Abnormal behavior determination method of bid evaluation expert in remote bid evaluation video conference process

Abstract

The invention provides a method for affirming an abnormal behavior of a bid evaluation expert in a remote bid evaluation video conference process, and the method comprises the steps: obtaining audio data and video data of the bid evaluation expert in a process that the bid evaluation expert carries out the remote bid evaluation of a preset item through a video conference bid evaluation tool; performing language detection on the audio data to obtain a language detection result; the language detection result comprises whether the bid evaluation expert has improper speech or not; performing face detection on the video data to obtain a face detection result; the face detection result comprises whether the bid evaluation expert has a leaving behavior or not; and on the basis of the language detection result and the face detection result, identifying the abnormal behavior of the bid evaluation expert to obtain an abnormal behavior identification result. According to the invention, in the process of identifying the abnormal behaviors of the bid evaluation experts, various factors are comprehensively considered, and the efficiency and accuracy of identifying the abnormal behaviors are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a method for identifying abnormal behavior of bid evaluation experts in a remote bid evaluation video conference. BACKGROUND

[0002] With the rapid development of information technology and artificial intelligence, more and more industries and fields have begun to use remote bid evaluation to conduct expert evaluation. However, remote bid evaluation lacks effective behavior monitoring mechanisms, which can easily lead to non-standard behavior and affect the fairness and accuracy of bid evaluation. Currently, existing technologies only consider a single factor in the process of identifying abnormal behavior of remote bid evaluation experts, which has poor accuracy and usually requires manual identification, resulting in low efficiency and poor real-time performance. SUMMARY

[0003] The present application provides a method for identifying abnormal behavior of bid evaluation experts in a remote bid evaluation video conference, which solves the defects of existing technologies that only consider a single factor in the process of identifying abnormal behavior of remote bid evaluation experts, resulting in poor accuracy and low efficiency.

[0004] The present application provides a method for identifying abnormal behavior of bid evaluation experts in a remote bid evaluation video conference, which includes: During the process of the bid evaluation expert remotely evaluating a predetermined project through a video conference bid evaluation tool, audio data and video data of the bid evaluation expert are obtained; Language detection is performed on the audio data to obtain a language detection result; the language detection result includes whether the bid evaluation expert has made inappropriate remarks; Face detection is performed on the video data to obtain a face detection result; the face detection result includes whether the bid evaluation expert has left the post; Based on the language detection result and the face detection result, the abnormal behavior of the bid evaluation expert is identified to obtain an abnormal behavior identification result.

[0005] In some embodiments, the language detection of the audio data to obtain a language detection result includes: Voice recognition is performed on the audio data to obtain text data; The text data is input into a pre-constructed text classification model to obtain a language detection result output by the text classification model; The text classification model is trained based on sample text data corresponding to sample audio data and language detection result labels corresponding to the sample audio data.

[0006] In some embodiments, the face detection of the video data to obtain a face detection result includes: inputting the video data into a pre-constructed face detection model to obtain a face detection result output by the face detection model; The face detection model is trained based on sample video data and a face detection result label corresponding to the sample video data.

[0007] In some embodiments, the inputting the video data into a pre-constructed face detection model to obtain a face detection result output by the face detection model comprises: performing face detection on the video data based on the face detection model, recording a time length during which the face of the bid evaluation expert is not detected continuously, and determining that the bid evaluation expert has an off-duty behavior if the time length exceeds a preset time length threshold.

[0008] In some embodiments, the identifying the abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result comprises: determining that the bid evaluation expert has an abnormal behavior if the bid evaluation expert has inappropriate remarks and / or has an off-duty behavior; determining that the bid evaluation expert has no abnormal behavior if the bid evaluation expert has no inappropriate remarks and has no off-duty behavior.

[0009] In some embodiments, after obtaining the abnormal behavior identification result, the method further comprises: generating a warning information and giving an alarm if the bid evaluation expert has inappropriate remarks and / or has an off-duty behavior.

[0010] In some embodiments, the training process of the text classification model comprises: obtaining sample audio data of a sample bid evaluation expert in a remote bid evaluation process; performing speech recognition on the sample audio data to obtain sample text data; determining a language detection result label corresponding to the sample audio data; inputting the sample text data into a pre-constructed initial text classification model to obtain a language detection prediction result output by the initial text classification model; the language detection prediction result comprises whether the sample bid evaluation expert has inappropriate remarks; calculating a first loss function value based on the language detection prediction result and the language detection result label; iteratively optimizing parameters of the initial text classification model based on the first loss function value to obtain the text classification model.

