Bank seat document knowledge base-based verbal skill generation method

Abstract

The invention discloses a bank seat document knowledge base-based verbal skill generation method, which comprises the following steps of: monitoring a conversation voice stream to obtain the voice stream; generating a text according to the voice stream; performing preprocessing analysis on the text to obtain an analysis result; generating an early warning result according to the analysis result; generating a verbal skill according to the early warning result; and performing seat monitoring intervention according to the verbal skill. And through dynamic rule updating, it is ensured that the quality inspection system can continuously adapt to a new violation mode, and the comprehensiveness and timeliness of detection are improved.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition, and in particular to a method for generating speech based on a bank agent document knowledge base. Background Art

[0002] With the rapid development of artificial intelligence technology, customer service centers have also ushered in a qualitative leap. How to respond to customer needs quickly and accurately has also become a challenge. Bank agents are important roles in banking business. They are responsible for receiving customers, handling various banking services and maintaining customer relationships. This requires agents to have not only professional business knowledge, but also good communication skills and problem-solving abilities, which is difficult for some new employees.

[0003] The current technology mainly uses offline voice quality inspection tools to conduct comprehensive quality inspections on customer service calls. The call recordings are converted into text information through voice transcription technology, and then the text information is compared and analyzed using quality inspection analysis technology to avoid missing problems that occurred during the call and free up quality inspection personnel. However, the post-call quality inspection work can only sort out the problems that occurred and reduce the occurrence of such problems in the future. It cannot stop the loss of the problems that occurred in time, and there is no way to remedy the problems that are happening. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide a method for generating speech based on a bank agent document knowledge base, which overcomes the above problems or at least partially solves the above problems.

[0005] According to one aspect of the present invention, a method for generating speech based on a bank agent document knowledge base is provided, the generating method comprising:

[0006] Monitor the call voice stream and obtain the voice stream;

[0007] generating text according to the speech stream;

[0008] Performing preprocessing analysis on the text to obtain analysis results;

[0009] Generate a warning result according to the analysis result;

[0010] Generate a speech based on the warning result;

[0011] Conduct agent monitoring and intervention based on the script.

[0012] Optionally, monitoring the call voice stream and obtaining the voice stream specifically includes:

[0013] Through recording equipment or API interface, the conversation content between bank customer service and customers can be collected in real time;

[0014] It also supports recognition processing of audio files manually uploaded by agents.

[0015] Optionally, generating text according to the voice stream specifically includes: utilizing the voice recognition capability of the model platform to convert the collected voice data into text data.

[0016] Optionally, the preprocessing analysis of the text to obtain the analysis result specifically includes:

[0017] Use the text analysis capabilities of the model platform to perform preprocessing operations such as cleaning, word segmentation, and part-of-speech tagging on the converted text data;

[0018] In the preprocessing stage, new rules are automatically extracted and generated and updated to the preset rule set to continuously improve the quality inspection rules. It also supports manual updates of the preset rule set by agents.

[0019] Optionally, generating a warning result according to the analysis result specifically includes:

[0020] Use the analysis model of the model platform and integrate the rule set to generate quality inspection warning result information;

[0021] And notify the seat staff in real time to provide sufficient background information for the next monitoring intervention.

[0022] Optionally, generating speech based on the warning result specifically includes: utilizing the speech model of the model platform and integrating key information of the knowledge base document to generate speech information corresponding to the warning result.

[0023] Optionally, performing agent monitoring intervention according to the speech technique specifically includes:

[0024] Agents spot-check warning information and automatically generated scripts in real time and intervene in any abnormal information.

[0025] The present invention provides a method for generating speech based on a bank agent document knowledge base, the method comprising: monitoring a call voice stream to obtain a voice stream; generating text according to the voice stream; preprocessing and analyzing the text to obtain an analysis result; generating an early warning result according to the analysis result; generating speech according to the early warning result; and performing agent monitoring intervention according to the speech. By dynamically updating rules, it is ensured that the quality inspection system can continuously adapt to new violation patterns and improve the comprehensiveness and timeliness of detection.

