047] In some embodiments, after obtaining the original conversation text, considering that there may be dialect or expression deviations in the communication between consumers and customer service, or homophonic objections, expression habits in specific fields and other problems, which may lead to inaccurate expression of the original conversation speech converted into the original conversation text, the embodiments of the present invention improve the accuracy of the original text by correcting errors in the original conversation text.

[0048] The key to building a keyword dictionary for a specific marketing field lies in establishing a highly customized dictionary template. First, the system needs to deeply explore a specific field (specifically, gas station and convenience store sales scenarios in this article) and extensively collect all keywords within that field. These keywords are not only the cornerstone of knowledge in that field, but also the benchmark for subsequent error correction work. For example, in gas station store conversation analysis, keywords can include various product names and professional terms. The collection process can be carried out through automated crawler technology to capture from authoritative websites and databases, or combined with expert manual review to ensure the comprehensiveness and accuracy of keywords.

[0049] These keywords also include a collection of word pairs that represent incorrect expressions of the keywords. This process can also be achieved through crawler technology. Detailed manual completion work can also be carried out around each keyword, aiming to build a set of word pairs containing the keyword and all possible incorrect expressions, namely, a keyword dictionary for a specific marketing field. For example, the correct keyword is "very cola" and the corresponding incorrect expressions include "very pitiful," "very happy," "very thirsty," etc. These expressions are also added to the keyword dictionary. This keyword dictionary consists of keys and key values, where each key is a keyword-related word and the corresponding key value is the pinyin of the keyword. For example, {'very cola': ['fei', 'chang', 'ke', 'le']; 'very thirsty': ['fei', 'chang', 'ke', 'le']; 'very pitiful': ['fei', 'chang', 'ke', 'lian']}, etc.

[0050] This process is crucial and challenging, as it requires designers to consider not only common errors like misspellings and homonyms, but also to understand the unique idioms and confusing concepts within a specific field. Through this combined approach of manual effort and algorithms, the generated dictionary template acts as a precise navigation map, providing a solid foundation for subsequent error correction.

[0051] Moreover, the embodiment of the present invention also needs to construct a standard dictionary set at the same time. The so-called standard dictionary set only contains accurate keywords and does not include incorrect expressions. For example, take commodities as an example: {'Very Cola': ['fei', 'chang', 'ke', 'le']}; 'Konjac Tripe'} ['mo', 'yu', 'su', 'mao', 'du']}, the standard dictionary set includes all commodities in the convenience store, and the standard dictionary set also includes commonly used terms in the field, specific industry terms, proper nouns and their frequency information. Specifically, a set of Chinese dictionary sets with more than 40,000 words containing specific industry terms, proper nouns and their frequency information is constructed, and the standard dictionary set is continuously updated, and proper nouns or phrases in specific industries or scenarios are supplemented in a timely manner. In this way, it is ensured that these words will not be incorrectly segmented during word segmentation.

[0052] In step S11, incorrect expressions in the original dialogue text are corrected using the keyword dictionary of the specific marketing field to obtain a standard dialogue text.

[0053] Furthermore, the use of the keyword dictionary of the specific marketing field to correct erroneous expressions in the original dialogue text to obtain a standard dialogue text includes: matching the original dialogue text with the keyword dictionary of the specific marketing field, and when an erroneous expression corresponding to the original dialogue text is matched in the keyword dictionary, correcting the original dialogue text to obtain a standard dialogue text.

[0054] In some embodiments, after having a rich and accurate keyword dictionary for a specific marketing field, the error correction system can enter the actual combat stage. When a text that needs to be verified (such as the original conversation text in the present invention) is input into the system, the algorithm will compare word by word and try to find a match in the keyword dictionary. If it is found that a certain word in the original conversation text matches the incorrect expression of a keyword in the dictionary, the system will immediately activate the error correction mechanism and replace the incorrect expression with the corresponding correct keyword. For example, when processing a conversation at a gas station store, if the original conversation text mistakenly writes "very cola" as "very pitiful", the system can quickly identify this error and automatically correct it to obtain a standard conversation text, that is, replace the incorrect expression in the original conversation text with the correct expression, thereby ensuring the professionalism and accuracy of the text to be identified. And in order to avoid the risk of erroneous corrections, the dictionary construction is required to be as comprehensive as possible to cover all possible error situations. Therefore, the keyword dictionary for a specific marketing field is continuously updated and improved, and a user feedback mechanism is introduced.