[0011] In some embodiments, the training process of the face detection model comprises: obtaining sample video data of a bid evaluation expert in a remote bid evaluation process; determining a face detection result label corresponding to the sample video data; inputting the sample video data into a pre-constructed initial face detection model to obtain a face detection prediction result output by the initial face detection model; calculating a second loss function value based on the face detection prediction result and the face detection result label; iteratively optimizing parameters of the initial face detection model based on the second loss function value to obtain the face detection model.

[0012] The application also provides an abnormal behavior identification system for a bid evaluation expert in a remote bid evaluation video conference process, comprising: an acquisition unit configured to acquire audio data and video data of the bid evaluation expert during a remote bid evaluation process of a predetermined project by the bid evaluation expert through a video conference bid evaluation tool; an audio detection unit configured to perform language detection on the audio data to obtain a language detection result, wherein the language detection result comprises whether the bid evaluation expert has inappropriate remarks; a video detection unit configured to perform face detection on the video data to obtain a face detection result, wherein the face detection result comprises whether the bid evaluation expert has an off-duty behavior; an identification unit configured to identify an abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result to obtain an abnormal behavior identification result.

[0013] The application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the abnormal behavior identification method for a bid evaluation expert in a remote bid evaluation video conference process according to any one of the above embodiments when executing the program.

[0014] The abnormal behavior identification method for a bid evaluation expert in a remote bid evaluation video conference process provided by the application comprises the following steps: acquiring audio data and video data of the bid evaluation expert in a remote bid evaluation process; performing language detection on the audio data to obtain a language detection result; performing face detection on the video data to obtain a face detection result; identifying an abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result to obtain an abnormal behavior identification result, which comprehensively considers multiple factors and improves the efficiency and accuracy of abnormal behavior identification. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative effort based on these drawings.

[0016] Figure 1 is a flowchart of the method for identifying the abnormal behavior of the bid evaluation expert in the remote bid evaluation video conference process provided by the embodiments of the present application.

[0017] Figure 2 is a flowchart of the training process of the text classification model.

[0018] Figure 3 is a flowchart of the training process of the face detection model.

[0019] Figure 4 is a structural diagram of the system for identifying the abnormal behavior of the bid evaluation expert in the remote bid evaluation video conference process provided by the embodiments of the present application.

[0020] Figure 5 is a structural diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0021] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort belong to the protection scope of the present application.

[0022] The terms "first", "second" and the like in the present application are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second" are usually a class, not limited to the number of objects, for example, the first object can be one or more. In addition, "and / or" in the present application means at least one of the connected objects, and the character " / " generally means that the front and rear associated objects are in an "or" relationship.

[0023] Figure 1 is a flowchart of the method for identifying the abnormal behavior of the bid evaluation expert in the remote bid evaluation video conference process provided by the embodiments of the present application. As Figure 1As shown, a remote bid evaluation video conference process of bid evaluation expert abnormal behavior identification method is provided, including the following steps: step 110, step 120, step 130 and step 140. The method flow steps are only as one possible implementation of the present application.

[0024] Step 110, in the process of bid evaluation expert through video conference bid evaluation tool for remote bid evaluation of the preset project, the audio data and video data of the bid evaluation expert are obtained.

[0025] Optionally, in the bid evaluation process, the audio data and video data of the bid evaluation expert are collected in real time by installing audio and video monitoring equipment. The audio equipment can clearly record the speech of the bid evaluation expert, and the video equipment can capture the facial expression, body language and location of the bid evaluation expert in real time.

[0026] It should be noted that the collected audio data and video data are uploaded to the server through encryption technology to ensure that the data is not disturbed, leaked or tampered with during transmission. The uploaded audio data and video data are stored in the cloud server for subsequent analysis and processing.

[0027] Step 120, language detection is performed on the audio data to obtain a language detection result; the language detection result includes whether the bid evaluation expert has inappropriate remarks.

[0028] Among them, inappropriate remarks include but are not limited to: biased remarks, induced remarks, discriminatory remarks, leakage remarks, inappropriate evaluation remarks, etc.

