TWI689843B - Intelligent appraisal system and method therof - Google Patents

Abstract

An intelligent appraisal system and method thereof are provided. The interactive appraisal system includes a user interface, a chatting robot device and a semantic analyzing engine. The chatting robot device receives a question message from an appraisal database to actively present the question message on the user interface. A reply message is sent to the semantic analyzing engine when the user interface receives the reply message. The semantic analysis engine extracts keywords from the reply message by using an algorithm module so as to parse the reply message. A classification module classifies the analyzed reply message into categories according to a field distribution. A score module gives scores according to contents of the classified reply message. A report generating device then compiles the reply message, the categories and the scores to generate a report.

Description

Intelligent inspection system and method

The invention is an interactive verification system and method, which is to make a query through a user and a chat robot and analyze the result of the query to be used as verification statistics.

The audit is used to collect the feedback from users and analyze the feedback for subsequent management. Generally speaking, the inspection department is used for the course content of teaching content, the follow-up of consumer behavior, feedback on service content or products, etc.

Existing inspections mainly use paper-based testing or online answering, but paper-based testing is susceptible to time and location constraints in actual use, which makes the response rate for answering questions generally low. In addition, the paper consumed by the paper also has environmental concerns. On the other hand, due to the lack of an interaction mechanism with the testee, the paper test under the same questionnaire may be explained by different questionnaires. It caused errors, and there was no way for the questionnaire questions to provide other explanations so that the testees could get different question descriptions through interaction, which in turn affected the accuracy of answering.

Although answering online can change the content of the paper questionnaire electronically, Multiple expression effects and methods, and can reduce the use limitations of the testee in time and place, but there is still no interactive mechanism that is lacking in the paper answering process. When the testee cannot understand the question, he cannot borrow it in real time. The interactive mechanism provides a further explanation of the question. Online answering also cannot greatly improve the accuracy of the questionnaire. In addition, when filling and answering qualitative questions, we cannot get more information through the interactive process.

To solve the aforementioned problems, the present invention provides an interactive verification system, which includes: a user interface; a chat robot device connected to the user interface and a verification database; the chat robot device is based on the verification data The library obtains the question message to actively present the question message to the user interface. After the user interface receives the reply message, the reply message is sent to the semantic analysis engine; the semantic analysis engine, which includes an algorithm module , A classification module, and a scoring module, wherein the algorithm module extracts the reply message through keywords to parse the reply message, and then sends the parsed reply message to the classification module and the scoring module, For the classification module to classify the parsed reply message into categories configured according to the field, and cause the scoring module to give a score based on the content of the classified reply message; and a report generation device that aggregates the The reply message parsed by the semantic analysis engine, the classification result of the classification module and the score given by the scoring module to generate a report.

The present invention also provides an interactive verification method, which includes: obtaining a question message from the verification database; actively presenting the question message to a user interface through a chat robot device; receiving a reply message to send the reply message to Semantic analysis engine; extract the reply message through keyword extraction through the semantic analysis engine's algorithm module; parse the reply message through the semantic analysis engine's classification module to classify the parsed reply message according to domain configuration The category of; the score module of the semantic analysis engine gives a score according to the content of the reply message after classification; and summarizes the reply message parsed by the semantic analysis engine, the classification result of the classification module, and the scoring module The scores given by the group to generate reports through the report generating device.

The aforementioned interactive verification system and method further include a reply database, which stores the reply message processed by the semantic analysis engine to provide the reply message to the report generation device.

The foregoing interactive verification system and method further include a learning program device that performs machine learning on the results of the semantic analysis engine through the model to transmit the results to the semantic analysis database.

In the aforementioned interactive verification system and method, the chat robot device judges the validity of the verification based on the time interval at which the reply message is continuously received by the user interface.

In the aforementioned interactive verification system and method, the chat robot device judges the validity of the verification based on the obtained question message with reverse meaning.

In the foregoing interactive verification system and method, the question message presented by the chat robot device is presented in the form of a text button, a picture selection answer, or a text answer, wherein the question message of the text button is given by a button Option for the chat robot device to convert the reply message selected by the text button into text; The video provides options for the chat robot device to convert the reply message selected by the picture into text; and the question message filled in the text expresses the question message in text form and receives the reply message in text form.

The foregoing interactive verification system and method, wherein the algorithm module performs keyword extraction through the semantic analysis database to generate a multi-dimensional word-breaking reply message to parse the reply after reducing the dimension message.

