A real estate industry-based recorded text label system construction method and system

Abstract

The embodiment of the application discloses a kind of based on real estate industry's recording text label system construction method and system, first transcribe the recording of project site, generate first transcription text set, utilize first transcription text set to generate primary label and primary label system with business requirement analysis, screen first transcription text set, obtain second transcription text set, the text in second transcription text set is executed word segmentation and the frequency of each word segmentation is calculated and sorted, select the high-frequency word of preset quantity as secondary label, according to the corresponding relationship of secondary label and the text of second transcription text set, obtain the first mapping relationship of secondary label and primary label, using first mapping relationship, secondary label is included in primary label system, generate secondary label system.The embodiment of the application is reduced by the above method, the dependence of system construction on the experience of designer judgment, can meet the demand of expanding new scene, new knowledge of business cognition and induction.

Description

Technical Field

[0001] This invention relates to the field of big data technology, specifically to a method and system for constructing a recording text tagging system based on the real estate industry. Background Technology

[0002] Currently, the technology for extracting tags from audio recordings and building a system in the real estate field is not yet mature. Most tags are generated from business data and lack a category system for organizing, managing, and planning tags. Therefore, it is necessary to refer to the tag systems of other fields to build a system that allows users to quickly find data and tags based on an understandable classification method.

[0003] The establishment of a tag system is inseparable from the design of tags. Tags can be divided into three types according to their source: basic, processing, and mining. Basic tags can be generated from business data. Processing tags are generally statistical indicators that can be calculated based on business data. Mining tags include data mining and text mining. Generally, the construction scheme of the mining tag system for text can be divided into two categories according to the scalability of the tag system: (1) The tag system is designed manually based on the text content and domain experience, and the tag information of the text data is marked by business personnel; (2) The tag system is designed based on the text data generated by the business and the business needs, and a keyword library and thesaurus are built with reference to domain experience. The word extraction and word frequency statistics of the business text are carried out in combination with the thesaurus and established regular expressions, and then the high-frequency words are included in the system thesaurus. Both of the above-mentioned tagging system construction schemes have their advantages and disadvantages. The former only requires designing the tagging system needed for the project and can be put into use at the lowest cost. However, the rationality of the tags depends on the domain knowledge of the designers, and the accuracy of the tags is affected by the users. It is a more practical scheme in the early stage of business development. As the business gradually develops and the text information becomes more and more diversified, the utilization of text data by this scheme will gradually become a constraint. It can only solve the current problem and cannot help the future business innovation scenarios.

[0004] The latter requires continuous cost investment, but effectively reduces the impact of designers on the tag system. Furthermore, the system's thesaurus has a certain expansion capability when facing diverse texts. However, the degree of system optimization depends on the richness of regular expressions and cannot provide inspiration for future business applications. Summary of the Invention

[0005] To address this, embodiments of the present invention provide a method and system for constructing a recording text tagging system based on the real estate industry, in order to solve the problem that the existing system construction relies too heavily on the experience and judgment of designers, and cannot meet the needs of cognition and summarization of new business scenarios and new knowledge.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] According to a first aspect of the present invention, a method for constructing a recording text tagging system based on the real estate industry is provided, comprising:

[0008] Collect all recorded conversations at the project site;

[0009] The recorded dialogue is transcribed to generate a first set of transcribed texts;

[0010] The first transcribed text set is used to perform business requirement analysis to generate primary tags and construct a primary tag system;

[0011] The first set of transcribed texts is filtered using the primary tags to obtain the second set of transcribed texts;

[0012] Perform word segmentation on the text in the second transcribed text set and calculate the word frequency of each segment;

[0013] Sort by word frequency and select a preset number of high-frequency words as secondary tags;

[0014] Based on the correspondence between the secondary tags and the texts in the second transcribed text set, a first mapping relationship between the secondary tags and the primary tags is obtained;

[0015] Using the first mapping relationship, the secondary tags are incorporated into the primary tag system to generate a secondary tag system.