[0029] Among them, biased remarks specifically refer to: the bid evaluation expert makes obvious biased evaluation or suggestion on a bidder or his product, service, etc. during the bid evaluation process, indicating his preference for a specific bid project; for example, "Do you agree with the feasibility of this project?" Among them, induced remarks specifically refer to: through suggestion, guidance and other ways, trying to influence the evaluation opinions of other bid evaluation experts, so that they evaluate according to the direction expected by themselves; discriminatory remarks specifically refer to: based on the region, scale, ownership nature of the bidder, making unfair evaluation or discrimination on the bidder, and making discriminatory remarks.

[0030] Among them, leakage remarks specifically refer to the bid evaluation expert revealing the identity of the bid evaluation committee members, the bid evaluation project, the evaluation situation, the recommendation situation of the bid winner candidate and other confidential information related to bid evaluation; inappropriate evaluation remarks specifically refer to: inappropriate evaluation on the evaluation opinions or ability of other bid evaluation experts.

[0031] Optionally, the audio data is preprocessed to obtain preprocessed audio data; the preprocessed audio data is subjected to speech recognition to obtain text data.

[0032] Optionally, the text data is subjected to semantic analysis to determine whether inappropriate remarks exist in the text data, to obtain a language detection result.

[0033] Optionally, the audio data is subjected to real-time analysis using artificial intelligence technologies such as deep learning and speech recognition to detect whether inappropriate remarks exist.

[0034] Step 130, performing face detection on the video data to obtain a face detection result; the face detection result includes whether the bid evaluation experts have off-duty behavior.

[0035] Optionally, the video data is preprocessed to obtain a plurality of frames of images, each frame of image containing time information.

[0036] The plurality of frames of images are subjected to frame-by-frame detection in chronological order to determine whether a face exists in each frame of image, and if so, the detected face is matched with pre-stored facial information of the bid evaluation experts to further determine whether a face of the bid evaluation experts exists in each frame of image.

[0037] Optionally, the plurality of frames of images are subjected to frame-by-frame detection in chronological order, and the duration of continuous non-detection of a face is calculated from the time when a face is first not detected.

[0038] Optionally, whether the bid evaluation experts have off-duty behavior is determined according to the duration of continuous non-detection of a face.

[0039] In some embodiments, the video data is subjected to body language detection to obtain a body language detection result, the body language detection result including whether the bid evaluation experts have non-standard interaction behavior.

[0040] The non-standard interaction behavior includes some body movements or behavior performances that do not conform to the bid evaluation process and standard requirements. For example, the bid evaluation experts privately whispering to each other, using body language to indicate intentions, and the like.

[0041] It should be noted that the non-standard interaction behavior has an adverse effect on the fairness and objectivity of the bid evaluation process, and through body language detection, these non-standard interaction situations can be discovered and recognized in a timely manner, so that subsequent corresponding measures can be taken for processing and correction.

[0042] Step 140, identifying the abnormal behavior of the bid evaluation experts based on the language detection result and the face detection result to obtain an abnormal behavior identification result.

[0043] Optionally, based on the language detection result and the face detection result, it is judged whether the bid evaluation expert has abnormal behavior, if yes, the type and risk level of the abnormal behavior are determined, and a warning is given according to the risk level.

[0044] Optionally, based on the language detection result, the face detection result and the body language detection result, it is judged whether the bid evaluation expert has abnormal behavior, and an abnormal behavior recognition result is obtained.

[0045] In the embodiment of the present application, by acquiring audio data and video data of the bid evaluation expert in the remote bid evaluation process, performing language detection on the audio data to obtain a language detection result, performing face detection on the video data to obtain a face detection result, and identifying the abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result to obtain an abnormal behavior recognition result, multiple factors are comprehensively considered, and the efficiency and accuracy of abnormal behavior identification are improved.

[0046] In some embodiments, the language detection on the audio data to obtain a language detection result comprises: performing speech recognition on the audio data to obtain text data; inputting the text data into a pre-constructed text classification model to obtain a language detection result output by the text classification model; The text classification model is obtained by training based on sample text data corresponding to sample audio data and language detection result labels corresponding to the sample audio data.

[0047] Optionally, the audio data is subjected to speech recognition to obtain original text data, and the original text data is subjected to word segmentation to obtain text data.

[0048] Optionally, the audio data is subjected to speech recognition based on an open-source speech recognition model (such as Open AI Whisper) to obtain text data.

