Electric customer service semantic analysis method and device based on AI large model and medium

Abstract

The invention provides an electric power customer service semantic analysis method and device based on an AI large model and a medium, and the method comprises the steps: carrying out the slicing of a received electric power consultation text based on a natural language, and obtaining a plurality of consultation text segments; performing semantic relationship analysis on each consultation text fragment based on the semantic understanding ability of the AI large model, and constructing a semantic relationship graph according to the semantic relationship between words in each consultation text fragment; determining a target power business scene associated with each consultation text fragment based on the semantic relationship graph and the power business knowledge graph; performing cross-fragment semantic integration on each consultation text fragment based on the logic relation between the consultation text fragments to obtain a cross-fragment semantic integration text; and outputting a reply text based on the cross-fragment semantic integration text and the target power business scene. According to the invention, the customer consultation can be comprehensively and accurately processed, the efficiency and accuracy of the electric power consultation service are remarkably improved, and thus the service experience of the user is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular, to a semantic analysis method, device, and medium for power customer service based on an AI large model. Background Art

[0002] In the power customer service business, currently, a semantic analysis method based on rule matching and simple keyword extraction is generally adopted. First, a knowledge base containing a large number of keywords related to power services and corresponding service scenarios is established. When the customer service system receives a customer's consultation text, the system performs word segmentation on the text, splitting the text into individual words. Then, keywords are extracted from these words and matched with the keywords in the knowledge base. If a corresponding keyword is matched, the customer's consultation intention is judged according to the preset rules in the knowledge base, and then associated with the corresponding service processing flow.

[0003] Although the semantic analysis method for power customer service based on rule matching and simple keyword extraction can handle common customer consultations to a certain extent, its ability to understand complex semantics is insufficient, and it is difficult to accurately handle customer problems with fuzzy semantics, diverse expressions, or involving multiple service scenario associations. Moreover, the existing method may only be able to identify some keywords and cannot accurately understand multiple complex intentions involved by customers at the same time, resulting in inaccurate and incomplete problem handling, affecting the customer experience and service efficiency. With the continuous expansion of power services and the increasing diversification of customer needs, this method is no longer able to meet the actual business requirements. Summary of the Invention

[0004] The present invention provides a semantic analysis method, device, and medium for power customer service based on an AI large model, aiming to improve the efficiency and accuracy of power consultation services and enhance the user service experience.

[0005] In a first aspect, the present invention provides a semantic analysis method for power customer service based on an AI large model, including:

[0006] Slicing the received power consultation text based on natural language to obtain multiple consultation text segments;

[0007] Performing semantic relationship parsing on each consultation text segment based on the semantic understanding ability of the AI large model, and constructing a semantic relationship graph according to the semantic relationships between the words in each consultation text segment;

[0008] Determining the target power service scenario associated with each consultation text segment based on the semantic relationship graph and the power service knowledge graph;

[0009] Performing cross-segment semantic integration on each consultation text segment based on the logical connection between each consultation text segment to obtain a cross-segment semantic integration text;

[0010] Output a response text based on the cross - segment semantic integrated text and the target power business scenario.

[0011] In a second aspect, the present invention also provides a power customer service semantic analysis device based on an AI large model, which is applied to the power customer service semantic analysis method based on an AI large model as described in the first aspect; the power customer service semantic analysis device based on an AI large model includes:

[0012] A text slicing module, configured to slice the received power consultation text based on natural language to obtain multiple consultation text segments;

[0013] A semantic relationship parsing module, configured to parse the semantic relationship of each consultation text segment based on the semantic understanding ability of the AI large model, and construct a semantic relationship graph according to the semantic relationship between words in each consultation text segment;

[0014] A business scenario association module, configured to determine the target power business scenario associated with each consultation text segment based on the semantic relationship graph and the power business knowledge graph;

[0015] A text integration module, configured to perform cross - segment semantic integration on each consultation text segment based on the logical connection between each consultation text segment to obtain a cross - segment semantic integrated text;

[0016] An intention recognition and output module, configured to output a response text based on the cross - segment semantic integrated text and the target power business scenario.

[0017] In a third aspect, the present invention also provides an electronic device, including: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby implementing the power customer service semantic analysis method based on an AI large model as described in any one of the above.

[0018] In a fourth aspect, the present invention also provides a non - transitory computer - readable storage medium, in which a computer software program is stored, and when the computer software program is executed by a processor, the power customer service semantic analysis method based on an AI large model as described in any one of the above is implemented.

[0019] In a fifth aspect, the present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the power customer service semantic analysis method based on an AI large model as described above is implemented.

[0020] The power customer service semantic analysis method based on the AI large model provided by the embodiments of the present invention can adapt to various expression methods through the powerful semantic understanding ability of the AI large model. As long as the semantics are the same or similar, the intention can be accurately understood. Further, through text segmentation, semantic relationship parsing, and cross-fragment semantic integration, the complex semantics in the power consultation text can be deeply understood. At the same time, through associated business scenario recognition and cross-fragment semantic integration, the relationships between different business scenarios can be clearly sorted out, so that multiple complex consultation intentions in the user consultation text can be accurately identified, and thus customer consultations can be comprehensively and accurately processed, significantly improving the efficiency and accuracy of power consultation services, and thereby improving the user service experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a schematic flowchart of the power customer service semantic analysis method based on the AI large model provided by the embodiments of the present invention;

[0022] Figure 2 is a schematic structural diagram of the power customer service semantic analysis device based on the AI large model provided by the embodiments of the present invention;

[0023] Figure 3 is an embodiment diagram of the electronic device provided by the embodiments of the present invention;

[0024] Figure 4 is an embodiment diagram of the computer-readable storage medium provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0026] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0027] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0028] Refer to Figure 1 , Figure 1 FIG. is a schematic flowchart of a semantic analysis method for power customer service based on an AI large model provided by the present invention. In the embodiments of the present invention, the execution subject of the semantic analysis method for power customer service based on the AI large model is a power customer service device. Therefore, the semantic analysis method for power customer service based on the AI large model includes:

[0029] Step 10: Slice the received power consultation text based on natural language to obtain multiple consultation text segments.

[0030] Optionally, when a user consults about a power problem, the user needs to input a power consultation text in the power customer service device. Therefore, after receiving the power consultation text, the power customer service device divides the text according to specific rules based on natural language processing technology to obtain multiple consultation text segments, as specifically described in Steps 101 to 105. Among them, the specific rules can be determined based on punctuation marks (such as full stops, question marks, exclamation marks, etc.), specific keywords (such as words indicating content conversion like "in addition", "moreover", etc.), and semantic coherence to determine the slicing positions. In one embodiment, the power customer service device receives the consultation text "The electricity bill of my home has suddenly become higher this month. What's the reason? In addition, the community has often had power outages recently. What's going on?". The power customer service device first recognizes the full stop and the keyword "in addition", and slices the text into two segments: Segment 1 is "The electricity bill of my home has suddenly become higher this month. What's the reason?"; Segment 2 is "The community has often had power outages recently. What's going on?".

[0031] Step 20: Parse the semantic relationships of each consultation text segment based on the semantic understanding ability of the AI large model, and construct a semantic relationship graph according to the semantic relationships between the words in each consultation text segment.

[0032] Further, for each consulting text segment, the power customer service device analyzes the semantic relationships between the words in the segment using the semantic understanding ability of the AI large model, such as subject-predicate relationships, verb-object relationships, modification relationships, etc., and presents these relationships in the form of a graph to construct a semantic relationship graph, specifically as described in steps 201 to 204. The AI large model is obtained by learning a large amount of text data and can accurately judge the semantic associations between words, providing a basis for determining the subsequent business scenarios.