[0055] By constructing a keyword dictionary for specific marketing fields, we can effectively identify and accurately correct errors in texts in specific fields, greatly improving the quality and efficiency of text processing.

[0056] Retrieve the conversation log between two customers. dia_id: A string, unique identifier for the conversation log, such as "A_1". customer_speech: A string, the customer's original speech. agent_response: A string, the agent's original response. Note: All conversations are merged into a single segment. Assume the standard conversation log information is as shown in Table 1.

[0057]

[0058]

[0059] Table 1

[0060] In step S12, the standard dialogue text is segmented using the standard dictionary set and the dictionary tree to obtain a dictionary corresponding to the standard dialogue text.

[0061] In some embodiments, after constructing a standard dictionary set for a specific marketing field in step S10, the original conversation text is radially segmented using a Trie tree (dictionary tree) using the standard dictionary set. The Trie tree allows the search process to quickly locate words in the dictionary, and word frequencies are counted and converted into frequency information during the construction process.

[0062] For each input sentence, a directed acyclic graph (DAG) is generated based on the dictionary tree. In this graph, each node represents a character in the sentence, and the edges represent the possible lexical relationships between characters. In this way, all possible word combinations in the sentence are represented.

[0063] Let's explain this with a specific example: we'll show how to use DAG to represent the word segmentation of the sentence "I love natural language processing." First, assume we have a simple standard dictionary set containing the following vocabulary (this is just an example; the actual word segmenter's dictionary is much larger and contains more details):

[0064] I 328841r;

[0065] Love 14878v;

[0066] Nature 20269d;

[0067] Language 7647n;

[0068] Processing 10840v;

[0069] Natural language 46l;

[0070] Language Processing 40nv.

[0071] Next, according to the dictionary, construct a DAG for each character in the sentence. Each node in the DAG represents a character in the original sentence, and each directed edge represents a possible word boundary. The specific construction method is as follows: Initialization: Start from the first character "我" of the sentence and create a starting node. Traverse the sentence: When encountering "我", since "我" is an independent word, draw an edge from the starting point to the end position of "我". Continue to "爱", and similarly, since "爱" is an independent word, draw an edge from the end position of "我" to the end position of "爱". Reach "自", where "自" can be an independent word or part of "自然", so draw two edges from the end position of "爱", one to the end of "自" and the other reserved for the possible end position of "自然". "自然" is in the dictionary, so draw an edge from the node of "自" to the end of "自然". Similarly, "语言" after "自然" is also a word in the dictionary, so draw an edge from the end position of "自然" to the end of "语言". At the same time, because "语言处理" is also a word, reserve another edge from the end position of "语言" for the possible end position of "语言处理". Finally, "处理" is a word in the dictionary, so draw an edge from the end position of "语言" to the end of "处理", completing the traversal of the entire sentence. As Figure 2 is the flowchart for word segmentation processing of text provided by the first embodiment of the present invention.

[0072] It can be understood that each character is a node, and the possible combinations of words can be seen from the edges between characters. For example, from "自" to "自然" to "语言" and finally to "语言处理" represents the recognition path of "自然语言处理" as a whole vocabulary.

[0073] Next, we need to calculate which of all the above paths has the greatest possibility, and take the segmentation corresponding to the path with the greatest possibility as our word segmentation result.