[0049] Optionally, based on the text classification model, feature extraction is performed on the text data to obtain a text feature vector, and based on the text feature vector, it is determined whether there is inappropriate speech, if yes, the type of inappropriate speech is further determined, such as biased speech, inducing speech, discriminatory speech, leakage speech, and inappropriate evaluation speech.

[0050] Optionally, the text data is input into a pre-constructed text classification model in real time to obtain a language detection result output by the text classification model.

[0051] It can be understood that by performing speech recognition on the audio data to obtain text data, and inputting the text data into a pre-constructed text classification model to obtain a language detection result output by the text classification model, the efficiency and accuracy of language detection can be improved, and real-time detection can be realized.

[0052] In some embodiments, the video data is subjected to face detection to obtain a face detection result, including: The video data is input into a pre-constructed face detection model to obtain a face detection result output by the face detection model; The face detection model is trained based on sample video data and face detection result labels corresponding to the sample video data.

[0053] Optionally, based on the face detection model, feature extraction is performed on the video data to obtain a video feature vector, and based on the video feature vector, face detection is performed to obtain the face detection result.

[0054] Optionally, the face detection model includes a multi-task cascaded convolutional neural network; the face detection model has high precision and multi-scale detection capability, and is suitable for face recognition tasks in complex backgrounds.

[0055] It can be understood that by inputting the video data into the pre-constructed face detection model, the face detection result output by the face detection model is obtained, thereby improving the efficiency and real-time performance of face detection.

[0056] In some embodiments, the video data is input into a pre-constructed face detection model to obtain a face detection result output by the face detection model, including: Optionally, based on the face detection model, face detection is performed on the video data, and the duration of continuous non-detection of the face of the bid evaluation expert is recorded. In a case where the duration is determined to exceed a preset duration threshold, it is determined that the bid evaluation expert has an off-duty behavior.

[0057] Optionally, the duration threshold is determined based on user input or according to empirical data, for example, 1800 seconds (i.e., 30 minutes). If no face is detected for 30 consecutive minutes, it is determined as “suspected off-duty behavior”.

[0058] Optionally, the video data is read frame by frame based on an image processing tool (such as Open CV), and video frame rate information (i.e., how many frames are included per second) is extracted.

[0059] Optionally, based on the video frame rate information, the duration threshold is converted into a frame number threshold, so that the continuous on-duty time of the bid evaluation expert in front of the camera can be accurately calculated.

[0060] Optionally, based on the face detection model, the following judgment is performed on each frame of image: If no face is detected in the current frame, the “no face frame counter” is incremented by one; If a face is detected, the counter is immediately reset; If the number of consecutive frames without a face exceeds the preset threshold, the off-duty determination is triggered, and a prompt is output and the analysis is terminated.

[0061] It should be noted that if the duration of consecutive frames without a face exceeds the duration threshold, the specific time point (in seconds) and prompt information are output for prompting the review system or staff. If the off-duty determination condition is not met throughout the video, a clear indication of "no obvious off-duty behavior detected" is displayed at the end of processing.

[0062] It can be understood that by performing face detection on video data, the duration of consecutive frames without detecting the face of the bid evaluation expert is recorded, and in the case where the duration exceeds the preset duration threshold, it is determined that the bid evaluation expert has off-duty behavior, which improves the accuracy of off-duty behavior recognition and avoids false positives caused by temporary occlusion or face recognition errors.

[0063] In some embodiments, based on the language detection result and the face detection result, the abnormal behavior of the bid evaluation expert is identified, including: In the case where the bid evaluation expert has inappropriate remarks and / or the bid evaluation expert has off-duty behavior, it is determined that the bid evaluation expert has abnormal behavior; In the case where the bid evaluation expert does not have inappropriate remarks and the bid evaluation expert does not have off-duty behavior, it is determined that the bid evaluation expert has no abnormal behavior.

[0064] Optionally, if the face of the bid evaluation expert cannot be recognized for a long time in the video data, it is automatically determined that the bid evaluation expert may have left the post, triggering an "off-duty" abnormality warning.

[0065] Optionally, if guiding or biased statements (such as attempts to influence review opinions or guide voting directions) are detected in the audio data, it is determined to be "abnormal speech" and a warning is triggered.