[0033] Continuing with the above embodiment, for segment 1 "The electricity bill of my home has suddenly become higher this month. What's the reason?", the AI large model analyzes that "my home" is the subject, "the electricity bill has become higher" is the predicate, "this month" is the time adverbial, and "reason" is the object being asked about. In the constructed semantic relationship graph, there is a subject-predicate association edge between "my home" and "the electricity bill has become higher", a time modification association edge between "this month" and "the electricity bill has become higher", and a causal association query pointing edge between "reason" and "the electricity bill has become higher".

[0034] Step 30: Determine the target power business scenario associated with each consulting text segment based on the semantic relationship graph and the power business knowledge graph.

[0035] Further, the power customer service device matches the constructed semantic relationship graph with the power business knowledge graph. The power business knowledge graph contains various power business concepts, attributes, and the relationships between them. Therefore, by comparing the elements in the semantic relationship graph with the nodes and relationships in the business knowledge graph, the target power business scenario that best matches each consulting text segment is obtained, specifically as described in steps 301 to 304.

[0036] Continuing with the above embodiment, for segment 1 "The electricity bill of my home has suddenly become higher this month. What's the reason?", the problem of abnormal electricity bills is highlighted in the semantic relationship graph. There is a "power bill anomaly analysis" business scenario node in the power business knowledge graph, and this node has relevant relationships with sub-nodes such as "power consumption situation", "tariff adjustment", and "electrical appliance failure". Through matching, it is determined that the target power business scenario associated with this segment is the "power bill anomaly analysis" scenario. For segment 2 "There have been frequent power outages in the community recently. What's going on?", the power outage problem is reflected in the semantic relationship graph, which matches the "power outage fault troubleshooting" business scenario node in the power business knowledge graph, and its target power business scenario is determined to be "power outage fault troubleshooting".

[0037] Step 40: Perform cross-segment semantic integration on each consulting text segment based on the logical connection between each consulting text segment to obtain a cross-segment semantic integration text.

[0038] Further, the power customer service device analyzes the logical connections between each consultation text segment, such as causal relationship, parallel relationship, progressive relationship, etc., and integrates each consultation text segment according to these logical connections to form a cross-segment semantic integration text with coherent semantics, so as to understand the user's consultation intention as a whole, as specifically described in Steps 401 to 404.

[0039] Continuing with the above embodiment, Segment 1 and Segment 2 are in a parallel relationship and are both consultations from the user regarding power issues. The power customer service device integrates them into "The user inquired about the reason for the sudden increase in electricity bills and the reason for frequent power outages in the community." If the text contains "Because of the power outage in the community, all the things in my family's refrigerator are broken, and the electricity bill also seems to have increased", then there is a causal relationship between the segments, and it is integrated into "The user stated that the power outage in the community caused damage to the refrigerator items and suspected increase in electricity bills, and inquired about the relevant reasons."

[0040] Step 50: Output a reply text based on the cross-segment semantic integration text and the target power business scenario.

[0041] Further, the power customer service device inputs each consultation text segment in the cross-segment semantic integration text into the business scenario reply module of its corresponding target power business scenario. The business scenario reply module will search the log library for similar historical consultations and reply records to match with the consultation text segment, and generate a reply result for each consultation text segment according to the matching result. Further, the power customer service device combines the reply results into a final reply text according to the priority of each consultation text segment in the cross-segment semantic integration text (which can be determined according to the order of user questions, the urgency of the questions, etc.).

[0042] Continuing with the above embodiment, for fragment 1 "The electricity bill of my family has suddenly become higher this month. What's the reason?", in the reply module of the "Analysis of Abnormal Electricity Bills" business scenario, the log library matches a similar query "The electricity bill suddenly increased last month", and the reply record is "The increase in the electricity bill may be due to the addition of high-power electrical appliances in the family this month, or a malfunction of the electricity meter. You can check the usage of the electrical appliances and contact us to arrange for someone to check the electricity meter." For fragment 2 "There have been frequent power outages in the community recently. What's going on?", in the reply module of the "Investigation of Power Outage Faults" business scenario, the log library matches "Inquiry about frequent power outages in the community", and the reply record is "The power outage in the community may be due to line maintenance or a sudden power failure. We are already investigating. Please pay attention to the community notice for the specific power restoration time." In the order of fragment 1 and fragment 2 (i.e., the order of user questions), the combined reply text is "The increase in the electricity bill may be due to the addition of high-power electrical appliances in the family this month, or a malfunction of the electricity meter. You can check the usage of the electrical appliances and contact us to arrange for someone to check the electricity meter. The power outage in the community may be due to line maintenance or a sudden power failure. We are already investigating. Please pay attention to the community notice for the specific power restoration time."

[0043] The embodiment of the present invention can adapt to various expression methods through the powerful semantic understanding ability of the AI large model. As long as the semantics are the same or similar, the intention can be accurately understood. Further, through text segmentation, semantic relationship parsing, and cross-fragment semantic integration, the complex semantics in the power consultation text can be deeply understood. At the same time, through associated business scenario recognition and cross-fragment semantic integration, the relationship between different business scenarios can be clearly sorted out, so that multiple complex consultation intentions in the user consultation text can be accurately identified, and thus the customer consultation can be processed comprehensively and accurately, significantly improving the efficiency and accuracy of the power consultation service, and thus improving the user service experience.

[0044] In one embodiment, the descriptions of steps 101 to 105 are as follows:

[0045] Step 101, split the power consultation text according to punctuation marks to obtain initial text fragments.

[0046] Optionally, after receiving the electricity consultation text, the power customer service device first divides the text into relatively independent initial text segments based on common punctuation marks, such as periods, question marks, exclamation marks, semicolons, etc., wherein punctuation marks are often used in natural language to distinguish complete semantic units. Continuing with the above embodiment, taking the electricity consultation text "My family's electricity bill has suddenly increased this month. What is the reason? In addition, there have been frequent power outages in the community recently. What's going on?" as an example, the power customer service device recognizes the question mark and comma in the text and divides it into two initial text segments. Segment 1 is "My family's electricity bill has suddenly increased this month. What is the reason?", and segment 2 is "In addition, there have been frequent power outages in the community recently. What's going on?".

[0047] Step 102: merge adjacent text segments in the initial text segments whose semantic coherence levels are greater than or equal to a preset coherence level threshold to obtain a merged text segment.

[0048] Furthermore, for the initial text segments, the power customer service device will evaluate the degree of semantic coherence between adjacent text segments, where the degree of semantic coherence is determined based on the text theme and the text context. If the degree of semantic coherence between adjacent text segments is greater than or equal to a preset coherence threshold, the power customer service device will merge the adjacent text segments into a new text segment. For example, if two adjacent segments both revolve around issues related to electricity bills and have a contextual connection, the merging condition is met.

[0049] In one embodiment, in a more complex electricity consultation text, there are two adjacent segments, segment A is "My family's electricity bill is ridiculously high this month", and segment B is "Is it because of the new high-power air conditioner?" The power customer service device determines that the semantic coherence between segments A and B is high, exceeding the preset coherence threshold, by analyzing the text theme (both are about the discussion of the reasons for the high electricity bill) and the context (segment B follows segment A to speculate on the reasons for the high electricity bill), and then merges segments A and B into "My family's electricity bill is ridiculously high this month, is it because of the new high-power air conditioner?" to obtain the merged text segment of segments A and B.

[0050] Step 103, identifying the power keywords of each merged text segment based on a preset power keyword dictionary, and performing topic clustering according to the power keywords and semantic information of each merged text segment to obtain a topic clustering result for each merged text segment.

[0051] Furthermore, a preset electric power keyword dictionary is built into the electric power customer service device. For each merged text segment, the electric power customer service device identifies the electric power keyword therein according to the electric power keyword dictionary.