[0074] Assume: w ij represents the frequency of the word formed by the characters from the i-th character to the j-th character in the sentence in the word frequency table. If the character combination in this interval is not a legal word, the default frequency is 1. Total is the sum of the frequencies of all legal words. P(w ij ) represents the probability of the word w ij , calculated as w ij / Total; R k represents the maximum log probability sum from position k to the end of the sentence in the optimal path; log represents the natural logarithm.

[0075] For each position i in the sentence, for each possible word segmentation end position j from i to N - 1, calculate the following probability (in logarithmic form): T ij = log(P(wij )) - log(Total) + R j+1 。

[0076] Converted to non - logarithmic form, which is equivalent to: P(w ij ) * R j+1 / Total. However, in the actual dynamic programming implementation, we directly operate in the logarithmic space to avoid floating - point underflow and improve computational stability. So the original mathematical expression should be the cumulative sum in logarithmic form, that is: T ij = log(w jj / Total) + R j+1 . Finally, find the maximum - probability path, calculate the probabilities of all possible paths and select the maximum one max(T ij ). The end point and the corresponding logarithmic probability difference are the solution of the current optimal path. In summary, this method, through the dynamic programming algorithm, uses the constructed DAG graph to calculate the maximum - probability word - segmentation path from the end of the sentence to each character forward. This process is a key step in Chinese text pre - processing, which can cut continuous text into word units with independent meanings, facilitating further text analysis and processing. After obtaining the word - segmentation results, convert Chinese characters into pinyin sequences, and support the handling of polyphonic characters. Here, the fuzzy mode is adopted, that is, only return the pinyin string without additional formatting processing, which can reduce memory consumption and improve processing speed.

[0077] Finally, construct a word - segmentation result dictionary, where each key is the word after word - segmentation, and the corresponding value is the pinyin form of the word. For example: {'你好': ['ni', 'hao'], ',': [','], '这里': ['zhe', 'li'], '是': ['shi'], '西安': ['xi', 'an'],'吗': ['ma'], '?': ['?'], '.': ['.']}. This data structure can directly query the pinyin of the word according to the word efficiently and accurately for further text processing, recognition, analysis, error correction and other tasks. Using a similar method, a dictionary corresponding to the standard benchmark text can be obtained.

[0078] In step S13, according to the multiple types of conversation - style templates in the specific marketing field, generate a dictionary corresponding to the multiple conversation - style keywords.

[0079] In some embodiments, the core goal of Chinese text recognition is to generate the mapping relationship between keywords and their pinyin according to a given template. The design idea of this method fully reflects the encapsulation and data - processing capabilities in object - oriented programming, and is widely applicable to scenarios that require processing a large number of keyword - pinyin matches, such as search - engine optimization, speech - recognition systems, intelligent recommendation systems, etc. The following will introduce this method in depth from multiple dimensions.

[0080] First, we build templates for various speech categories in specific marketing fields. For the convenience store marketing field covered by this invention, these templates typically cover a wide range of scenario needs and can be roughly divided into four categories: Greeting speech templates: These templates are designed to establish friendly initial interactions, including but not limited to "Hello, how may I help you?" and "Welcome, what would you like to learn today?" They can quickly connect with users and create a positive atmosphere for communication.

[0081] Promotional script templates are designed for sales or promotional scenarios, such as "Buy now and enjoy 20% off! Miss out and you'll regret it for a year!" or "Our new product has launched, exclusive offer only for today! Hurry and order now!" These templates focus on conveying value and stimulating customer purchase desire. Keyword templates are represented in dictionary format as: key: a string representing the keywords / phrases of the promotional slogan, such as "Buy one, get one free, super value." Value: a string array, such as ["mai","yi","zeng","yi","chao","zhi","xiang","shou"].

[0082] Blacklist script templates: These are used to filter out inappropriate or potentially illegal content, such as insulting words and spam, to ensure a healthy and safe conversation environment. For example, "Please refrain from sending offensive messages, or your service will be terminated." By setting up these templates, the system can automatically identify and respond to negative interactions.