[0026] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0028] Figure 1 A flowchart of a method for generating speech based on a bank agent document knowledge base is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0030] The terms "comprises" and "having" and any variations thereof in the description embodiments, claims and drawings of the present invention are intended to cover non-exclusive inclusions, for example, including a series of steps or units.

[0031] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0032] like Figure 1 As shown, a method for generating speech based on a bank agent document knowledge base includes:

[0033] Step 1: Monitor the call audio stream: Use recording equipment or API interface to collect the conversation content between bank customer service and customers in real time. It also supports the recognition and processing of audio files manually uploaded by agents.

[0034] Step 2: Generate text: Use the speech recognition capabilities of the model platform to convert the collected speech data into text data.

[0035] Step 3: Preprocessing and analyzing text: Use the text analysis capabilities of the model platform to perform preprocessing operations such as cleaning, word segmentation, and part-of-speech tagging on the converted text data to improve the accuracy of subsequent analysis. At the same time, some new rules will be automatically extracted and generated in the preprocessing stage and updated to the preset rule set, so that the quality inspection rules are continuously improved. At the same time, it also supports the manual update of the preset rule set by the seat staff.

[0036] Step 4: Generate early warning results: Use the analysis model of the model platform and integrate the rule set to generate quality inspection early warning result information. Notify the seat staff in real time to provide sufficient background information for the next step of monitoring intervention.

[0037] Step 5: Generate speech: Use the speech model of the model platform and integrate the key information of the knowledge base document to generate speech information corresponding to the warning results.

[0038] Step 6: Agent monitoring and intervention: Agents spot-check warning information and automatically generated scripts in real time, and intervene in any abnormal information.

[0039] The specific detailed process includes:

[0040] 1. Voice stream monitoring: The voice stream monitoring model is used to capture, analyze, and monitor continuous voice data streams in real time. This model is often used for voice calls, meeting records, or any scenario where real-time monitoring of voice content is required. The main technologies include: Audio capture: Use a microphone or audio input device to capture raw audio signals. Preprocessing: Includes technologies such as noise suppression, echo cancellation, and audio enhancement to improve voice quality. Feature extraction: Convert audio signals into feature vectors suitable for machine learning model processing, such as MFCC (Mel frequency cepstral coefficients), spectrograms, etc. Real-time analysis: Use real-time data stream processing technologies such as Apache Kafka, Spark Streaming, etc. to perform real-time analysis on the extracted features. Anomaly detection: Detect abnormal behavior or content based on changes in voice features, such as emotional changes, keyword triggers, etc.

[0041] 2. Speech recognition model explanation: The speech recognition model converts human speech into text representation. This is the key technology to achieve speech-to-text conversion (STT). The main technologies include: Acoustic model: Learn how to convert audio signals into sequences of phonemes or subword units. Deep learning models such as DNN-HMM, RNN, LSTM, Transformer, etc. are usually used. Language model: Used to predict the sequence of words that may appear in a given context. This helps solve the problem of homophones and generate grammatically correct sentences. Decoder: Combine the phoneme or subword sequence output by the acoustic model with the language model to generate the final text output. Common decoding methods include Viterbi algorithm, beam search, etc.

[0042] 3. Explanation of the text processing model: The text processing model is used to perform cleaning, word segmentation, part-of-speech tagging, named entity recognition, sentiment analysis and other processing on text data. The main technologies include: Text cleaning: remove noise from the text, such as HTML tags, special characters, stop words, etc. Word segmentation: divide the text into words or tokens, and sentence segmentation is also required for languages ​​such as Chinese. Part-of-speech tagging: assign a part-of-speech tag to each word, such as noun, verb, etc. Named entity recognition: identify entities with specific meanings from the text, such as names of people, places, and institutions. Sentiment analysis: determine the emotional tendency of the text expression, such as positive, negative, or neutral.