[0052] Further, the power customer service device performs topic clustering on the merged text segments with similar power keywords and semantics according to the power keywords of each merged text segment and the semantic information of the combined text. For example, keywords related to electricity bills such as "electricity bill", "electricity price", "power consumption", etc. If multiple merged text segments contain these keywords and the semantics revolve around electricity bill-related issues, they will be grouped into one category.

[0053] In one embodiment, there are three merged text segments. Segment C is "The electricity bill of my family has suddenly increased this month. I don't know why", Segment D is "Recently, the electricity price has been adjusted. Will it affect my electricity bill?", and Segment E is "After the power outage in the community, when the power came back on, the electricity meter showed abnormal power consumption". The power customer service device, based on the power keyword dictionary, identifies that both Segment C and Segment D contain the power keyword "electricity bill", and the semantics revolve around electricity bill issues, and clusters Segment C and Segment D under the "Electricity Bill Problem" topic; although Segment E is related to electricity, the keywords and semantics are more inclined to the abnormal power consumption of the electricity meter after the power outage, and it is clustered under the "Power Outage and Electricity Meter Abnormality" topic.

[0054] Step 104, based on a preset correlation word dictionary, identify the correlation words associated with each merged text segment in the power consultation text, and perform a structural hierarchy division on the power consultation text based on the correlation words and topic clustering results of each merged text segment to obtain the text structural hierarchy of each merged text segment.

[0055] Further, a preset correlation word dictionary is built into the power customer service device. Therefore, the power customer service device, based on the preset correlation word dictionary, identifies the correlation words associated with each merged text segment in the original power consultation text, such as "in addition", "moreover", "because", "so", etc. Among them, these correlation words can reflect the logical relationship between texts.

[0056] Further, the power customer service device performs a structural hierarchy division on the power consultation text according to the correlation words of each merged text segment combined with the topic clustering results of each merged text segment, clarifies the structural position of each merged text segment in the entire consultation content, and determines its text structural hierarchy. For example, two segments connected by "in addition" usually belong to a parallel structural hierarchy.

[0057] Returning to the original example, Segment 1 "The electricity bill of my family has suddenly become higher this month. What's the reason?" and Segment 2 "In addition, the community has been experiencing frequent power outages recently. What's going on?", the power customer service device identifies the correlation word "in addition" in Segment 2, and combines the topic clustering results (Segment 1 is under the "Electricity Bill Problem" topic, and Segment 2 is under the "Power Outage Problem" topic), and determines that these two segments belong to a parallel structural hierarchy. The text structural hierarchy of Segment 1 is the first layer in the parallel structure, and Segment 2 is the second layer in the parallel structure.

[0058] Step 105 , dividing the merged text segments of the same topic clustering result and text structure level into the same slice to obtain a plurality of consulting text segments.

[0059] Furthermore, the electric power customer service device divides the merged text segments of the same subject clustering result and text structure level into the same slice, thereby obtaining multiple consulting text segments, which can make the internal semantics of each consulting text segment closely related, and have a clear positioning in the entire consulting text structure, so as to facilitate subsequent targeted semantic understanding and processing. Based on the analysis of the previous embodiment, segment 1 "My family's electricity bill has suddenly increased this month. What is the reason?" and segment 2 "In addition, there are frequent power outages in the community recently. What's going on?", because their subject clustering results are different ("electricity bill problem" and "power outage problem" respectively), but the text structure level is in a parallel relationship. So in the end, two consulting text segments are obtained, segment 1 is a slice around "electricity bill problem", and segment 2 is a slice around "power outage problem".

[0060] The embodiments of the present invention can systematically disassemble and classify complex electricity consultation texts to obtain multiple consultation text fragments with clear logic, coherent semantics and clear themes. These fragments can more accurately reflect different aspects of user consultations, facilitate subsequent semantic relationship analysis and determination of related electricity business scenarios based on the AI ​​big model, greatly improve the efficiency and accuracy of subsequent processing, enable the electricity customer service device to understand the user's intentions more accurately, and thus be able to comprehensively and accurately handle customer consultations, significantly improve the efficiency and accuracy of electricity consulting services, and thus improve the user service experience.

[0061] In one embodiment, the description of steps 201 to 205 is as follows:

[0062] Step 201: For each consultation text segment, perform grammatical analysis on the consultation text segment based on the AI ​​big model to identify words of various parts of speech in the consultation text segment.

[0063] Optionally, for each consultation text segment, the electric power customer service device uses the AI ​​big model to perform grammatical analysis on each consultation text segment. During this process, the AI ​​big model will identify words of various parts of speech in the consultation text segment, such as nouns, verbs, adjectives, adverbs, etc.

[0064] Taking the consultation text fragment "The electricity bill of my family has suddenly increased this month" as an example, the power customer service device uses the AI large model to perform grammatical parsing. It identifies that "my family" is a noun and serves as the subject of the sentence, representing the doer or the object of the action; "this month" is a noun phrase and acts as an adverbial of time, used to limit the time range when the event of "the electricity bill increasing" occurs; "the electricity bill" is a noun and is the core object of the sentence discussion; "suddenly" is an adverb that modifies the verb "increased", describing the state when the action of "increasing" occurs; "increased" is a verb and serves as the predicate of the sentence, expressing the change that has occurred to "the electricity bill of my family".

[0065] Step 202: Based on the third semantic feature vectors of each word, perform semantic relationship parsing on the consultation text fragment to identify the preliminary semantic relationships between each word in the consultation text fragment.

[0066] Furthermore, each word has a specific position in the semantic space corresponding to the consultation text fragment, and this position is characterized by the semantic feature vector of the word. Among them, the semantic feature vector is determined based on the contribution degree and semantic feature values of each semantic feature in the word. Therefore, the power customer service device uses the semantic feature vectors of each word to perform semantic relationship parsing on the consultation text fragment. By calculating relationships such as the similarity and direction between vectors, the preliminary semantic relationships between each word in the consultation text fragment are identified. For example, if the semantic feature vectors of two words are similar, they may have a close semantic connection.

[0067] Continuing with the above embodiment, the semantic feature vector of "my family" contains features related to the family, such as feature values related to the living place and family members; the semantic feature vector of "the electricity bill" contains feature values related to electricity consumption and cost calculation. Through calculation, it is found that the semantic feature vectors of "my family" and "the electricity bill" have a high degree of overlap in certain dimensions, indicating that there is an ownership relationship between them, that is, the "electricity bill" of "my family". At the same time, the semantic feature vector of "suddenly" and the semantic feature vector of "increased" are closely related in the dimension representing state change, initially determining that "suddenly" has a modifying semantic relationship with "increased".

[0068] Step 203: Based on the context information of each word, adjust the preliminary semantic relationships to obtain the strengthened semantic relationships between each word in the consultation text fragment.

[0069] Furthermore, although the preliminary semantic relationships provide the basic connections between words, there may be inaccuracies based solely on semantic feature vectors. Therefore, the power customer service device further adjusts the preliminary semantic relationships by considering the context information of each word, where the context information includes other words before and after the word, the context of the entire sentence, and the overall theme of the consultation text where the text fragment is located. By combining this information, the semantic relationships between words can be determined more accurately, resulting in enhanced semantic relationships.

[0070] Continuing with the above embodiment, in the text fragment "The electricity bill of my home has suddenly become higher this month. What's the reason?", from the perspective of the preliminary semantic relationship, there doesn't seem to be a directly close connection between "become higher" and "reason". However, considering the context, the whole sentence is asking about the reason for the "higher electricity bill". Therefore, by considering the context information, the power customer service device strengthens the causal semantic relationship between "become higher" and "reason", that is, "the electricity bill has become higher" is the result, and "reason" is the inquiry about the factor that causes this result.