[0083] Custom templates: Adding a new category of custom templates to the existing script template system aims to meet specific business scenarios or personalized needs, providing more flexible and diverse interaction strategies. These templates allow businesses or individuals to design unique scripts based on their service characteristics, customer base, or temporary event needs, thereby improving service relevance and user satisfaction. Custom keywords are represented in a dictionary format as follows: key: a string representing the user-specified template keyword, such as "Coca-Cola" or "Wuyi Mountain." value: a string array, such as ["ke","kou","ke","le"].

[0084] After obtaining various types of speech, the next step is to convert them into a standard dictionary form for efficient subsequent processing. Each speech template can be regarded as a key-value pair, where the key usually represents the category or purpose of the speech, and the value is the specific expression content. For example: {'15243536254253': {'Coca-Cola': ['ke', 'kou', 'ke', 'le'], 'Pizza Hut Pizza': ['bi', 'sheng', 'ke', 'pi', 'sa'], 'Konjac Tripe': ['mo', 'yu', 'su', 'mao', 'du'], 'Maltes': ['mai', 'li', 'su'], 'Wuyi Mountain': ['w u', 'yi', 'shan'], 'Xi'an': ['xi', 'an']}, '15243536254236': {'Welcome': ['huan', 'ying', 'guang', 'lin'], 'Thank you for coming': ['you', 'lao', 'da', 'jia', 'guang', 'lin'], 'It's my honor to have you visit my humble home': ['guang', 'li n', 'han', 'she', 'bu', 'sheng', 'rong', 'xing'], 'It makes my humble home glorious': ['ling', 'peng', 'bi', 'sheng', hui'], 'I've long admired you for a long time': ['jiu', 'yang', 'jiu', 'yang'], 'I feel deeply honored': ['shen', 'gan', 'rong', 'xing'], 'Welcome to our guidance': ['huan', 'ying', 'li', 'lin', 'zhi', 'dao'], 'You are cordially invited to visit us': ['jing', 'qing', 'guang', 'lin'], 'You are cordially invited to give us your guidance': ['jing', 'qing', 'zhi', 'dao'], 'We have been waiting for your guidance': ['yi', 'gong', 'hou', 'duo', 'shi']},}.

[0085] The template IDs 15243536254253 and 15243536254236 are any of the four template types listed above. The following values ​​are the keywords or sentences to be identified within the template. This data structure makes template management more organized and simplifies template storage, retrieval, and identification.

[0086] In step S14, the dictionary corresponding to the standard dialogue text is matched with the dictionaries corresponding to the plurality of speech keywords to identify whether the target speech keyword exists in the standard dialogue text.

[0087] Furthermore, the multiple speech keywords include greeting keywords, promotion speech keywords, blacklist speech keywords and custom keywords.

[0088] The matching of the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords to identify whether the standard dialogue text contains the target speech keywords includes: when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any one of the multiple speech keywords is not matched, determining that the standard dialogue text does not contain the target speech keywords; or, when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple speech keywords is matched, comparing the matched speech keywords with the target speech keywords; and, if the matched keywords are consistent with the target speech keywords, determining that the standard dialogue text contains the target speech keywords.

[0089] In some embodiments, the various script templates described in step S13 above may include greeting keywords such as "Hello," "Welcome," and "Hello, what can I do for you?" depending on the template type. Promotional script keywords include "points redemption," "full discount," and "20% discount." Blacklisted scripts include "uncivilized language." Customized keywords include "mineral water," "windshield washer fluid," "self-service car wash," and "green energy" related keywords.

[0090] The standard conversation text (i.e., the corrected conversation text) is matched against a dictionary of various speech keywords. This step is used to identify whether the standard conversation text contains the speech terms in the dictionary corresponding to the various speech keywords. These keywords are organized by category, and each word is pre-converted into the dictionary format described above (i.e., the primary key is the word, and the value is a unique sequence of pinyin segments) for comparison. This process mainly serves functions such as quality monitoring of the dialogue system or automated responses triggered by specific keywords. The working principles and steps of speech recognition are detailed below.