[0016] Furthermore, the first transcribed text set is used to perform business requirement analysis to generate primary tags, and a primary tag system is constructed, including:

[0017] Based on the framework of people, things, and relationships, the attribute labels of the root node of the system are determined by analyzing data and business needs in the real estate sector.

[0018] Set static tags, dynamic tags, and mining tags for each attribute tag;

[0019] Based on the static tags, dynamic tags, and mining tags under each attribute tag, the first transcribed text set is used to analyze business requirements, generate corresponding first-level tags, and construct the first tag system.

[0020] The attribute tags include: person tags, project tags, and prefix tags.

[0021] Further, the first transcribed text set is filtered using the primary tags to obtain a second transcribed text set, including:

[0022] Retrieve the audio text data of the current first-level tag on a sentence-by-sentence basis;

[0023] The recorded text data is preprocessed, and a corpus is constructed.

[0024] New words are extracted using the corpus.

[0025] The extracted new words are filtered and selected using existing words in a general vocabulary database to generate usable new words;

[0026] The general lexicon can be expanded using the available new words;

[0027] Based on the expanded general lexicon and the preset first word segmentation model, the corpus in the corpus is segmented into the first word segmentation corpus.

[0028] The first word segmentation model is trained using the first word segmentation corpus as training data to obtain the second word segmentation model;

[0029] The second word segmentation model is used to perform second word segmentation on the corpus to obtain the second word segmentation corpus;

[0030] The first sentence vector and the second sentence vector are obtained by using the second word segmentation corpus corresponding to each first-level tag and the audio text data, respectively.

[0031] Calculate the cosine similarity between the first sentence vector and the second sentence vector corresponding to each first-level label;

[0032] Determine whether the cosine similarity reaches the first threshold;

[0033] If the cosine similarity reaches the first threshold, the second word segment corpus is used as the secondary tag corpus text, and it is determined whether there is a primary tag corresponding to the text that has not been processed into sentence vectors in the second word segment corpus and the audio text data;

[0034] If the cosine similarity does not reach the first threshold, it is directly determined whether there is a text corresponding to a first-level label in the second word segmentation corpus and the audio text data that has not undergone sentence vector processing;

[0035] If there are texts corresponding to first-level tags in the second word segmentation corpus and the audio text data that have not been processed into sentence vectors, then the sentence vector processing is performed on the next unprocessed text corresponding to the first-level tags in the loop.

[0036] If there are no texts corresponding to the first-level labels in the second word segmentation corpus and the recorded text data that have not undergone sentence vector processing, then the second transcribed text set is generated using the currently obtained second-level label corpus text.

[0037] Furthermore, extracting new words from the corpus also includes:

[0038] Perform third-order word segmentation on the corpus;

[0039] Calculate the left and right information entropy H and mutual information cohesion PMI of each word in the third word segmentation result;

[0040] Determine whether the left and right information entropy H reaches the second threshold and whether the mutual information cohesion degree PMI reaches the third threshold;

[0041] If the left and right information entropy H reaches the second threshold and the mutual information cohesion degree PMI reaches the third threshold, then the word segmentation is determined as a new word;

[0042] If the left and right information entropy H does not reach the second threshold or the mutual information cohesion degree PMI does not reach the third threshold, then the word segmentation is determined to be a non-new word.

[0043] Preferably, a method for constructing a recording text tagging system based on the real estate industry further includes:

[0044] The second transcribed text set is analyzed according to business requirements to obtain basic class tags, processing class tags, and a second mapping relationship between the basic class tags, the processing class tags and the first-level tags;

[0045] Using the second mapping relationship, the basic category tags and the processing category tags are incorporated into the secondary tag system;

[0046] Among them, the basic category tags are basic business attribute tags, and the processing category tags are statistical indicators.

[0047] Furthermore, the semantic words of the second tag are analyzed, and the semantic words are abstracted and processed;

[0048] Three-level tags are generated based on the abstracted and processed semantic words.

[0049] Based on the correspondence between the third-level tags and the second-level tags, the third-level tags are incorporated into the second-level tag system to construct a third-level tag system.