[0071] Step 204, based on the AI large model, determine the semantic relationship direction between each pair of words according to the vector direction of the third semantic feature vector between each pair of words, and determine the semantic relationship level of the enhanced semantic relationship between each pair of words according to the semantic relationship direction and the enhanced semantic relationship between each pair of words.

[0072] Furthermore, the power customer service device once again relies on the AI large model to determine the semantic relationship direction between each pair of words according to the vector direction of the semantic feature vector between each pair of words. For example, in the subject-predicate relationship, the vector direction can point from the subject to the predicate, indicating that the subject performs the action of the predicate. At the same time, combining the enhanced semantic relationship, determine the semantic relationship level of the enhanced semantic relationship between each pair of words, where the semantic relationship level can reflect the primary-secondary, subordinate, etc. relationships between different semantic relationships. For example, the core semantic relationships such as the subject-predicate-object relationship are at a higher level, while the modifying relationship is relatively at a lower level.

[0073] Continuing with the above embodiment, in "The electricity bill of my home has suddenly become higher this month", from the perspective of the semantic feature vector direction, the vector direction from "my home" to "become higher" reflects the direction of the subject-predicate relationship, that is, "my home" is the doer of the action "become higher". The vector direction from "suddenly" to "become higher" reflects the direction of the modifying relationship, that is, "suddenly" modifies "become higher". From the perspective of the semantic relationship level, the subject-predicate relationship between "my home" and "become higher" is at a higher level because it constitutes the core structure of the sentence; while the modifying relationship between "suddenly" and "become higher" is at a lower level, which is a further description of the predicate verb in the core structure.

[0074] Step 205: Use the words in each consultation text segment as graph nodes, and construct the edges between the nodes based on the enhanced semantic relationships, semantic relationship directions, and semantic relationship levels between each pair of words in each consultation text segment, thereby obtaining a semantic relationship graph.

[0075] Furthermore, the power customer service device uses the words in each consultation text segment as graph nodes, and takes the enhanced semantic relationships, semantic relationship directions, and semantic relationship levels between each pair of words as the basis for constructing the edges between the nodes. Specifically, the type of the edge is determined according to the semantic relationship (such as the subject-predicate relationship edge, the modification relationship edge, etc.), the direction of the edge is determined according to the semantic relationship direction, and the weight or level of the edge can be determined according to the semantic relationship level. In this way, a semantic relationship graph that can intuitively display the semantic relationships between the words in the consultation text segment is constructed. Continuing with the above embodiment, for the text segment "The electricity bill of my home has suddenly become higher this month", "my home", "this month", "electricity bill", "suddenly", and "become higher" are each a graph node. There is a subject-predicate relationship edge between "my home" and "become higher", with the direction pointing from "my home" to "become higher", and this edge is at a higher level; there is a time modification relationship edge between "this month" and "become higher", with the direction pointing from "this month" to "become higher", at a lower level; there is a state modification relationship edge between "suddenly" and "become higher", with the direction pointing from "suddenly" to "become higher", also at a lower level. These edges and nodes together form a semantic relationship graph that clearly displays this text segment.

[0076] The embodiment of the present invention can transform the consultation text segment into a structured semantic relationship graph. The semantic relationship graph comprehensively and meticulously displays the semantic connections between the words in the text segment, including the relationship type, direction, and level, providing a solid data foundation for subsequently determining the target power business scenario associated with the consultation text segment. Therefore, the power customer service device can quickly and accurately match the corresponding power business scenario by analyzing the key nodes and relationships in the graph, greatly improving the accuracy and efficiency of understanding the consultation text, thereby being able to comprehensively and accurately handle customer consultations, significantly improving the efficiency and accuracy of power consultation services, and thus enhancing the user service experience.

[0077] In one embodiment, the descriptions of steps 301 to 304 are as follows:

[0078] Step 301: Combine the first graph nodes in the semantic relationship graph and the second graph nodes in the power business knowledge graph to obtain target graph node pairs.

[0079] Optionally, based on the above embodiments, the semantic relationship graph is constructed based on the consultation text fragments, showing the semantic connections between words in the text; the power business knowledge graph contains various concepts, attributes, and relationships in the power field. Therefore, the power customer service device traverses each first graph node in the semantic relationship graph and simultaneously traverses each second graph node in the power business knowledge graph. For each pair of nodes, the semantic similarity between the first graph node and the second graph node in the pair of nodes is calculated. Among them, the calculation of semantic similarity can be based on various methods, such as cosine similarity based on word vectors, etc. If the semantic similarity of a pair of nodes is greater than the preset similarity threshold, then this pair of nodes meets the preset conditions, and the first graph node and the second graph node in the pair of nodes are combined into a target graph node pair, where the preset similarity threshold is set according to the actual situation.

[0080] In one embodiment, the preset similarity threshold is 0.7. There is a node "electricity bill" in the semantic relationship graph, which is connected to other words through various semantic relationships in the semantic relationship graph, such as "become higher", "reason", etc. There is a node "electricity bill accounting" in the power business knowledge graph. The power customer service device uses a word vector model to calculate the semantic similarity between "electricity bill" and "electricity bill accounting". Assuming that through cosine similarity calculation, the similarity value is 0.8. Since 0.8 > 0.7, "electricity bill" (the first graph node) and "electricity bill accounting" (the second graph node) meet the preset conditions and form a target graph node pair.

[0081] Step 302, for each target graph node pair, perform path traversal based on the first adjacent node corresponding to the first graph node and the second adjacent node corresponding to the second graph node to obtain an initial association path.

[0082] Furthermore, for each target graph node pair, the power customer service device respectively obtains the first adjacent node corresponding to the first graph node and the second adjacent node corresponding to the second graph node, where the adjacent node refers to a node directly connected to the current node through a certain relationship.

[0083] Furthermore, the power customer service device performs path traversal on the first adjacent node and the second adjacent node. During the traversal process, only when the semantic similarity between the first adjacent node and the second adjacent node is also greater than the preset similarity threshold (that is, meets the preset conditions), these two adjacent nodes are included in the initial association path. By continuously finding the next-level adjacent nodes that meet the conditions from the current adjacent node, an initial association path starting from the first graph node, passing through a series of adjacent nodes that meet the conditions, and finally reaching the second graph node is gradually constructed. Therefore, it can be understood that the initial association path is composed of the first adjacent node and the second adjacent node that meet the preset conditions.

[0084] In one embodiment, for the pair of target graph nodes of "electricity bill" (the first graph node) and "electricity bill accounting" (the second graph node) obtained previously. In the semantic relationship graph, the first adjacent node of "electricity bill" is "increase", and in the power business knowledge graph, the second adjacent node of "electricity bill accounting" is "power consumption". The power customer service device calculates the semantic similarity between "increase" and "power consumption", and assumes that the similarity value obtained is 0.75 (greater than the preset similarity threshold of 0.7). Continuing from "increase" and "power consumption", the adjacent node of "increase" is "reason", and the adjacent node of "power consumption" is "abnormal increase", and the semantic similarity between "reason" and "abnormal increase" is calculated, assumed to be 0.8. In this way, through continuous traversal of adjacent nodes that meet the conditions, an initial association path is obtained: electricity bill - increase - reason (semantic relationship graph) and electricity bill accounting - power consumption - abnormal increase (power business knowledge graph).

[0085] Step 303: Determine the target association path based on the semantic consistency between the first semantic relationship of the initial association path in the semantic relationship graph and the second semantic relationship in the power business knowledge graph.