[0091] When the key words in the standard dialogue text do not match any of the multiple speech keywords, that is, there is neither greeting nor product recommendation or sales activity speech, it means that the target speech keywords are not mentioned in the standard dialogue text. The target speech keywords can be regarded as the speech that the operation manager hopes the customer service staff will take the initiative to greet and launch corresponding products in a timely manner according to customer needs.

[0092] Conversely, imagine you're the operations manager of a gas station chain. You want to analyze conversations between store employees and customers to ensure they're appropriately mentioning the company's key services. For example, target keywords like "points redemption discounts," "self-service car wash," and "introduction to green energy." To monitor the effectiveness of these key messages, create a dictionary for each keyword. Because everyday conversations at gas stations may contain variations in dialect or accent, using pinyin allows for more accurate keyword capture.

[0093] First, obtain a dictionary corresponding to various scripts and organize them by service category. For example, a custom template might include promotions, value-added services, and environmental education. "Points redemption discount" falls under "Promotions," "Self-service car wash" falls under "Value-added Services," and "Introduction to green energy" falls under "Environmental Education." Each keyword has a corresponding specific pinyin sequence, such as "jī fēn duì hu àn yōu huì," "zì zhù xǐ chē fú wù," and "lǜ sè néng yuán jiè shào."

[0094] When a customer asks about services or discounts, the store staff will give corresponding answers. For example, the staff may say: "Hello, welcome! We currently have a points redemption promotion. With 300 points or more, you can enjoy a 5 yuan car wash discount. In addition, we also provide self-service car wash services, which are environmentally friendly and convenient. There is also a detailed introduction to the use of green energy vehicles, which you can learn about." The content of this conversation needs to be systematically analyzed. First, this conversation is split into individual words based on the word segmentation method described above, and then converted into a dictionary. Then, it will compare the pinyin of each word with the entries in the keyword pinyin mapping table. Using the method provided in the embodiment of the present invention, the specific output parameters for the above conversation text are as follows: Top-level key (STORE001, STORE002): represents different store identifiers.

[0095] Conversation record identifier (A_1, A_2): Specifies a specific customer service interaction record within a store.

[0096] keyword_templates(matched,keywords):

[0097] ■matched: Boolean value indicating whether the keyword template is matched in the customer's speech.

[0098] ■keywords: string array, listing the specific keywords that were matched (if any).

[0099] welcome(matched,keywords):

[0100] ■matched: Boolean value indicating whether the welcome message is recognized.

[0101] ■keywords: A string array listing the recognized welcome keywords (if any).

[0102] promotion(matched,keywords):

[0103] ■matched: A Boolean value indicating whether the promotion was mentioned.

[0104] ■keywords: string array, listing the matching promotion keywords (if any).

[0105] For example, by identifying the pinyin for "points redemption discount," "self-service car wash service," and "introduction to green energy" from employee responses and confirming whether these keywords are mentioned, the customer service attitude can be evaluated, facilitating effective customer service management across stores and improving customer satisfaction. A multi-dimensional quality inspection framework can also be established, encompassing politeness, professionalism, rationality, and demand satisfaction. Each metric can be quantified and scored, enabling a comprehensive assessment of customer service delivery. Compared to single-dimensional evaluations, this comprehensive approach improves the objectivity and accuracy of service evaluations and is expected to increase customer satisfaction by at least 15%, contributing to brand image building. See Table 2 for the output recognition results.

[0106]

[0107]

[0108]

[0109] Table 2

[0110] Finally, a report will be generated to tell the user that in this conversation, "points redemption discount" under the "promotional activities" category was mentioned, which is a match; "self-service car wash service" under the "value-added services" category was also mentioned, which is also a match; and "green energy introduction" under the "environmental education" category was also mentioned. This way, it is clear whether employees have successfully promoted the various services that the company wants to emphasize.