The aforementioned interactive verification system and method, wherein the classification module classifies the content of the reply message according to the classification model in the semantic analysis database, and the category configured according to the domain is set in the classification module Category database.

In the foregoing interactive verification system and method, wherein the scoring module is based on the positive or negative value range of the emotional vocabulary of the content in the reply message.

With the popularization of the current network and smart mobile devices, the bi-directional interactive intelligent inspection system of the present invention can be inspected without restriction of the time and location of the answer, and can be more intuitively interactive Conversation, the verification of interactive queries with chatbots on the instant messaging platform, not only can increase the answer rate, but also can increase the accuracy of the questionnaire through the real-time interactive mechanism and the mechanism to judge the validity of the questionnaire during the verification process .

11: User interface

12: chat robot device

13: Audit database

14: Semantic analysis engine

15: Reply database

16: learning program device

17: Report generation device

S1-S14: Step

Figure 1 is a schematic diagram of the system architecture of the interactive inspection system of the present invention; Figures 2 to 3 are the user interface and chat robot module of the present invention Fig. 4 is a schematic diagram of the reply message of the present invention after word breaking; and Fig. 5 is a schematic diagram of the operation flow of the interactive verification system of the invention.

The present invention provides an interactive verification system and an interactive verification method. As shown in FIG. 1, the interactive verification system includes: a user interface 11, a chat robot device 12, a verification database 13, and a semantic analysis engine 14. Answer database 15, learning program device 16, and report generating device 17.

The user interface 11 is an operation interface for receiving user's voice, text, pictures, options and other reply messages. It can be a social software, a communication platform, a website, a chat room, or an application. The user can use the user interface 11 Talk or communicate with chatbots by voice input, text input, and click.

The chat robot device 12 is connected to the user interface 11 and the audit database 13. After the chat robot device 12 obtains the question message that has been built in the audit database 13 from the audit database 13, it uses buttons, text, Pictures, or any combination thereof, actively present the question message to the user interface 11, as shown in Figures 2 and 3, after the user interface 11 receives the user's reply message, the reply message is sent to the semantic analysis Engine 14. Wherein, when the chat robot device 12 provides a text button, a picture selection answer, or a text filled question message, the question message of the text button presents an option in the form of a button or a card. After the user selects an answer according to the option, the chat robot The device 12 responds to the text button selected by the user The information is converted into text and stored in the answer database 15; the question message of the picture selection is provided in different blocks of the picture design to indicate different options. The user clicks the block to answer the reply message, and the chat robot device 12 sends the user The reply message of the selected picture block is converted into corresponding text, and the corresponding text is stored in the reply database 15; the question message filled in the text is sent by the chat robot device 12 to the verification database 13 Message, and the user answers the reply message in text form, the chatbot device 12 stores the user's text reply message in the reply database 15; the reply database 15 stores the aforementioned reply message processed by the semantic analysis engine to provide the reply message The reply message is to the report generating device 17.

The semantic analysis engine 14 includes an algorithm module, a classification module, and a scoring module, wherein the algorithm module analyzes the semantics of the user's reply message with a natural language processing algorithm, mainly through keyword extraction, through A large amount of text is a semantic analysis database composed of lexical parts of speech, and the sentence of the reply message is segmented to analyze the reply message, and the vocabulary used for model training is generated in the same way, as shown in Figure 4. Word-breaking sentences have a multi-dimensional structure, which makes the sentences easier to parse after lowering the dimensions.

The classification module of the semantic analysis engine 14 assigns the parsed reply messages to multiple categories configured according to the industrial field. The classification module includes a category database and a classification mechanism, where the category database is configured with different pre-built The categories correspond to the different industrial fields under inspection. The classification mechanism is a classification model built on the semantic analysis database. It initially assigns and categorizes the related words of each category and uses a large number of documents. The accuracy of this repeated training classification model is based on the examination of the learning effectiveness of the education industry. The categories used are "teaching method", "course content", "examination scoring method", etc. Taking Figure 4 as an example, The reply message will be categorized in the "Teaching Method" category.