[0086] Further, for each initial association path, the power customer service device compares the first semantic relationship in the semantic relationship graph with the second semantic relationship in the power business knowledge graph, where the semantic relationships include causal relationships, parallel relationships, subordinate relationships, etc. If the two semantic relationships are semantically consistent, then this initial association path may become the target association path. For example, in the semantic relationship graph, "electricity bill" and "increase" are subject-predicate relationships, and in the power business knowledge graph, there is also a causal relationship between "electricity bill accounting" and "increase in power consumption", that is, the increase in power consumption affects the electricity bill accounting. Here, the semantic relationships are semantically relevant and consistent. The device will perform such a consistency judgment on each group of corresponding semantic relationships on the entire initial association path. Only when all corresponding semantic relationships meet the consistency requirements, the initial association path is determined as the target association path.

[0087] In one embodiment, for the initial association path obtained previously: electricity bill - increase - reason (semantic relationship graph) and electricity bill accounting - power consumption - abnormal increase (power business knowledge graph). In the semantic relationship graph, "electricity bill" and "increase" are subject-predicate relationships, and "increase" and "reason" are causal relationships; in the power business knowledge graph, "electricity bill accounting" and "power consumption" are a kind of accounting basis relationship, "power consumption" and "abnormal increase" are state description relationships, and there is a causal relationship between "abnormal increase in power consumption" and "change in electricity bill accounting". Overall, the semantic relationships in the semantic relationship graph and the power business knowledge graph are semantically consistent, so this initial association path is determined as the target association path.

[0088] Step 304: Determine the target power business scenario for each consultation text segment based on the positions of each node in the target association path in the power business knowledge graph.

[0089] Furthermore, the power customer service device checks the positions of each node in the target association path in the power business knowledge graph. Usually, the nodes in the power business knowledge graph correspond to different power business concepts, processes, or scenarios. Therefore, by analyzing the positions of the nodes in the target association path and their relationships, the target power business scenario associated with the consultation text segment can be determined. For example, if the key nodes in the target association path are concentrated under the branch of "abnormal electricity bill handling" in the power business knowledge graph, then the target power business scenario can be determined as the abnormal electricity bill handling scenario, as described in detail in Steps 3041 to 3044.

[0090] The embodiment of the present invention can deeply associate the semantic relationship graph constructed based on the consultation text with the power business knowledge graph. Therefore, it can accurately determine the corresponding target power business scenario for each consultation text segment, without simply relying on keyword matching on the surface of the text. Therefore, when faced with a large number of complex and diverse power consultation texts, it can quickly and accurately classify different consultations into the corresponding business scenarios, laying a solid foundation for providing targeted responses subsequently, greatly improving the efficiency and accuracy of power customer service in handling consultations, thereby being able to comprehensively and accurately handle customer consultations, significantly improving the efficiency and accuracy of power consultation services, and thus enhancing the user service experience.

[0091] In one embodiment, the descriptions of Steps 3041 to 3044 are as follows:

[0092] Step 3041: Obtain the first target node in the target association path.

[0093] Optionally, the power customer service device starts looking for nodes that meet specific conditions in the target association path. The power business knowledge graph has a certain structure, and there are central nodes in the graph. These central nodes usually represent core power business concepts or processes. Therefore, it will judge the positions of each node in the target association path in the power business knowledge graph to check whether it is within the preset range of the central nodes of the graph. The preset range can be defined based on the topological structure of the graph. For example, it is the node range that can be reached starting from the central node after a certain number of hops (such as 2 - 3 hops). Further, the nodes in the target association path that meet this condition are determined as the first target nodes by the power customer service device.

[0094] In one embodiment, the central node of the power service knowledge graph is "power service", with a preset range of 2 hops based on it. Among the previously determined target association paths, there is a node "electricity bill calculation". Starting from "power service" and passing through the intermediate node "power cost management", it is possible to reach "electricity bill calculation" within 2 hops. Therefore, "electricity bill calculation" is determined as the first target node by the power customer service device.

[0095] Step 3042: Determine the initial power service scenarios that have relationship links with the first target node in the power service knowledge graph.

[0096] Furthermore, the power customer service device searches in the power service knowledge graph for nodes that have relationship links with the first target node. These relationship links can be of various types, such as causal relationships, subordinate relationships, upstream and downstream process relationships, etc. Therefore, it can be understood that the nodes connected to the first target node through these relationship links often represent specific power service scenarios. Thus, the power customer service device determines these service scenarios as the initial power service scenarios.

[0097] Continuing with the above embodiment, for the previously determined first target node "electricity bill calculation", in the power service knowledge graph, the nodes that have relationship links with "electricity bill calculation" are "abnormal electricity bill handling" (because the electricity bill calculation result may be abnormal, there is a causal relationship) and "electricity bill generation" (electricity bill calculation is the pre - process of electricity bill generation, there is an upstream and downstream relationship). The power customer service device determines "abnormal electricity bill handling" and "electricity bill generation" as the initial power service scenarios.

[0098] Step 3043: Rank the initial power service scenarios based on the semantic fit degree between the first semantic feature vector of the consultation text segment and the second semantic feature vector of each initial power service scenario to obtain the ranked power service scenarios.

[0099] Furthermore, in order to more accurately determine the target service scenarios from the initial power service scenarios, an in - depth analysis of each initial power service scenario is carried out. The power customer service device extracts the first semantic feature vector of the consultation text segment, which can be obtained through natural language processing techniques such as word vector models, text embedding models, etc., and it can comprehensively reflect the semantic information of the consultation text segment. At the same time, the power customer service device extracts the second semantic feature vector of each initial power service scenario, and this vector is obtained based on the processing of information such as text descriptions and knowledge definitions related to the service scenario, representing the semantic characteristics of the service scenario.

[0100] Further, the power customer service device calculates the semantic fitness degree between the first semantic feature vector of the consultation text segment and the second semantic feature vector of each initial power business scenario. Among them, the semantic fitness degree can be calculated by various methods, such as cosine similarity, Euclidean distance, etc. According to the calculated semantic fitness degree, the power customer service device sorts the initial power business scenarios, and the higher the semantic fitness degree of the initial power business scenario, the higher the priority in the sorting.

[0101] Continuing the above embodiment, for the consultation text segment "The electricity bill of my family has suddenly become higher this month. What's the reason?", the power customer service device obtains its first semantic feature vector through the natural language processing model. For the initial power business scenario "Electricity bill anomaly handling", based on the knowledge description of this business scenario (such as including content such as analysis of reasons for electricity bill anomalies and anomaly handling processes), its second semantic feature vector is obtained through text processing. Through cosine similarity calculation, the semantic fitness degree between the two is 0.85. For another initial power business scenario "Electricity bill generation", after the same processing, the semantic fitness degree with the consultation text segment is 0.4. Sorting according to the semantic fitness degree from high to low, the sorting is: "Electricity bill anomaly handling" first, and "Electricity bill generation" second.

[0102] Step 3044, determine the power business scenario with the highest priority in the sorted power business scenarios as the target power business scenario for each consultation text segment.

[0103] Further, the power customer service device determines the power business scenario with the highest priority in the sorted power business scenarios as the target power business scenario for each consultation text segment. Among them, this business scenario with the highest priority, because it is the most semantically compatible with the consultation text segment, is most likely to be the actual business scenario involved in the user's consultation. In this way, the corresponding power business scenario can be accurately matched for the consultation text segment, preparing for providing targeted responses in the future.

[0104] In an embodiment, based on the previous sorting result, "Electricity bill anomaly handling" has the highest priority in the sorted power business scenarios. Therefore, the power customer service device determines the target power business scenario of the consultation text segment "The electricity bill of my family has suddenly become higher this month. What's the reason?" as the "Electricity bill anomaly handling" scenario.