[0111] In this way, not only can service quality be monitored, but training focus can also be adjusted based on actual conversation content, ensuring that every customer interaction receives the expected information and service experience.

[0112] This invention significantly improves speech recognition accuracy, reducing the error rate by at least 30% compared to traditional methods. This means that under the same conditions, the system can more accurately capture key information in customer service conversations, reduce misjudgments, and improve the reliability and efficiency of quality inspections.

[0113] This invention supports cloud-based configuration and instant updating of script templates. Companies can quickly adjust their marketing strategies based on market feedback or policy changes with simple operations, without having to wait for software upgrades. This flexibility reduces response time to hours, shortening the update cycle by 90% compared to traditional fixed script libraries, significantly improving market adaptability and competitiveness.

[0114] This invention will focus on developing efficient and accurate speech recognition algorithms and introduce a speech recognition error correction module. Through deep learning and big data technology, it will optimize the accuracy of speech recognition, reduce quality inspection errors caused by recognition errors, and ensure the credibility and effectiveness of quality inspection results.

[0115] Realize flexible configuration and dynamic update of standard speech templates: The present invention aims to provide a method and system that supports flexible configuration and real-time update of standard speech templates, so that enterprises can adjust and expand speech standards conveniently and quickly according to business needs, market feedback and changes in customer service scenarios to meet quality inspection needs in different scenarios.

[0116] The present invention can also further construct a comprehensive quality inspection framework covering multiple dimensions such as politeness and compliance of the speech, professionalism and rationality of promotion, and satisfaction of customer needs when identifying that the marketing speech contains the target speech, so as to conduct a comprehensive, objective and detailed identification and evaluation of customer service speech, thereby more effectively guiding customer service personnel to improve service quality and promote the improvement of the overall service level of the enterprise.

[0117] The embodiments of this invention offer the following potential: 1. Intelligently drive sales growth. By accurately identifying the effectiveness of marketing pitches, the system provides timely feedback on which pitches are most likely to lead to sales and which links are lacking, guiding employees in optimizing their communication strategies. It is estimated that within one year of system deployment, the average sales generated by each employee will increase by 20%, directly driving overall revenue growth for gas stations. 2. The massive amount of data collected by the data-driven management decision-making system not only facilitates quality control but also provides valuable insights for management, such as which time periods and services are most attractive to customers, and which promotions receive the best customer feedback. Based on this data, management can formulate more precise sales strategies and resource allocation, achieving an estimated 40% improvement in decision-making efficiency and reducing resource waste. 3. Combined with the evaluation results of the intelligent recognition system, companies can design targeted training programs to enhance employee professional skills and sales techniques. Furthermore, the increased transparency of performance can establish fair incentive mechanisms, improve employee motivation and loyalty, and increase employee satisfaction by over 10%. This invention not only takes into account the problems of low efficiency, strong subjectivity, and insufficient data utilization of traditional quality inspection methods, but also realizes the intelligent and refined management of marketing language recognition through technological innovation, bringing significant economic and social benefits to the gas station industry and promoting the intelligent transformation of the service industry.

[0118] In summary, this invention aims to provide a system for intelligently identifying marketing language used by gas station employees based on speech recognition and natural language processing technologies. This system addresses the subjectivity, inefficiency, inadequate data collection and analysis, and poor technical adaptability of traditional evaluation methods. This system enables efficient collection, in-depth analysis, and accurate evaluation of marketing language, thereby improving sales management capabilities and employee sales and service capabilities, and promoting the intelligent, data-driven transformation of sales management in the gas station industry.

[0119] like Figure 3 It is a marketing speech recognition system provided by the second embodiment of the present invention. A marketing speech recognition system, the recognition system 30 includes: a construction module 301, used to construct a keyword dictionary and a standard dictionary set for a specific marketing field; a first acquisition module 302, used to use the keyword dictionary of the specific marketing field to correct incorrect expressions in the original dialogue text to obtain a standard dialogue text; a second acquisition module 303, used to use the standard dictionary set and the dictionary tree to perform word segmentation processing on the standard dialogue text to obtain a dictionary corresponding to the standard dialogue text; a generation module 304, used to generate a dictionary corresponding to the multiple speech keywords based on the multiple speech category templates in the specific marketing field; and an identification module 305, used to match the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords to identify whether there are target speech keywords in the standard dialogue text. As Figure 4This is a diagram of a speech recognition system provided by the second embodiment of the present invention.