The scoring module of the semantic analysis engine 14 gives a score based on the content of the reply message classified by the classification module, for example, filling in the vocabulary in the content according to the user's text, especially the vocabulary defined with emotional meaning, Judging its value range, the value range is divided into positive and negative directions, and the magnitude and width of the value range will be determined according to the inspection needs of different industrial fields. For example, in the case of the inspection of the learning effectiveness of the education industry, the value range It can be set in the range of [-1,1]. As in the example in Figure 4 above, if the word "uninteresting" is a negative word in the scoring tag database, the scoring module will The sentences belonging to the "teaching method" category in the figure are scored according to the negative degree of the emotional vocabulary, for example, "-0.63 points", and the aforementioned values are aggregated in the report generating device.

Table 1 below is an example of classification and scoring by the aforementioned classification module and scoring module:

The following example is an example of the training phase and prediction phase of the semantic analysis engine 14. In the training phase, the sentence input by the user is subjected to word segmentation and part-of-speech analysis to obtain the matrix O in Table 2 below. Depending on part of speech, it is possible Obtain multiple matrices (for example: O ₁ , O ₂ ,..., On _n ); reduce the dimensions of the high-dimensional matrix obtained after the aforementioned word-breaking to obtain the matrix S of the following Table 3 (according to the difference of parts of speech , It is possible to get the matrix S ₁ ,S ₂ ,...,S _n ); put the matrix S and the category value C into the classification module to create a classification model to get the model A; then according to the use of each category value The matrix S and the score V input by the user are modeled using the scoring module to obtain the model B (depending on the category, it is possible to obtain the matrix B ₁ , B ₂ , ..., B _n ).

In the prediction stage, the sentence input by the user is subjected to word segmentation and part-of-speech analysis to obtain the aforementioned matrix O (according to different parts of speech, it is possible to obtain multiple matrices O ₁ , O ₂ , ..., On _n ); Reduce the dimensions of the high-dimensional matrix obtained after the aforementioned word breaking to obtain the matrix S (depending on the part of speech, it is possible to obtain the matrix S ₁ , S ₂ ,..., S _n ); put the aforementioned matrix S Enter model A to obtain the category value C; select the corresponding model B according to the aforementioned category value C and obtain the predicted score through the matrix S.

For example: User input: The content of this lesson is too interesting

The predicted category is: Course content

The predicted score is: 4

The report generating device 17 of the present invention aggregates the parsed and word-breaking reply message, the classification result of the classification module and the score given by the scoring module from the answer database 15, after statistics, in the form of text or chart Generate reports, and use the algorithm module to summarize key indicators that facilitate decision-making.

The present invention further includes a learning program device 16, which performs machine learning on the analysis results of the semantic analysis engine through the model to transmit the results to the semantic analysis database.

In order to ensure the validity of the verification process and to confirm whether the user actually answers the question, the chat robot device 12 can continuously receive the reply message through the user interface through the user interface to confirm the user Fill in the time of the reply to judge the validity of the audit. The chat robot device 12 can also randomly send question messages with reverse meaning between multiple question messages to judge the effectiveness of the audit Sex (for example, ask if you like the activity in the first question, and ask the user if you don’t like the activity in question 5).

Please refer to the operation flow diagram of the interactive verification system in FIG. 5. When the user activates the interactive verification system of the present invention, the interactive verification system system determines the user's way of entering the system S1, and the chat robot device will ask The user intends S2 to confirm the identity (if the verification needs to confirm the user's identity, it is necessary to provide consent identity binding, and if there is no such requirement, the identity binding will not be performed) to determine the verification system the user enters There are three types of inquiries about intention to enter the verification system: providing code to enter the verification, direct statement to enter the verification (for example, talking to the chat robot device: I want to fill in the questionnaire for the OO lecture), and click the button to enter the verification (use When the user interacts with the chat robot device, when the chat robot device queries the user's intention, the user will be provided with a quick reply button for the user to click), then the user is allowed to enter the verification S4, and the chat robot device Check the category of S4, determine whether there is a temporary storage S5, and check whether there is a historical answer record in the identity according to whether the review has temporary characteristics, and ask whether to continue the previous answer S7, If it continues, continue with the previous verification S8; or after judging that there is no temporary storage content, the chat robot device will explain the verification content S9, and interact with the chat robot device verification process S10, after the interactive verification process is completed, the verification The interactive system confirms whether to submit the reply message S11, if it is, it is imported into the database S12, if not, the chat robot device asks the intention S2; then after S12, ask whether to continue to answer other questions S13, if yes, go back to step S4, if not, then end the audit S14; if asking whether to continue the previous answer S7 or whether to submit a reply If No is selected in the message S11, returning to the chat robot device will ask the user's intention S2.