[0105] The embodiments of the present invention can start from the target association path, and through multiple steps of screening and analysis, accurately determine the target power service scenario corresponding to each consultation text segment, avoiding the limitations of simply matching service scenarios based on surface keywords or simple relationships. Combining with the semantic features of the consultation text, it realizes a more intelligent and accurate business scenario matching, greatly improving the accuracy of power customer service in handling consultations, enabling the customer service to quickly locate the business area to which the user's problem belongs, so as to comprehensively and accurately handle customer consultations, significantly improving the efficiency and accuracy of power consultation services, and thus improving the user service experience.

[0106] In one embodiment, the descriptions of steps 401 to 404 are as follows:

[0107] Step 401: Classify the logical connections between each consultation text segment to obtain the logical connection categories between each consultation text segment, and construct a logical connection graph with each consultation text segment as a node and the logical connection categories between each consultation text segment as node edges.

[0108] Optionally, the power customer service device analyzes multiple consultation text segments, identifies the logical connections existing between each consultation text segment, and classifies each identified logical connection into the corresponding logical connection category. Among them, the logical connections may include causal relationships, parallel relationships, progressive relationships, turning relationships, etc. For example, if one consultation text segment describes a power outage event and another segment describes the electrical appliance damage caused by the power outage, then there is a causal relationship between these two segments.

[0109] Furthermore, the power customer service device constructs a logical connection graph with each consultation text segment as a node and the logical connection categories between each consultation text segment as node edges.

[0110] In one embodiment, the power customer service device receives three consultation text segments: Segment A "Why has my electricity bill suddenly become higher this month?", Segment B "There have been frequent power outages in the community recently, bringing a lot of inconvenience to life", and Segment C "The food in my refrigerator has gone bad after the power outage". Analysis shows that there is a causal relationship between Segment B and Segment C because the power outage in the community caused the food in the refrigerator to be damaged; Segment A has no direct logical relationship with Segment B and C and belongs to a parallel relationship. Therefore, in the constructed logical connection graph, Segments A, B, and C are nodes, the edge between Segment B and Segment C is marked as "causal relationship", and the edges between Segment A and B, and C are marked as "parallel relationship".

[0111] Step 402: For each second target node in the logical connection graph, traverse the paths in the logical connection graph based on each node edge to determine the number of paths starting from or / and ending with the second target node.

[0112] Further, for each node (i.e., the second target node) in the logical connection diagram, the power customer service device performs path traversal in the logical connection diagram according to each node edge, where the path traversal is to find all possible paths from one node to other nodes along the node edges in the diagram.

[0113] Further, for each second target node, the power customer service device determines the number of paths starting from or / and ending at it. The number of paths reflects the importance and connection tightness of the node in the entire logical connection diagram. If a node has more paths connected to it, it indicates that it plays a more crucial role in the logical connection and the logical association with other segments is more complex and diverse.

[0114] In an embodiment, for segment B (“Recently, the community often has power outages, which brings a lot of inconvenience to life”) in the above logical connection diagram, starting from segment B, there is one path reaching segment C through the “causal relationship” edge; ending at segment B, there is no path (assuming reverse paths are not considered in the current analysis). Therefore, the number of paths of segment B is 1. For segment C (“The food in my refrigerator has gone bad after the power outage”), ending at segment C, there is one path reaching from segment B through the “causal relationship” edge; starting from segment C, there is no path. Therefore, the number of paths of segment C is also 1. For segment A (“The electricity bill of my family has suddenly become very high this month. What's the reason?”), starting from segment A, there are two paths reaching segment B and segment C respectively through the “parallel relationship” edge; ending at segment A, there are also two paths. Therefore, the number of paths of segment A is 4.

[0115] Step 403, determine the consultation text segments with the number of paths greater than the preset number as key text segments.

[0116] Further, setting a preset number as the judgment criterion, the power customer service device determines the consultation text segments with the number of paths greater than the preset number as key text segments. Among them, the key text segments are in an important position in the logical structure of the entire consultation content. Their logical connections with other segments are more abundant and tight, and they play a key role in understanding the overall consultation intention of the user.

[0117] Continuing the above embodiment, the preset number is 2. According to the previously calculated number of paths, the number of paths of segment A is 4, which is greater than the preset number 2; the number of paths of segment B and segment C is 1, which is less than the preset number 2. Therefore, the power customer service device determines segment A as the key text segment.

[0118] Step 404, based on the key text segments and combining the text semantic themes of each consultation text segment, perform cross-segment semantic integration on each consultation text segment to obtain the cross-segment semantic integration text.

[0119] Further, the power customer service device takes the determined key text segment as the core, and combines the text semantic themes of each consultation text segment for cross-segment semantic integration. Specifically: The power customer service device analyzes the semantic themes of each consultation text segment. For example, the semantic theme of segment A revolves around abnormal electricity bills, the semantic theme of segment B revolves around power outages in the community, and the semantic theme of segment C revolves around food damage in the refrigerator caused by power outages. Further, guided by the semantics of the key text segment, the power customer service device integrates the semantics of other relevant segments according to logical connections. During the integration process, appropriate language organization and adjustment will be performed on the text to form a coherent and complete cross-segment semantic integration text, so as to clearly present the overall consultation intention of the user, as specifically described in steps 4041 to 4043.

[0120] In the embodiment of the present invention, by constructing a logical connection diagram, analyzing the number of analysis paths to determine the key text segment, and then integrating it with the semantic theme, simple text splicing is avoided, enabling accurate identification of multiple complex consultation intentions in the user's consultation text, so as to comprehensively and accurately process customer consultations, significantly improving the efficiency and accuracy of power consultation services, and thus improving the user service experience.

[0121] In one embodiment, the descriptions of steps 4041 to 4043 are as follows:

[0122] Step 4041, for any two target consultation text segments, determine the semantic theme correlation degree of the text semantic themes of the target consultation text segments under the logical connection between the target consultation text segments.

[0123] Optionally, when performing cross-segment semantic integration, the power customer service device analyzes any two target consultation text segments. For each consultation text segment, natural language processing technology and semantic analysis models are used to determine its text semantic theme. For example, through methods such as keyword extraction and LDA topic model, clarify the power-related themes mainly centered around each segment, such as electricity bill problems, power outage faults, equipment maintenance, etc.

[0124] Further, based on the determined logical connection (such as causal relationship, parallel relationship, progressive relationship, etc.) between the two target consultation text segments, the power customer service device calculates the semantic theme correlation degree between their text semantic themes. Among them, the calculation of the semantic theme correlation degree can be based on the semantic similarity algorithm and weighted adjustment in combination with the characteristics of the logical connection. For example, for two segments with a causal relationship, if the theme of one segment is "power outage" and the other is "electrical appliance damage", when calculating the correlation degree, the causal relationship will be given a higher weight, making their correlation degree relatively high.

[0125] In one embodiment, there are two consultation text fragments, fragment D "Power outage caused by community transformer failure" and fragment E "Data loss of my home computer after the power outage". First, determine that the text semantic theme of fragment D is "Power outage cause (transformer failure)", and the text semantic theme of fragment E is "Power outage consequence (computer data loss)". Since there is a causal logical connection between these two fragments, use a semantic similarity algorithm (such as cosine similarity) to calculate the similarity of the two themes, and give a weight according to the causal relationship. For example, the unweighted cosine similarity is 0.6. After considering the causal relationship weight (set to 1.5), the semantic theme correlation degree is 0.6 * 1.5 = 0.9.

[0126] Step 4042, based on the position of the key text fragment in the logical connection diagram and the semantic theme correlation degree between the key text fragment and other consultation text fragments, construct a cross-fragment semantic integration framework; the other consultation text fragments are the consultation text fragments other than the key text fragment.