[0050] Furthermore, the three-level tagging system is run in the business data according to a preset time period;

[0051] Based on the results, calculate the coverage rate of each mining category label;

[0052] Determine whether the coverage rate has reached the fourth threshold;

[0053] If the coverage rate does not reach the fourth threshold, the mining category label is determined to be a redundant label, and the redundant label is deleted.

[0054] If the coverage rate reaches the fourth threshold, the mining category label is determined to be a non-redundant label, and the non-redundant label is retained.

[0055] The coverage rate is the proportion of texts with the actual tag present in the results to the total number of texts.

[0056] According to a second aspect of the present invention, a system for constructing a recording text tagging system based on the real estate industry is provided, the system comprising:

[0057] The audio recording module is used to comprehensively collect audio recordings of conversations on-site during the project.

[0058] The first transcription module is used to transcribe the dialogue recording to generate a first transcribed text set.

[0059] The primary tag system construction module is used to generate primary tags by performing business requirement analysis using the first transcribed text set, and to construct the primary tag system.

[0060] The two-level tag system construction module is used to perform the following steps:

[0061] The first set of transcribed texts is filtered using the primary tags to obtain the second set of transcribed texts;

[0062] Perform word segmentation on the text in the second transcribed text set and calculate the word frequency of each segment;

[0063] Sort by word frequency and select a preset number of high-frequency words as secondary tags;

[0064] Based on the correspondence between the secondary tags and the texts in the second transcribed text set, a second mapping relationship between the secondary tags and the primary tags is obtained;

[0065] Using the second mapping relationship, the secondary tags are incorporated into the primary tag system to generate a secondary tag system.

[0066] Furthermore, the system also includes a three-level tag system construction module, used to perform the following steps:

[0067] Analyze the semantic words of the second tag and perform abstract processing on the semantic words;

[0068] Three-level tags are generated based on the abstracted and processed semantic words.

[0069] Based on the correspondence between the third-level tags and the second-level tags, the third-level tags are incorporated into the second-level tag system to construct a third-level tag system.

[0070] Furthermore, the system also includes a condensation module for performing the following steps:

[0071] The three-level tagging system is run in the business data according to a preset time period;

[0072] Based on the results, calculate the coverage rate of each mining category label;

[0073] Determine whether the coverage rate has reached the fourth threshold;

[0074] If the coverage rate does not reach the fourth threshold, the mining category label is determined to be a redundant label, and the redundant label is deleted.

[0075] If the coverage rate reaches the fourth threshold, the mining category label is determined to be a non-redundant label, and the non-redundant label is retained.

[0076] The coverage rate is the proportion of texts with the actual tag present in the results to the total number of texts.

[0077] The embodiments of the present invention have the following advantages:

[0078] This invention discloses a method and system for constructing a text tagging system for audio recordings in the real estate industry. First, it comprehensively collects audio recordings of conversations at project sites. Then, it transcribes these recordings to generate a first transcribed text set. Using this first transcribed text set, it performs business requirement analysis to generate primary tags and constructs a primary tagging system. The primary tags are then used to filter the first transcribed text set, resulting in a second transcribed text set. The text in the second transcribed text set is segmented, and the word frequency of each segment is calculated. Based on the word frequency, a preset number of high-frequency words are selected as secondary tags. According to the correspondence between the secondary tags and the text in the second transcribed text set, a first mapping relationship between the secondary tags and the primary tags is obtained. Using this first mapping relationship, the secondary tags are incorporated into the primary tagging system, generating a secondary tagging system. Thus, this invention, through unsupervised learning methods using big data, effectively reduces the reliance on designers during the system construction process, improves data mining and natural language processing technologies, meets the needs of recognizing and summarizing new business scenarios and knowledge, and, through innovative tag abstraction link technology, enables the streamlined construction of a text tagging system for audio recordings in the real estate industry, applicable to various other application scenarios. Attached Figure Description

[0079] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0080] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0081] Figure 1 A schematic diagram of the logical structure of a recording text tagging system based on the real estate industry provided in this embodiment of the invention;

[0082] Figure 2 A flowchart illustrating a method for constructing a recording text tagging system based on the real estate industry, provided in an embodiment of the present invention;

[0083] Figure 3 This is a flowchart illustrating the construction of a primary tag system in a method for constructing a recording text tag system based on the real estate industry, as provided in an embodiment of the present invention.