[0066] Optionally, if no abnormal behavior is detected in the video data and the audio data, it is determined to be "no abnormality", and the bid evaluation process is considered normal.

[0067] Optionally, the review results of each bid evaluation expert are obtained, and the review results of different bid evaluation experts are compared. If the review result of a certain expert deviates significantly from the review results of other bid evaluation experts, it is marked as abnormal, and an automatic feedback mechanism is used to display the abnormal information to the administrator for review. In this way, the transparency of the review process can be ensured, and abnormal review behavior that may affect the fairness of the results can be discovered in a timely manner.

[0068] It should be noted that in order to realize long-term storage and traceable query of analysis results, a lightweight database structure based on SQLite is designed. The following key information fields are stored in the database: Client IP address, audio file name, video file name, bid evaluation expert name, whether there is inappropriate speech (Boolean judgment), analysis time, whether normal (complementary to the above-mentioned fields), geographic location (such as the city where the expert is located), etc.

[0069] Optionally, the detection result is written into the database table audio_info through a standardized structure, realizing complete data closed-loop management, facilitating subsequent manual review, statistical analysis or automatic alarm.

[0070] Table 1 is an example of basic information of a video file in a database. As shown in Table 1, the video file contains client IP address, video file name, bid evaluation expert name, and whether to leave the post, etc.

[0071] Table 1 is an example of basic information of a video file in a database. As shown in Table 1, the video file contains client IP address, video file name, bid evaluation expert name, and whether to leave the post, etc. Table 2 is an example of basic information of an audio file in a database. As shown in Table 1, the audio file contains client IP address, audio file name, bid evaluation expert name, and whether to leave the post, etc.

[0072] Table 2 is an example of basic information of an audio file in a database. As shown in Table 1, the audio file contains client IP address, audio file name, bid evaluation expert name, and whether to leave the post, etc. It can be understood that, by taking the language detection result and the face detection result as key basis, the behavior of the bid evaluation expert is analyzed and judged to determine whether there is abnormal behavior, which can more comprehensively and accurately reflect the behavior performance of the bid evaluation expert in the bid evaluation process; this comprehensive judgment based on multiple detection results can effectively improve the accuracy and reliability of the bid evaluation expert behavior monitoring, and ensure that the fairness, objectivity and seriousness of the bid evaluation work are guaranteed.

[0073] In some embodiments, after obtaining the abnormal behavior recognition result, further comprising: In the case where it is determined that the bid evaluation expert has inappropriate speech, and / or the bid evaluation expert has off-duty behavior, generating a warning information for alarm.

[0074] The warning information includes the type and risk level of the abnormal behavior, and the suggested measures.

[0075] It can be understood that, by generating a warning information for alarm in the case where it is determined that the bid evaluation expert has inappropriate speech, and / or the bid evaluation expert has off-duty behavior, it is helpful to discover potential risks in advance, control and handle factors that may affect the fairness of the bid evaluation in the early stage, reduce the probability of major mistakes or irregularities in the bid evaluation process, and ensure that the fairness and objectivity of the bid evaluation result are fully guaranteed.

[0076] Figure 2 A flowchart of the training process of the text classification model provided by the embodiments of the present application is shown. As shown in some embodiments, the training process of the text classification model includes: Figure 2 Step 210, obtaining sample audio data of a sample bid evaluation expert in a remote bid evaluation process; Step 220, performing speech recognition on the sample audio data to obtain sample text data; Step 230, determining a language detection result label corresponding to the sample audio data; Step 240, inputting the sample text data into a pre-constructed initial text classification model to obtain a language detection prediction result output by the initial text classification model; the language detection prediction result includes whether the sample bid evaluation expert has inappropriate remarks; Step 250, calculating a first loss function value based on the language detection prediction result and the language detection result label; Step 260, based on the first loss function value, iteratively optimizing the parameters of the initial text classification model to obtain a text classification model.

[0077] Wherein, the inappropriate remarks in the sample text data are trained as follows: (1) Guiding sentence example (for training to detect guiding tendency): “Do you think this proposal is innovative?” “How do you evaluate this candidate?” “Do you agree with the feasibility of this project?” (2) Unregulated sentence example (for training to detect unregulated expressions): “This research direction may not have practical application value at the current stage.” “I suggest we focus on reviewing other projects.” “I have cooperated with the team and understand their working style.” (3) Other risk language examples (such as revealing privacy, evaluation discrimination): “The project leader has performed well in past projects and is worth paying attention to.” “Our unit usually gives priority to projects with international cooperation background.” “The technical solution of the project meets the evaluation standards of our unit.” Optionally, inappropriate remarks include but are not limited to: biased remarks, leading remarks, discriminatory remarks, confidential remarks, inappropriate evaluation remarks, etc.