[0127] Furthermore, the power customer service device determines the position of the key text fragment in the logical connection diagram. Among them, the key text fragment often occupies a core position or connects multiple important paths in the logical connection diagram, playing a key guiding role in the overall semantic integration. Therefore, the power customer service device analyzes the semantic theme correlation degree between the key text fragment and other consultation text fragments (that is, the fragments other than the key text fragment).

[0128] Furthermore, the power customer service device constructs a cross-fragment semantic integration framework according to the semantic theme correlation degree between the key text fragment and other consultation text fragments. Among them, the cross-fragment semantic integration framework takes the key text fragment as the center, and determines the connection method and hierarchical structure between other text fragments and the key text fragment according to the semantic theme correlation degree and logical connection. For example, if a non-key text fragment has a high semantic theme correlation degree and a close logical connection with the key text fragment, then in the framework it will be closer to the key text fragment and be connected by the corresponding logical relationship edge. The cross-fragment semantic integration framework can be represented by a tree structure or a graph structure, clearly showing the semantic and logical relationships between each text fragment.

[0129] In one embodiment, the key text segment F is "The electricity bill of my family has increased abnormally this month", and the other consultation text segments include segment G "New high-power electrical appliances have been added this month" and segment H "Neighbors have also reported that the electricity bill has become higher". The key text segment F is in the central position in the logical connection diagram because it is the core problem of the user's consultation. The semantic theme correlation degree between segment F and segment G is calculated to be 0.8 (since the addition of new high-power electrical appliances may cause abnormal electricity bills, there is a causal relationship), and the semantic theme correlation degree between segment F and segment H is 0.6 (both are centered around the problem of the increased electricity bill, which is a parallel relationship). Therefore, in the reconstructed cross-segment semantic integration framework, segment F is in the center, segment G is closely connected to segment F through the "causal relationship" edge, segment H is connected to segment F through the "parallel relationship" edge, and segment G is structurally closer to segment F because its correlation degree is higher.

[0130] Step 4043, based on the cross-segment semantic integration framework, perform cross-segment semantic integration on each consultation text segment to obtain the cross-segment semantic integration text.

[0131] Furthermore, according to the constructed cross-segment semantic integration framework, the power customer service device starts to perform cross-segment semantic integration on each consultation text segment. Specifically: starting from the key text segment, in accordance with the connection methods and hierarchical structures determined in the framework, sequentially integrate the semantics of other relevant text segments. During the integration process, the power customer service device will organize and adjust the text to make the semantics of the integrated text coherent and logically smooth. For example, appropriate conjunctions such as "because... so...", "meanwhile", "moreover", etc. will be added according to the logical relationship in the expression. In this way, each independent consultation text segment is integrated into a complete and clear cross-segment semantic integration text that reflects the user's consultation intention.

[0132] Continuing with the above embodiment, based on the previously constructed cross-segment semantic integration framework, the power customer service device starts from the key text segment F "The electricity bill of my family has increased abnormally this month". Since segment G "New high-power electrical appliances have been added this month" is closely connected to F through a causal relationship, it is integrated into "The electricity bill of my family has increased abnormally this month because new high-power electrical appliances have been added this month". Then, segment H "Neighbors have also reported that the electricity bill has become higher" is added through a parallel relationship, and finally the cross-segment semantic integration text obtained is "The electricity bill of my family has increased abnormally this month because new high-power electrical appliances have been added this month, and at the same time neighbors have also reported that the electricity bill has become higher".

[0133] In the embodiments of the present invention, an integration framework is constructed by quantifying the semantic topic correlation degree, and then text integration is performed according to the framework, so that the cross - segment semantic integrated text after integration can highly restore the complex consultation intention of the user, enabling the overall picture of the user's consultation to be quickly and accurately understood, avoiding the understanding deviation caused by fragment dispersion, thus being able to comprehensively and accurately process customer consultations, significantly improving the efficiency and accuracy of power consultation services, and thereby enhancing the user service experience.

[0134] Furthermore, the power customer service semantic analysis device based on the AI large - model provided by the present invention will be described below. The power customer service semantic analysis device based on the AI large - model described below can be mutually referred to corresponding to the power customer service semantic analysis method described above.

[0135] Optionally, referring to Figure 2 , Figure 2 is the structural schematic diagram of the power customer service semantic analysis device based on the AI large - model provided by the present invention. The power customer service semantic analysis device based on the AI large - model includes.

[0136] A text slicing module 210, configured to slice the received power consultation text based on natural language to obtain a plurality of consultation text fragments;

[0137] A semantic relationship parsing module 220, configured to perform semantic relationship parsing on each consultation text fragment based on the semantic understanding ability of the AI large - model, and construct a semantic relationship graph according to the semantic relationship between words in each consultation text fragment;

[0138] A business scenario association module 230, configured to determine the target power business scenario associated with each consultation text fragment based on the semantic relationship graph and the power business knowledge graph;

[0139] A text integration module 240, configured to perform cross - segment semantic integration on each consultation text fragment based on the logical connection between each consultation text fragment to obtain a cross - segment semantic integrated text;

[0140] An intention recognition and output module 250, configured to output a reply text based on the cross - segment semantic integrated text and the target power business scenario.

[0141] In the embodiments of the present invention, the powerful semantic understanding ability of the AI large model can adapt to various expression methods. As long as the semantics are the same or similar, the intention can be accurately understood. Further, through text segmentation, semantic relationship parsing, and cross-fragment semantic integration, the complex semantics in the power consultation text can be deeply understood. At the same time, through associated business scenario recognition and cross-fragment semantic integration, the relationships between different business scenarios can be clearly sorted out, so that multiple complex consultation intentions in the user consultation text can be accurately identified, and thus customer consultations can be comprehensively and accurately processed, significantly improving the efficiency and accuracy of power consultation services, and thus improving the user service experience.

[0142] Please refer to Figure 3 , Figure 3 which is the embodiment diagram of the electronic device provided by the embodiments of the present invention. As Figure 3 shown, the embodiments of the present invention provide an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0143] Slice the received power consultation text based on natural language to obtain multiple consultation text fragments;

[0144] Parse the semantic relationships of each consultation text fragment based on the semantic understanding ability of the AI large model, and construct a semantic relationship graph according to the semantic relationships between the words in each consultation text fragment;

[0145] Determine the target power business scenario associated with each consultation text fragment based on the semantic relationship graph and the power business knowledge graph;

[0146] Integrate the semantics of each consultation text fragment across fragments based on the logical connection between each consultation text fragment to obtain a cross-fragment semantic integration text;

[0147] Output a reply text based on the cross-fragment semantic integration text and the target power business scenario.

[0148] Please refer to Figure 4 , Figure 4 which is the embodiment diagram of the computer-readable storage medium provided by the embodiments of the present invention. As Figure 4 shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0149] Slice the received power consultation text based on natural language to obtain multiple consultation text fragments;

[0150] Perform semantic relationship parsing on each consultation text segment based on the semantic understanding ability of the AI large model, and construct a semantic relationship graph according to the semantic relationships between words in each consultation text segment;

[0151] Determine the target power business scenario associated with each consultation text segment based on the semantic relationship graph and the power business knowledge graph;

[0152] Perform cross-segment semantic integration on each consultation text segment based on the logical connection between each consultation text segment to obtain a cross-segment semantic integration text;

[0153] Output a reply text based on the cross-segment semantic integration text and the target power business scenario.

[0154] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the power customer service semantic analysis method based on the AI large model provided by the above methods. The method includes:

[0155] Slice the received power consultation text based on natural language to obtain multiple consultation text segments;

[0156] Perform semantic relationship parsing on each consultation text segment based on the semantic understanding ability of the AI large model, and construct a semantic relationship graph according to the semantic relationships between words in each consultation text segment;

[0157] Determine the target power business scenario associated with each consultation text segment based on the semantic relationship graph and the power business knowledge graph;

[0158] Perform cross-segment semantic integration on each consultation text segment based on the logical connection between each consultation text segment to obtain a cross-segment semantic integration text;

[0159] Output a reply text based on the cross-segment semantic integration text and the target power business scenario.