[0084] Figure 4 A schematic diagram illustrating the process of obtaining the second transcribed text set in a method for constructing a recording text tagging system based on the real estate industry, provided in an embodiment of the present invention;

[0085] Figure 5 This is a flowchart illustrating the process of extracting new words from a corpus in a method for constructing a recording text tagging system based on the real estate industry, as provided in an embodiment of the present invention. Detailed Implementation

[0086] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0087] Tags can be categorized into three types based on their source: basic, processed, and mined. Basic tags can be generated from business data. Processed tags are generally statistical indicators that can be calculated from business data. Mined tags include data mining and text mining. Generally, the construction scheme for a text mining tag system can be divided into two categories based on the scalability of the tag system: manual design of the tag system based on text content and domain experience, with the tag information of the text data being labeled by business personnel; and design of the tag system based on the text data generated by the business and business needs, and construction of a keyword library and thesaurus with reference to domain experience. The thesaurus and established regular expressions are used to extract words and perform word frequency statistics on the business text, and then high-frequency words are included in the system thesaurus.

[0088] Both of the above solutions have their advantages and disadvantages. The former only requires designing the labeling system needed for the project and can be put into use at the lowest cost. However, the rationality of the labels depends on the domain knowledge of the designers, and the accuracy of the labels is affected by the users. It is a more practical solution in the early stage of business development. As the business gradually develops and the text information becomes more and more diversified, the utilization of text data by this solution will gradually become a constraint. It can only solve the current problem and cannot help the future business innovation scenarios.

[0089] The latter requires continuous cost investment, but effectively reduces the impact of designers on the tag system. Furthermore, the system's lexicon has a certain expansion capability when facing diverse texts. However, the degree of system optimization depends on the richness of regular expressions and cannot provide inspiration for future business applications.

[0090] To address the issue that the aforementioned labeling system relies too heavily on the experience and judgment of designers and lacks the technical capacity to organize, manage, and plan the label category system.

[0091] refer to Figure 1 This invention discloses a recording text tagging system construction system based on the real estate industry. The system includes: a recording acquisition module 1, a first transcription module 2, and a first-level tagging system construction module 3.

[0092] This invention discloses a system for constructing a recording text tagging system based on the real estate industry. After collecting recordings from a project site, the system first transcribes the recorded conversations to generate a first transcribed text set. Then, it uses the first transcribed text set to perform business requirement analysis to generate primary tags and construct a primary tagging system. The first transcribed text set is then filtered using the primary tags to obtain a second transcribed text set. The text in the second transcribed text set is segmented, and the word frequency of each segment is calculated. Based on the word frequency, a preset number of high-frequency words are selected as secondary tags. According to the correspondence between the secondary tags and the text in the second transcribed text set, a first mapping relationship between the secondary tags and the primary tags is obtained. Using the first mapping relationship, the secondary tags are incorporated into the primary tagging system to generate a secondary tagging system.

[0093] Preferably, refer to Figure 1 The system further includes a secondary tag system construction module 4, which performs the following steps: filtering the first transcribed text set using the primary tags to obtain a second transcribed text set; performing word segmentation on the text in the second transcribed text set and calculating the word frequency of each segment; sorting by word frequency and selecting a preset number of high-frequency words as secondary tags; obtaining a first mapping relationship between the secondary tags and the primary tags based on the correspondence between the secondary tags and the text in the second transcribed text set; and incorporating the secondary tags into the primary tag system using the first mapping relationship to generate a secondary tag system.

[0094] Preferably, refer to Figure 1 The system further includes a three-level tag system construction module 5, which is used to perform the following steps: analyze the semantic words of the second tag, and abstract and process the semantic words; generate a three-level tag based on the abstracted and processed semantic words; and incorporate the three-level tag into the two-level tag system according to the correspondence between the three-level tag and the second-level tag to construct a three-level tag system.