[0120] The principles and technical effects of the marketing speech recognition system provided by the second embodiment of the present invention are the same as those of the first embodiment described above, and will not be described in detail here.

[0121] Furthermore, the construction module for constructing a keyword dictionary in a specific marketing field includes: constructing a keyword dictionary in the specific marketing field according to keywords in the specific marketing field and incorrect expressions corresponding to the keywords.

[0122] Furthermore, the first acquisition module is used to correct incorrect expressions in the original dialogue text using the keyword dictionary of the specific marketing field to obtain a standard dialogue text, including: matching the original dialogue text with the keyword dictionary of the specific marketing field, and when an incorrect expression corresponding to the original dialogue text is matched in the keyword dictionary, correcting the original dialogue text to obtain a standard dialogue text.

[0123] Furthermore, the multiple speech keywords include greeting keywords, promotional speech keywords, blacklist speech keywords and custom keywords; the identification module is used to match the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords, and identify whether there are target speech keywords in the standard dialogue text, including: when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any of the multiple corresponding speech keywords are not matched, it is determined that there are no target speech keywords in the standard dialogue text; or, when the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple corresponding speech keywords is matched, the matched speech keywords are compared with the target speech keywords; and if the matched keywords are consistent with the target speech keywords, it is determined that the standard dialogue text contains the target speech keywords.

[0124] Furthermore, the recognition system also includes: a third acquisition module, which is used to obtain the original conversation voice between customer service and customers in the target store within a set time period before executing the step of constructing the keyword dictionary and standard dictionary set for the specific marketing field; and a fourth acquisition module, which is used to perform voice recognition on the original conversation voice to obtain the original conversation text.

[0125] The third embodiment of the present invention also provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the marketing speech recognition method as described above.

[0126] The method and effect of identifying marketing speech executed by an electronic device provided in the third embodiment of the present invention are the same as those in the first embodiment above, and will not be described in detail here.

[0127] A fourth embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the marketing speech recognition method as described above.

[0128] The method and effect of identifying marketing words executed by a computer-readable storage medium provided in the fifth embodiment of the present invention are the same as those of the first embodiment above and will not be described in detail here.

[0129] A fifth embodiment of the present invention further provides a computer program product, comprising a computer program, which implements the marketing speech recognition method as described above when executed by a processor.

[0130] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0131] An embodiment of the present invention provides a device comprising a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are performed: (method claim steps, sole claim + subordinate claim). The device herein may be a server, a PC, a PAD, a mobile phone, or the like.

[0132] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0133] The present application 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 application. 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 produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. 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.

[0134] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0136] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0137] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0138] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0139] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0140] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for identifying marketing rhetoric, characterized in that: The identification method comprises: Construct keyword dictionaries and standard dictionary collections for specific marketing fields; Correcting incorrect expressions in the original dialogue text using a keyword dictionary in the specific marketing field to obtain a standard dialogue text; Using the standard dictionary set and the dictionary tree to perform word segmentation processing on the standard dialogue text, and obtaining a dictionary corresponding to the standard dialogue text; Generating a dictionary corresponding to the multiple speech keywords according to the multiple speech category templates in the specific marketing field; and The dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the plurality of speech keywords to identify whether the target speech keyword is included in the standard dialogue text.

2. The identification method according to claim 1, characterized in that: Building a keyword dictionary for a specific marketing area includes: A keyword dictionary for the specific marketing field is constructed based on the keywords for the specific marketing field and the incorrect statements corresponding to the keywords.