The above detailed description is a specific description of a feasible embodiment of the present invention, but this embodiment is not intended to limit the patent scope of the present invention, and any equivalent implementation or change without departing from the technical spirit of the present invention should be included in The patent scope of this case.

To sum up, this case is not only innovative in terms of technical ideas, but also possesses the above-mentioned multiple functions that traditional methods do not match, which has fully met the requirements of novelty and progressive legal invention patents. Approve this application for a patent for invention to encourage invention and achieve good results.

11: User interface

12: chat robot device

13: Audit database

14: Semantic analysis engine

15: Reply database

16: learning program device

17: Report generation device

Claims (10)

An intelligent inspection system includes: a user interface; a chat robot device connected to the user interface, the chat robot device is used to obtain a problem message to actively present the problem message to the user interface, so as to After receiving the reply message, the user interface sends the reply message; the semantic analysis engine includes an algorithm module, a classification module, and a scoring module. After receiving the reply message, the algorithm module will The reply message is extracted by keywords to parse the reply message, and then the parsed reply message is sent to the classification module and the scoring module for the classification module to classify the parsed reply message to The category configured by the domain, and the scoring module is given a score according to the content of the categorized reply message; and a report generation device that aggregates the reply message parsed by the semantic analysis engine and the classification result of the classification module And the score given by the scoring module to generate a report; wherein, the chat robot device is judged according to the time interval of continuously receiving the reply message by the user interface and according to the obtained question message with reverse meaning For the validity of this check, the scoring module is based on the positive or negative value range of the emotional vocabulary in the reply message. The intelligent verification system according to item 1 of the patent application scope further includes a reply database, which is used to store the semantic analysis engine The processed reply message is used to provide the reply message to the report generation device, or further includes a learning program device, which performs machine learning on the analysis result of the semantic analysis engine through the model to send the result to the semantic analysis database. According to the intelligent verification system described in item 1 of the patent application scope, wherein the question message presented by the chat robot device is presented in the form of a text button, a picture selection answer, or a text answer, in which the question of the text button The message is given an option via a button for the chat robot device to convert the reply message selected by the text button into text; the question message selected by the picture is selected by the picture for the chat robot device to select the answer message by the picture The reply message is converted into text; and the question message filled in the text is to express the question message in text form and receive the reply message in text form. According to the intelligent verification system described in item 1 of the patent application scope, the algorithm module executes keyword extraction through the semantic analysis database to generate the multi-dimensional word-breaking reply message in order to reduce After the dimension, parse the reply message. The intelligent verification system according to item 1 of the patent application scope, wherein the classification module classifies the content of the reply message according to the classification model in the semantic analysis database, and the category configured according to the domain is set at The category database of the classification module. An intelligent inspection method, including: Obtain the question message; actively present the question message to the user interface through the chat robot device; receive the reply message to send the reply message to the semantic analysis engine; extract the reply message through keyword extraction through the algorithm module of the semantic analysis engine Parse the reply message; classify the parsed reply message to the category configured by the domain through the classification module of the semantic parsing engine; give the scoring module of the semantic parsing engine based on the content of the reply message after classification A score; and a summary of the reply message parsed by the semantic analysis engine, the classification result of the classification module and the score given by the scoring module to generate a report through a report generation device; wherein the chat robot device is based on the continuous The time interval of the received reply message and the question message obtained with the reverse meaning are used to judge the validity of the review. The scoring module is based on the positive or negative emotion words in the reply message. Scores are given in the negative range. The intelligent verification method according to item 6 of the patent application scope further includes: storing the reply message processed by the semantic analysis engine in the reply database, or, further comprising: analyzing the result of the semantic analysis engine through the model Perform machine learning to send the results to the semantic analysis database. According to the intelligent verification method described in item 6 of the patent application scope, wherein, The chat robot device provides: an option given by a button for the chat robot device to convert the reply message selected by the text button into text; a choice provided by a picture for the chat robot device to convert the reply message selected by the picture As text; and express the question message in text form and receive the reply message in text form. According to the intelligent verification method described in Item 6 of the patent application scope, the algorithm module performs keyword extraction through the semantic analysis database on the reply message, and generates a multi-dimensional word-breaking reply message to reduce After the dimension, parse the reply message. The intelligent verification method according to item 6 of the patent application scope, wherein the classification module classifies the content of the reply message according to the classification model in the semantic analysis database, and the category configured according to the domain is set at The category database of the classification module.

Images