[0095] Preferably, refer to Figure 1 The system further includes a refinement module 6, used to perform the following steps: running the three-level tag system in business data according to a preset time period; calculating the coverage rate of each mining category tag based on the running results; determining whether the coverage rate reaches a fourth threshold; if the coverage rate does not reach the fourth threshold, determining that the mining category tag is a redundant tag and deleting the redundant tag; if the coverage rate reaches the fourth threshold, determining that the mining category tag is a non-redundant tag and retaining the non-redundant tag; wherein, the coverage rate is: the proportion of the number of texts with the actual tag in the running results to the total number of texts.

[0096] Corresponding to the aforementioned system for constructing a recording text tagging system based on the real estate industry, this invention also discloses a method for constructing a recording text tagging system based on the real estate industry. The following details the method for constructing a recording text tagging system based on the real estate industry disclosed in this invention, in conjunction with the aforementioned system for constructing a recording text tagging system based on the real estate industry.

[0097] refer to Figure 2 This invention discloses a method for constructing a recording text tagging system based on the real estate industry, comprising: comprehensively collecting dialogue recordings at the project site through a recording acquisition module 1; transcribing the dialogue recordings using a first transcription module 2 to generate a first transcribed text set; using a first-level tagging system construction module 3 to perform business requirement analysis using the first transcribed text set to generate first-level tags and construct a first-level tagging system; and using a second-level tagging system construction module 4 to complete the following steps: filtering the first transcribed text set using the first-level tags to obtain a second transcribed text set; performing word segmentation on the text in the second transcribed text set and calculating the word frequency of each segment; sorting by word frequency and selecting a preset number of high-frequency words as second-level tags; obtaining a first mapping relationship between the second-level tags and the first-level tags based on the correspondence between the second-level tags and the text in the second transcribed text set; and incorporating the second-level tags into the first-level tagging system using the first mapping relationship to generate a second-level tagging system.

[0098] The evaluation criteria for the first transcribed text set of the first transcription module include: AB role hit rate: the probability of correctly distinguishing the speaker in the recording; omission rate: the probability of missing words in the recording transcription process; sentence error rate: the probability of errors in the recording transcription; and BLEU score: the score for the correct transcription of core domain words is evaluated with reference to the machine translation BLEU evaluation criteria.

[0099] Further, refer to Figure 3 The system utilizes the first transcribed text set to perform business requirement analysis, generate primary tags, and construct a primary tag system. This includes: using people, things, and relationships as the basic framework, and analyzing real estate data and business requirements to determine the attribute tags of the root node of the system; setting static tags, dynamic tags, and mining tags for each attribute tag; and using the first transcribed text set to perform business requirement analysis based on the static tags, dynamic tags, and mining tags under each attribute tag, generating corresponding primary tags and constructing the first tag system. The attribute tags include: person tags, project tags, and prerequisite tags.

[0100] Further, refer to Figure 4The process involves filtering the first transcribed text set using the first-level tags to obtain a second transcribed text set, including: acquiring the recorded text data matched by the current first-level tag on a sentence-by-sentence basis; preprocessing the recorded text data and constructing a corpus; extracting new words from the corpus; filtering and selecting usable new words using existing words in a general lexicon; expanding the general lexicon using the usable new words; performing first word segmentation on the corpus based on the expanded general lexicon and a preset first word segmentation model to obtain a first word segmentation corpus; training the first word segmentation model using the first word segmentation corpus as training data to obtain a second word segmentation model; performing second word segmentation on the corpus using the second word segmentation model to obtain a second word segmentation corpus; and obtaining a first sentence vector and a second sentence vector using the second word segmentation corpus corresponding to each first-level tag and the recorded text data, respectively. The process involves: calculating the cosine similarity between the first sentence vector and the second sentence vector corresponding to each first-level label; determining whether the cosine similarity reaches a first threshold; if the cosine similarity reaches the first threshold, using the second word segment corpus as the second-level label corpus text, and determining whether there are any first-level label corresponding texts in the second word segment corpus and the recorded text data that have not undergone sentence vector processing; if the cosine similarity does not reach the first threshold, directly determining whether there are any first-level label corresponding texts in the second word segment corpus and the recorded text data that have not undergone sentence vector processing; if there are first-level label corresponding texts in the second word segment corpus and the recorded text data that have not undergone sentence vector processing, then iteratively processing the next unprocessed first-level label corresponding text; if there are no first-level label corresponding texts in the second word segment corpus and the recorded text data that have not undergone sentence vector processing, then generating a second transcribed text set using the currently obtained second-level label corpus text.