[0078] Optionally, the initial text classification model is constructed based on the Chinese BERT language model. ​

[0079] Optionally, the language detection result label is 1 or 0, and "1" represents the existence of improper speech, and "0" represents the non-existence of improper speech.

[0080] Optionally, the sample audio data and the corresponding language detection result label are divided into training data and test data, and the proportion of the training data is about 80%, and the proportion of the test data is about 20%.

[0081] Optionally, the initial text classification model is trained based on the training data to obtain a text classification model, the text classification model is tested based on the test data to obtain a test result, and the parameters of the text classification model are optimized according to the test result.

[0082] Optionally, the sample text data is segmented by using a pre-trained Chinese BERT segmenter (such as a "bert-base-chinese" model).

[0083] Optionally, the sample text data is batch-processed by using the segmenter, long texts are truncated, and short texts are padded to make the text lengths consistent.

[0084] Optionally, before training the initial text classification model, the following parameters are set: Directory of training output: place for saving training results and models; Evaluation frequency: evaluation after each training round; Learning rate: parameter for controlling the learning speed of the model; Batch size on each device: number of samples fed to the model at a time; Number of training rounds: number of times of repeating training of the entire data set; Weight decay: a way of regularization, used to prevent overfitting; Log output directory and step number: frequency and location of saving log information of the training process.

[0085] Optionally, the initial text classification model, the training data, the evaluation data, and the training parameters are integrated by using a Trainer trainer to start model training. During the entire training process, a series of operations such as forward propagation, backward propagation, loss calculation, and model updating are automatically processed; after the training is completed, the text classification model is saved in a directory (such as:. / results), including model parameters, config, tokenizer, and the like.

[0086] Figure 3 A flowchart of a training process of a face detection model provided by an embodiment of the present application is shown in FIG. 1. Figure 3 As shown in FIG. 1, in some embodiments, the training process of the face detection model includes: Step 310: Obtain sample video data of the sample evaluation experts during the remote evaluation process; Step 320: Determine the label of the face detection result corresponding to the sample video data; Step 330: Input the sample video data into the pre-built initial face detection model to obtain the face detection prediction results output by the initial face detection model; Step 340: Calculate the second loss function value based on the face detection prediction result and the face detection result label; Step 350: Based on the second loss function value, iteratively optimize the parameters of the initial face detection model to obtain the face detection model.

[0087] Optionally, the face detection prediction results may include whether the sample evaluation experts have left their posts.

[0088] Optionally, based on a face detection model, face detection is performed on the sample video data, and the predicted duration when no face of the sample evaluation expert is detected for a continuous period is recorded. If the predicted duration exceeds a preset duration threshold, it is predicted that the evaluation expert has left their post.

[0089] The following describes the system for identifying abnormal behavior of bid evaluation experts during remote bid evaluation video conferences provided in this embodiment of the invention. The system for identifying abnormal behavior of bid evaluation experts during remote bid evaluation video conferences described below can be referred to in correspondence with the method for identifying abnormal behavior of bid evaluation experts during remote bid evaluation video conferences described above.

[0090] Figure 4 This is a schematic diagram of the system for identifying abnormal behavior of bid evaluation experts during remote bid evaluation video conferences provided in this embodiment of the invention. Figure 4 As shown, the abnormal behavior detection system 400 for evaluation experts during the remote bidding video conference includes: The acquisition unit 410 is used to acquire the audio and video data of the evaluation experts during the process of remotely evaluating the pre-set project through a video conferencing evaluation tool. The audio detection unit 420 is used to perform language detection on the audio data and obtain language detection results; the language detection results include: whether the evaluation experts made inappropriate remarks; The video detection unit 430 is used to perform face detection on video data and obtain face detection results; the face detection results include: whether the evaluation experts have left their posts; The recognition unit 440 is used to identify abnormal behavior of the evaluation experts based on language detection results and face detection results, and obtain abnormal behavior recognition results.