3. The identification method according to claim 2, characterized in that: The method of correcting incorrect expressions in the original dialogue text by using the keyword dictionary in the specific marketing field to obtain the standard dialogue text includes: The original dialogue text is matched with the keyword dictionary of the specific marketing field. When an erroneous statement corresponding to the original dialogue text is matched in the keyword dictionary, the original dialogue text is corrected to obtain a standard dialogue text.

4. The identification method according to claim 1, characterized in that: The multiple speech keywords include greeting keywords, promotion speech keywords, blacklist speech keywords and custom keywords; The step of matching the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the plurality of speech keywords to identify whether the target speech keyword is in the standard dialogue text comprises: When the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any of the multiple speech keywords is not matched, it is determined that there is no target speech keyword in the standard dialogue text; or, When the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple speech keywords is matched, the matched speech keyword is compared with the target speech keyword; and, If the matched keywords are consistent with the target speech keywords, it is determined that the target speech keywords are included in the standard dialogue text.

5. The identification method according to claim 1, characterized in that: Before executing the step of constructing a keyword dictionary and a standard dictionary set for a specific marketing field, the identification method further includes: The original conversation voice between the customer service and the customer of the target store within a set time period is obtained; and the original conversation voice is subjected to voice recognition to obtain the original conversation text.

6. A marketing speech recognition system, characterized in that: The identification system comprises: Building modules for building keyword dictionaries and standard dictionary sets for specific marketing fields; A first acquisition module is used to correct incorrect expressions in the original dialogue text using the keyword dictionary of the specific marketing field to obtain a standard dialogue text; A second acquisition module is used to perform word segmentation processing on the standard dialogue text using the standard dictionary set and the dictionary tree to acquire a dictionary corresponding to the standard dialogue text; A generating module, for generating a dictionary corresponding to the multiple speech keywords according to the multiple speech category templates in the specific marketing field; and The identification module is used to match the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords, and identify whether the target speech keyword exists in the standard dialogue text.

7. The identification system according to claim 6, characterized in that: The building blocks for constructing a keyword dictionary for a specific marketing field include: A keyword dictionary for the specific marketing field is constructed based on the keywords for the specific marketing field and the incorrect statements corresponding to the keywords.

8. The identification system according to claim 7, characterized in that: The first acquisition module is used to correct incorrect expressions in the original dialogue text using the keyword dictionary in the specific marketing field to obtain the standard dialogue text, including: The original dialogue text is matched with the keyword dictionary of the specific marketing field. When an erroneous statement corresponding to the original dialogue text is matched in the keyword dictionary, the original dialogue text is corrected to obtain a standard dialogue text.

9. The identification system according to claim 6, characterized in that: The multiple speech keywords include greeting keywords, promotion speech keywords, blacklist speech keywords and custom keywords; The identification module is used to match the dictionary corresponding to the standard dialogue text with the dictionary corresponding to the multiple speech keywords, and to identify whether the target speech keyword is in the standard dialogue text, including: When the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if any of the multiple speech keywords is not matched, it is determined that there is no target speech keyword in the standard dialogue text; or, When the dictionary corresponding to the standard dialogue text is matched with the dictionary corresponding to the multiple speech keywords, if at least one of the multiple speech keywords is matched, the matched speech keyword is compared with the target speech keyword; and, If the matched keywords are consistent with the target speech keywords, it is determined that the target speech keywords are included in the standard dialogue text.

10. The identification system according to claim 6, characterized in that: The identification system further comprises: A third acquisition module is used to acquire the original conversation voice between the customer service and the customer of the target store within a set time period before executing the step of constructing the keyword dictionary and the standard dictionary set in the specific marketing field; and The fourth acquisition module is used to perform speech recognition on the original conversation speech to obtain the original conversation text.

11. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the marketing speech recognition method according to any one of claims 1-5.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the marketing speech recognition method according to any one of claims 1 to 5.

13. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the method for identifying marketing speech as described in any one of claims 1-5.

Images