[0101] Furthermore, refer to Figure 5 Extracting new words from the corpus includes: performing third-order word segmentation on the corpus; calculating the left and right information entropy H and mutual information cohesion degree PMI of each word segment in the third-order word segmentation result; the formula for calculating the left and right information entropy H is as follows:

[0102] H(X)=-∑ x∈X p(x)log2p(x)

[0103] Where p(x) is the probability of the text T1' appearing, and X is a random variable; the formula for calculating the mutual information cohesion degree PMI is as follows:

[0104]

[0105] Where x and y are variables; it is determined whether the left and right information entropy H reaches the second threshold and whether the mutual information cohesion PMI reaches the third threshold; if the left and right information entropy H reaches the second threshold and the mutual information cohesion PMI reaches the third threshold, then the word segmentation is determined as a new word; if the left and right information entropy H does not reach the second threshold or the mutual information cohesion PMI does not reach the third threshold, then the word segmentation is determined as a non-new word.

[0106] Preferably, the second transcribed text set is analyzed according to business needs to obtain basic category tags, processing category tags, and a second mapping relationship between the basic category tags, the processing category tags, and the first-level tags; using the second mapping relationship, the basic category tags and the processing category tags are incorporated into the second-level tag system; wherein, the basic category tags are basic business attribute tags, and the processing category tags are statistical indicators.

[0107] Furthermore, the semantic words of the second tag are analyzed and abstracted; a third-level tag is generated based on the abstracted semantic words; and the third-level tag is incorporated into the second-level tag system according to the correspondence between the third-level tag and the second-level tag to construct a third-level tag system.

[0108] Furthermore, the three-level tagging system is run in the business data according to a preset time period; based on the running results, the coverage rate of each mining category tag is calculated; it is determined whether the coverage rate reaches a fourth threshold; if the coverage rate does not reach the fourth threshold, the mining category tag is determined to be a redundant tag and the redundant tag is deleted; if the coverage rate reaches the fourth threshold, the mining category tag is determined to be a non-redundant tag and the non-redundant tag is retained; wherein, the coverage rate is: the proportion of the number of texts with the actual tag in the running results to the total number of texts.

[0109] This invention, through unsupervised learning methods using big data, can effectively reduce the reliance on designers during the system construction process, improve data mining and natural language processing technologies, meet the needs of cognition and summarization of new business scenarios and knowledge, and through the innovation of tag abstraction link technology, can construct a real estate industry dialogue recording text tagging system in a process-oriented manner and be applicable to other various application scenarios.