[0160] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0161] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A semantic analysis method for power customer service based on the AI large model, characterized in that, Including: Slice the received power consultation text based on natural language to obtain multiple consultation text fragments; Based on the semantic understanding ability of the AI large model, parse the semantic relationships of each consultation text fragment, and construct a semantic relationship graph according to the semantic relationships between the words in each consultation text fragment; Based on the semantic relationship graph and the power business knowledge graph, determine the target power business scenario associated with each consultation text fragment; Based on the logical connection between each consultation text fragment, perform cross-fragment semantic integration on each consultation text fragment to obtain a cross-fragment semantic integration text; Based on the cross-fragment semantic integration text and the target power business scenario, output a reply text.

2. The semantic analysis method of power customer service based on the AI large model according to claim 1, characterized in that, The determining the target power business scenario associated with each consultation text fragment based on the semantic relationship graph and the power business knowledge graph includes: Combine the first graph nodes in the semantic relationship graph and the second graph nodes in the power business knowledge graph to obtain a target graph node pair; the first graph node and the second graph node meet a preset condition; the preset condition is that the semantic similarity between the nodes is greater than a preset similarity threshold; For each target graph node pair, perform path traversal based on the first adjacent node corresponding to the first graph node and the second adjacent node corresponding to the second graph node to obtain an initial association path; the initial association path is composed of the first adjacent node and the second adjacent node that meet the preset condition; Based on the semantic consistency between the first semantic relationship of the initial association path in the semantic relationship graph and its second semantic relationship in the power business knowledge graph, determine the target association path; Based on the positions of each node in the target association path in the power business knowledge graph, determine the target power business scenario of each consultation text fragment.

3. The method for semantic analysis of power customer service based on the AI large model according to claim 2, wherein, The determining the target power business scenario of each consultation text fragment based on the positions of each node in the target association path in the power business knowledge graph includes: Obtain the first target node in the target association path; the position of the first target node in the power business knowledge graph is within a preset range of the graph center node of the power business knowledge graph; Determine the initial power business scenario whose business scenario is related to the first target node in the power business knowledge graph; Sort the initial power business scenarios based on the semantic fit between the first semantic feature vector of the consultation text fragment and the second semantic feature vector of each initial power business scenario to obtain the sorted power business scenarios; Determine the power business scenario with the highest priority in the sorted power business scenarios as the target power business scenario of each consultation text fragment.

4. The method for semantic analysis of power customer service based on the AI large model according to claim 1, wherein, The performing cross-fragment semantic integration on each consultation text fragment based on the logical connection between each consultation text fragment to obtain a cross-fragment semantic integration text includes: Classify the logical connections between each consultation text segment to obtain the logical connection categories between each consultation text segment. Then, construct a logical connection graph with each consultation text segment as a node and the logical connection categories between each consultation text segment as the node edges. For each second target node in the logical connection graph, traverse the paths in the logical connection graph based on each node edge to determine the number of paths starting from and / or ending at the second target node. Determine the key text segments as those with a path number greater than a preset number. Based on the key text segments and combined with the text semantic themes of each consultation text segment, perform cross-segment semantic integration on each consultation text segment to obtain the cross-segment semantic integration text.

5. The method for semantic analysis of power customer service based on the AI large model according to claim 4, wherein The step of performing cross-segment semantic integration on each consultation text segment based on the key text segments and combined with the text semantic themes of each consultation text segment to obtain the cross-segment semantic integration text includes: For any two target consultation text segments, determine the semantic theme correlation degree of the text semantic themes of the target consultation text segments under the logical connection between the target consultation text segments. Based on the position of the key text segments in the logical connection graph and the semantic theme correlation degree between the key text segments and other consultation text segments (where the other consultation text segments are those other than the key text segments), construct a cross-segment semantic integration framework. Based on the cross-segment semantic integration framework, perform cross-segment semantic integration on each consultation text segment to obtain the cross-segment semantic integration text.

6. The method for semantic analysis of power customer service based on the AI large model according to claim 1, wherein The step of parsing the semantic relationships of each consultation text segment based on the semantic understanding ability of the AI large model and constructing a semantic relationship graph according to the semantic relationships between the words in each consultation text segment includes: For each consultation text segment, perform grammatical parsing on the consultation text segment based on the AI large model to identify the words of various parts of speech in the consultation text segment. Perform semantic relationship parsing on the consultation text segment based on the third semantic feature vector of each word to identify the preliminary semantic relationships between each word in the consultation text segment. The third semantic feature vector of each word represents the position of each word in the semantic space corresponding to the consultation text segment and is determined based on the contribution degree and semantic feature value of each semantic feature in the word. Adjust the preliminary semantic relationships based on the context information of each word to obtain the strengthened semantic relationships between each word in the consultation text segment. Based on the AI large model, determine the semantic relationship direction between each word according to the vector direction of the third semantic feature vector between each word, and determine the semantic relationship level of the strengthened semantic relationship between each word according to the semantic relationship direction and the strengthened semantic relationship between each word. Use the words in each consultation text segment as graph nodes and construct the edges between the nodes with the strengthened semantic relationships, semantic relationship directions, and semantic relationship levels between each word in each consultation text segment to obtain the semantic relationship graph.

7. The method for semantic analysis of power customer service based on the AI large model according to any one of claims 1 to 6, characterized in that Slice the received power consultation text based on natural language to obtain multiple consultation text segments, including: Split the power consultation text according to punctuation marks to obtain initial text segments; Merge adjacent text segments in the initial text segments whose semantic coherence degree is greater than or equal to a preset coherence degree threshold to obtain merged text segments; the semantic coherence degree is determined according to the text theme and text context between adjacent text segments; Based on a preset power keyword dictionary, identify the power keywords of each merged text segment, and perform topic clustering according to the power keywords and semantic information of each merged text segment to obtain the topic clustering result of each merged text segment; Based on a preset correlation word dictionary, identify the correlation words associated with each merged text segment in the power consultation text, and divide the power consultation text into structural levels based on the correlation words and topic clustering results of each merged text segment to obtain the text structural level of each merged text segment; Divide the merged text segments with the same topic clustering result and text structural level into the same slice to obtain multiple said consultation text segments.

8. An AI large model-based semantic analysis device for power customer service, characterized in that, Applied to the power customer service semantic analysis method based on the AI large model as described in any one of claims 1 to 7; the power customer service semantic analysis device based on the AI large model includes: A text slicing module for slicing the received power consultation text based on natural language to obtain multiple consultation text segments; A semantic relationship parsing module for parsing the semantic relationship of each consultation text segment based on the semantic understanding ability of the AI large model, and constructing a semantic relationship graph according to the semantic relationship between words in each consultation text segment; A business scenario association module for determining the target power business scenario associated with each consultation text segment based on the semantic relationship graph and the power business knowledge graph; A text integration module for performing cross-segment semantic integration on each consultation text segment based on the logical connection between each consultation text segment to obtain a cross-segment semantic integration text; An intention recognition output module for outputting a reply text based on the cross-segment semantic integration text and the target power business scenario.

9. An electronic device, comprising: A memory for storing computer software programs; A processor for reading and executing the computer software program, characterized in that when the processor executes the computer software program, it implements the power customer service semantic analysis method based on the AI large model as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing a computer software program, characterized in that, When the computer software program is executed by the processor, it implements the power customer service semantic analysis method based on the AI large model as described in any one of claims 1 to 7.

Images