[0043] 4. Rule analysis model explanation: The rule analysis model parses and judges text or voice data based on a predefined set of rules. These rules can be based on business logic, grammatical rules, or standards in a specific field. The main technologies include: Rule definition: clarify the specific content and conditions of the rules, such as regular expressions, conditional logic, etc. Rule engine: a software component that implements rule matching and execution, used to traverse input data and make judgments based on the defined rule set. Result output: Generate corresponding output based on the results of rule matching, such as marking anomalies, classification labels, etc.

[0044] 5. Explanation of the speech generation model: The speech generation model automatically generates natural language responses based on given inputs (such as user requests, conversation history, etc.). The main technologies include: Template matching: Generate responses based on predefined templates and variables. This method is simple and fast, but has low flexibility. Sequence-to-sequence model: Such as RNN, LSTM, Transformer, etc., used to convert input sequences into output sequences. This method can generate more natural and diverse responses. Context management: Track the conversation history to ensure that the generated responses are consistent and coherent with the previous conversation content. Post-processing: Perform post-processing such as grammar checking and spelling correction on the generated responses to improve the output quality.

[0045] Beneficial effects: Enhanced configuration flexibility: Through dynamic rule updates, ensure that the quality inspection system can continuously adapt to new violation patterns and improve the comprehensiveness and timeliness of detection.

[0046] Realize early warning and prevention functions: Introduce real-time voice analysis technology to realize immediate early warning and intervention of illegal behaviors, reduce the occurrence of customer service violations, and protect corporate interests and customer rights.

[0047] Improve quality and efficiency: The automated or semi-automated quality inspection model building process significantly reduces manual intervention and development costs, while speeding up model iteration and improving quality inspection work efficiency.

[0048] Improve service quality: Promote continuous improvement of service quality and enhance customer satisfaction and brand loyalty through accurate and efficient quality inspection methods.

[0049] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating speech based on a bank agent document knowledge base, characterized in that: The generation method comprises: Monitor the call voice stream and obtain the voice stream; generating text according to the speech stream; Performing preprocessing analysis on the text to obtain analysis results; Generate a warning result according to the analysis result; Generate a speech based on the warning result; Conduct agent monitoring and intervention based on the script.

2. A method for generating speech based on a bank agent document knowledge base according to claim 1, characterized in that: The monitoring of the call voice stream and obtaining the voice stream specifically includes: Through recording equipment or API interface, the conversation content between bank customer service and customers can be collected in real time; It also supports recognition processing of audio files manually uploaded by agents.

3. A method for generating speech based on a bank agent document knowledge base according to claim 1, characterized in that: The generating text according to the voice stream specifically includes: utilizing the voice recognition capability of the model platform to convert the collected voice data into text data.

4. A method for generating speech based on a bank agent document knowledge base according to claim 1, characterized in that: The preprocessing analysis of the text to obtain the analysis result specifically includes: Use the text analysis capabilities of the model platform to perform preprocessing operations such as cleaning, word segmentation, and part-of-speech tagging on the converted text data; In the preprocessing stage, new rules are automatically extracted and generated and updated to the preset rule set to continuously improve the quality inspection rules. It also supports manual updates of the preset rule set by agents.

5. The method for generating speech based on bank agent document knowledge base according to claim 1 is characterized in that: Generating a warning result according to the analysis result specifically includes: Use the analysis model of the model platform and integrate the rule set to generate quality inspection warning result information; And notify the seat staff in real time to provide sufficient background information for the next monitoring intervention.

6. A method for generating speech based on a bank agent document knowledge base according to claim 1, characterized in that: The generating of speech based on the warning result specifically includes: utilizing the speech model of the model platform and integrating key information of the knowledge base document to generate speech information corresponding to the warning result.

7. A method for generating speech based on a bank agent document knowledge base according to claim 1, characterized in that: The agent monitoring intervention according to the speech technique specifically includes: Agents spot-check warning information and automatically generated scripts in real time and intervene in any abnormal information.

Images