[0110] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0111] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0112] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0113] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0114] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0115] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0116] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0117] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for constructing a recording text tagging system based on the real estate industry, characterized in that, The method includes: Collect all recorded conversations at the project site; The recorded dialogue is transcribed to generate a first set of transcribed texts; The first transcribed text set is used to perform business requirement analysis to generate primary tags and construct a primary tag system; The first set of transcribed texts is filtered using the primary tags to obtain the second set of transcribed texts; Perform word segmentation on the text in the second transcribed text set and calculate the word frequency of each segment; Sort by word frequency and select a preset number of high-frequency words as secondary tags; Based on the correspondence between the secondary tags and the texts in the second transcribed text set, a first mapping relationship between the secondary tags and the primary tags is obtained; Using the first mapping relationship, the secondary tags are incorporated into the primary tag system to generate a secondary tag system; The first transcribed text set is filtered using the primary tags to obtain a second transcribed text set, which includes: Retrieve the audio text data of the current first-level tag on a sentence-by-sentence basis; The recorded text data is preprocessed, and a corpus is constructed. New words are extracted using the corpus. The extracted new words are filtered and selected using existing words in a general vocabulary database to generate usable new words; The general lexicon can be expanded using the available new words; Based on the expanded general lexicon and the preset first word segmentation model, the corpus in the corpus is segmented into the first word segmentation corpus. The first word segmentation model is trained using the first word segmentation corpus as training data to obtain the second word segmentation model; The second word segmentation model is used to perform second word segmentation on the corpus to obtain the second word segmentation corpus; The first sentence vector and the second sentence vector are obtained by using the second word segmentation corpus corresponding to each first-level tag and the audio text data, respectively. Calculate the cosine similarity between the first sentence vector and the second sentence vector corresponding to each first-level label; Determine whether the cosine similarity reaches the first threshold; If the cosine similarity reaches the first threshold, the second word segment corpus is used as the secondary tag corpus text, and it is determined whether there is a primary tag corresponding to the text that has not been processed into sentence vectors in the second word segment corpus and the audio text data; If the cosine similarity does not reach the first threshold, it is directly determined whether there is a text corresponding to a first-level label in the second word segmentation corpus and the audio text data that has not undergone sentence vector processing; If there are texts corresponding to first-level tags in the second word segmentation corpus and the audio text data that have not been processed into sentence vectors, then the sentence vector processing is performed on the next unprocessed text corresponding to the first-level tags in the loop. If there are no texts corresponding to the first-level labels in the second word segmentation corpus and the recorded text data that have not undergone sentence vector processing, then the second transcribed text set is generated using the currently obtained second-level label corpus text.

2. The method for constructing a recording text tagging system based on the real estate industry as described in claim 1, characterized in that, Using the first transcribed text set, business requirements analysis is performed to generate primary tags, and a primary tag system is constructed, including: Based on the framework of people, things, and relationships, the attribute labels of the root node of the system are determined by analyzing data and business needs in the real estate sector. Set static tags, dynamic tags, and mining tags for each attribute tag; Based on the static tags, dynamic tags, and mining tags under each attribute tag, the first transcribed text set is used to analyze business requirements, generate corresponding first-level tags, and construct the first tag system. The attribute tags include: person tags, project tags, and prefix tags.

3. The method for constructing a recording text tagging system based on the real estate industry as described in claim 1, characterized in that, Extracting new words from the corpus includes: Perform third-order word segmentation on the corpus; Calculate the left and right information entropy H and mutual information cohesion PMI of each word in the third word segmentation result; Determine whether the left and right information entropy H reaches the second threshold and whether the mutual information cohesion degree PMI reaches the third threshold; If the left and right information entropy H reaches the second threshold and the mutual information cohesion degree PMI reaches the third threshold, then the word segmentation is determined as a new word; If the left and right information entropy H does not reach the second threshold or the mutual information cohesion degree PMI does not reach the third threshold, then the word segmentation is determined to be a non-new word.

4. The method for constructing a recording text tagging system based on the real estate industry as described in claim 1, characterized in that, The method further includes: The second transcribed text set is analyzed according to business requirements to obtain basic class tags, processing class tags, and a second mapping relationship between the basic class tags, the processing class tags and the first-level tags; Using the second mapping relationship, the basic category tags and the processing category tags are incorporated into the secondary tag system; Among them, the basic category tags are basic business attribute tags, and the processing category tags are statistical indicators.

5. The method for constructing a recording text tagging system based on the real estate industry as described in claim 4, characterized in that, The method further includes: Analyze the semantic words of the second tag and perform abstract processing on the semantic words; Three-level tags are generated based on the abstracted and processed semantic words. Based on the correspondence between the third-level tags and the second-level tags, the third-level tags are incorporated into the second-level tag system to construct a third-level tag system.