[0091] Optionally, language detection is performed on the audio data to obtain language detection results, including: performing speech recognition on the audio data to obtain text data; inputting the text data into a pre-constructed text classification model to obtain a language detection result output by the text classification model; The text classification model is obtained based on sample text data corresponding to sample audio data and language detection result labels corresponding to the sample audio data.

[0092] Optionally, the video data is subjected to face detection to obtain a face detection result, including: inputting the video data into a pre-constructed face detection model to obtain a face detection result output by the face detection model; The face detection model is obtained based on sample video data and face detection result labels corresponding to the sample video data.

[0093] Optionally, the video data is input into a pre-constructed face detection model to obtain a face detection result output by the face detection model, including: based on the face detection model, performing face detection on the video data, recording a time length during which no face of the bid evaluation expert is continuously detected, and determining that the bid evaluation expert has an off-duty behavior if the time length exceeds a preset time length threshold.

[0094] Optionally, the abnormal behavior of the bid evaluation expert is identified based on the language detection result and the face detection result, including: determining that the bid evaluation expert has an abnormal behavior if the bid evaluation expert has inappropriate remarks and / or the bid evaluation expert has an off-duty behavior; determining that the bid evaluation expert has no abnormal behavior if the bid evaluation expert has no inappropriate remarks and the bid evaluation expert has no off-duty behavior.

[0095] Optionally, the abnormal behavior identification system of the bid evaluation expert in the remote bid evaluation video conference further includes: a warning unit configured to generate a warning information and give an alarm if the bid evaluation expert has inappropriate remarks and / or the bid evaluation expert has an off-duty behavior.

[0096] Optionally, the training process of the text classification model includes: obtaining sample audio data of a sample bid evaluation expert in a remote bid evaluation process; performing speech recognition on the sample audio data to obtain sample text data; determining language detection result labels corresponding to the sample audio data; inputting the sample text data into an initial text classification model to obtain a language detection prediction result output by the initial text classification model, the language detection prediction result including whether the sample bid evaluation expert has improper remarks; calculating a first loss function value based on the language detection prediction result and the language detection result label; performing iterative optimization on parameters of the initial text classification model based on the first loss function value to obtain the text classification model.

[0097] Optionally, the training process of the face detection model includes: obtaining sample video data of the sample bid evaluation expert in the remote bid evaluation process; determining a face detection result label corresponding to the sample video data; inputting the sample video data into an initial face detection model to obtain a face detection prediction result output by the initial face detection model; calculating a second loss function value based on the face detection prediction result and the face detection result label; performing iterative optimization on parameters of the initial face detection model based on the second loss function value to obtain the face detection model.

[0098] It should be noted that the abnormal behavior identification system of the bid evaluation expert in the remote bid evaluation video conference process provided by the embodiments of the present application can realize all the method steps realized by the method embodiment of the abnormal behavior identification of the bid evaluation expert in the remote bid evaluation video conference process, and can achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiment in this embodiment will not be described in detail.

[0099] Figure 5 The structural schematic diagram of the electronic device provided by the embodiments of the present application is as follows: Figure 5As shown, the electronic device can include a processor 510, a communications interface 520, a memory 530, and a communications bus 540, wherein the processor 510, the communications interface 520, and the memory 530 complete mutual communication through the communications bus 540. The processor 510 can invoke a logic instruction in the memory 530 to execute the method for identifying abnormal behavior of a bid evaluation expert in a remote bid evaluation video conference process, which includes: obtaining audio data and video data of the bid evaluation expert in a process in which the bid evaluation expert performs remote bid evaluation on a preset project through a video conference bid evaluation tool; performing language detection on the audio data to obtain a language detection result; the language detection result includes whether the bid evaluation expert has inappropriate speech; performing face detection on the video data to obtain a face detection result; the face detection result includes whether the bid evaluation expert has an off-duty behavior; and identifying abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result to obtain an abnormal behavior identification result.

[0100] In addition, the logic instruction in the memory 530 described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0101] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement it without creative labor.

[0102] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of 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 the various embodiments or some parts of the embodiments.

[0103] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; 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 application.

Claims

1. A method for identifying abnormal behavior of bid evaluation experts in a remote bid evaluation video conference process, characterized in that, The method comprises the following steps: In the process of remote bid evaluation of a preset project by bid evaluators through a video conference bid evaluation tool, audio data and video data of the bid evaluators are acquired; Language detection is performed on the audio data to obtain a language detection result; The language detection result comprises whether the bid evaluators have inappropriate remarks; Face detection is performed on the video data to obtain a face detection result; the face detection result comprises whether the bid evaluators have off-duty behavior; Based on the language detection result and the face detection result, abnormal behavior of the bid evaluators is identified to obtain an abnormal behavior identification result. 2.The method of claim 1, wherein, The language detection on the audio data to obtain the language detection result comprises: Speech recognition is performed on the audio data to obtain text data; The text data is input into a pre-constructed text classification model to obtain a language detection result output by the text classification model; The text classification model is obtained by training based on sample text data corresponding to sample audio data and language detection result labels corresponding to the sample audio data. 3.The method of claim 1, wherein, The face detection on the video data to obtain the face detection result comprises: The video data is input into a pre-constructed face detection model to obtain a face detection result output by the face detection model; The face detection model is obtained by training based on sample video data and face detection result labels corresponding to the sample video data.

4. The method of claim 3, wherein the method further comprises: The inputting of the video data into the pre-constructed face detection model to obtain the face detection result output by the face detection model comprises: Based on the face detection model, face detection is performed on the video data, and a time length of continuous non-detection of a face of the bid evaluator is recorded; in a case where it is determined that the time length exceeds a preset time length threshold, it is determined that the bid evaluator has off-duty behavior.

5. The method of claim 1, wherein the method further comprises: The identification of the abnormal behavior of the bid evaluators based on the language detection result and the face detection result comprises: In a case where it is determined that the bid evaluators have inappropriate remarks and / or the bid evaluators have off-duty behavior, it is determined that the bid evaluators have abnormal behavior; In a case where it is determined that the bid evaluators do not have inappropriate remarks and the bid evaluators do not have off-duty behavior, it is determined that the bid evaluators have no abnormal behavior. 6.The method of claim 1, wherein, After the abnormal behavior identification result is obtained, the method further comprises: In a case where it is determined that the bid evaluators have inappropriate remarks and / or the bid evaluators have off-duty behavior, warning information is generated to perform warning.

7. The method of claim 2, wherein the method further comprises: determining whether the bid evaluation expert has an abnormal behavior in the remote bid evaluation video conference process. The training process of the text classification model comprises: Sample audio data of sample bid evaluators in a remote bid evaluation process is acquired; Speech recognition is performed on the sample audio data to obtain sample text data; Language detection result labels corresponding to the sample audio data are determined; The sample text data is input into a pre-constructed initial text classification model to obtain a language detection prediction result output by the initial text classification model; the language detection prediction result comprises whether the sample bid evaluators have inappropriate remarks; The first loss function value is calculated based on the language detection prediction result and the language detection result label; The parameters of the initial text classification model are iteratively optimized based on the first loss function value, to obtain the text classification model.

8. The method of claim 3, wherein the method further comprises: determining whether the bid evaluation expert has an abnormal behavior in the remote bid evaluation video conference process. The training process of the face detection model includes: Obtaining sample video data of a sample bid evaluation expert in a remote bid evaluation process; Determining a face detection result label corresponding to the sample video data; Inputting the sample video data into a pre-constructed initial face detection model to obtain a face detection prediction result output by the initial face detection model; The second loss function value is calculated based on the face detection prediction result and the face detection result label; The parameters of the initial face detection model are iteratively optimized based on the second loss function value, to obtain the face detection model.

9. An abnormal behavior identification system for bid evaluation experts in a remote bid evaluation video conference process, characterized in that, Comprise: An acquisition unit, configured to acquire audio data and video data of a bid evaluation expert in a process in which the bid evaluation expert performs remote bid evaluation on a preset project by using a video conference bid evaluation tool; An audio detection unit, configured to perform language detection on the audio data to obtain a language detection result; The language detection result includes whether the bid evaluation expert has inappropriate remarks; A video detection unit, configured to perform face detection on the video data to obtain a face detection result; the face detection result includes whether the bid evaluation expert has an off-duty behavior; An identification unit, configured to identify an abnormal behavior of the bid evaluation expert based on the language detection result and the face detection result, to obtain an abnormal behavior identification result.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the computer program to implement the method for identifying an abnormal behavior of a bid evaluation expert in a remote bid evaluation video conference process according to any one of claims 1 to 8.

Images