6. The method for constructing a recording text tagging system based on the real estate industry as described in claim 5, characterized in that, The method further includes: The three-level tagging system is run in the business data according to a preset time period; Based on the results, calculate the coverage rate of each mining category label; Determine whether the coverage rate has reached the fourth threshold; If the coverage rate does not reach the fourth threshold, the mining category label is determined to be a redundant label, and the redundant label is deleted. If the coverage rate reaches the fourth threshold, the mining category label is determined to be a non-redundant label, and the non-redundant label is retained. The coverage rate is the proportion of texts with the actual tag present in the results to the total number of texts.

7. A system for constructing a recording text tagging system based on the real estate industry, characterized in that, The system includes: The audio recording module is used to comprehensively collect audio recordings of conversations on-site during the project. The first transcription module is used to transcribe the dialogue recording to generate a first transcribed text set. The primary tag system construction module is used to generate primary tags by performing business requirement analysis using the first transcribed text set, and to construct the primary tag system. The two-level tag system construction module is used to perform the following steps: The first set of transcribed texts is filtered using the primary tags to obtain the second set of transcribed texts; Perform word segmentation on the text in the second transcribed text set and calculate the word frequency of each segment; Sort by word frequency and select a preset number of high-frequency words as secondary tags; Based on the correspondence between the secondary tags and the texts in the second transcribed text set, a first mapping relationship between the secondary tags and the primary tags is obtained; Using the first mapping relationship, the secondary tags are incorporated into the primary tag system to generate a secondary tag system; The first transcribed text set is filtered using the primary tags to obtain a second transcribed text set, which includes: Retrieve the audio text data of the current first-level tag on a sentence-by-sentence basis; The recorded text data is preprocessed, and a corpus is constructed. New words are extracted using the corpus. The extracted new words are filtered and selected using existing words in a general vocabulary database to generate usable new words; The general lexicon can be expanded using the available new words; Based on the expanded general lexicon and the preset first word segmentation model, the corpus in the corpus is segmented into the first word segmentation corpus. The first word segmentation model is trained using the first word segmentation corpus as training data to obtain the second word segmentation model; The second word segmentation model is used to perform second word segmentation on the corpus to obtain the second word segmentation corpus; The first sentence vector and the second sentence vector are obtained by using the second word segmentation corpus corresponding to each first-level tag and the audio text data, respectively. Calculate the cosine similarity between the first sentence vector and the second sentence vector corresponding to each first-level label; Determine whether the cosine similarity reaches the first threshold; If the cosine similarity reaches the first threshold, the second word segment corpus is used as the secondary tag corpus text, and it is determined whether there is a primary tag corresponding to the text that has not been processed into sentence vectors in the second word segment corpus and the audio text data; If the cosine similarity does not reach the first threshold, it is directly determined whether there is a text corresponding to a first-level label in the second word segmentation corpus and the audio text data that has not undergone sentence vector processing; If there are texts corresponding to first-level tags in the second word segmentation corpus and the audio text data that have not been processed into sentence vectors, then the sentence vector processing is performed on the next unprocessed text corresponding to the first-level tags in the loop. If there are no texts corresponding to the first-level labels in the second word segmentation corpus and the recorded text data that have not undergone sentence vector processing, then the second transcribed text set is generated using the currently obtained second-level label corpus text.

8. The audio recording text tagging system construction system based on the real estate industry as described in claim 7, characterized in that, The system also includes a three-level tag system construction module, used to perform the following steps: Analyze the semantic words of the second tag and perform abstract processing on the semantic words; Three-level tags are generated based on the abstracted and processed semantic words. Based on the correspondence between the third-level tags and the second-level tags, the third-level tags are incorporated into the second-level tag system to construct a third-level tag system.

9. A system for constructing a recording text tagging system based on the real estate industry as described in claim 8, characterized in that, The system also includes a condensation module for performing the following steps: The three-level tagging system is run in the business data according to a preset time period; Based on the results, calculate the coverage rate of each mining category label; Determine whether the coverage rate has reached the fourth threshold; If the coverage rate does not reach the fourth threshold, the mining category label is determined to be a redundant label, and the redundant label is deleted. If the coverage rate reaches the fourth threshold, the mining category label is determined to be a non-redundant label, and the non-redundant label is retained. The coverage rate is the proportion of texts with the actual tag present in the results to the total number of texts.

Images