JP2019016046A - Questionnaire evaluation system and questionnaire evaluation method - Google Patents

Abstract

To provide a questionnaire evaluation system evaluating a degree of seriousness of a reply to a questionnaire and selecting contents of a reward (processing contents to be provided to a user) in accordance with this degree.SOLUTION: A questionnaire evaluation system according to the present invention comprises a seriousness degree analyzing unit performing natural language processing of text data of a sentence described by a user in a questionnaire and evaluating a degree of seriousness of the user in a reply to the questionnaire and a processing selection unit selecting data contents to be provided to the user in accordance with the degree.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a questionnaire evaluation system and a questionnaire evaluation method for evaluating the contents described in a questionnaire.

In recent years, questionnaires have been conducted on products to be evaluated (target products), responses from multiple respondents have been analyzed, and the degree of recognition of the target products, whether or not they have been purchased, and the evaluation of products has been investigated. (For example, refer to Patent Document 1).
And in order to urge an answer to the questionnaire, a reward such as a point is given to the user who answered the questionnaire.

JP 2007-87339 A

As described above, since the reward is given, the response rate from the user to the questionnaire can be improved.
However, although it is unclear whether all users answering the questionnaire are seriously answering the questionnaire, the same reward is given to all the users who answered.

When a questionnaire from a user is aggregated and sales promotion or product development is performed, unless the content of the questionnaire response is serious, it cannot be used as information for reflecting in the examination when performing sales promotion or product development.
For this reason, as a reward for the answer to the questionnaire, the same reward is given to all of the users who answered, so the result is a lack of fairness for the users who are seriously responding.

  The present invention has been made in view of such circumstances, and evaluates the degree of seriousness of the contents of the answer to the questionnaire, and the contents of the reward so as to be fair to each user according to the degree of seriousness. An object is to provide a questionnaire evaluation system and a questionnaire evaluation method for selecting (processing contents to be provided to a user).

  The present invention has been made to solve the above-described problems, and the questionnaire evaluation system of the present invention performs natural language processing of text data of a sentence described by a user in a questionnaire, and the user's seriousness in answering the questionnaire A seriousness analysis unit that evaluates the degree of data, and a process selection unit that selects data content to be provided to the user corresponding to the degree.

  The questionnaire evaluation method of the present invention includes a process of performing natural language processing of text data of a sentence described by a user in a questionnaire and evaluating the degree of seriousness of the user in answering the questionnaire, and the user corresponding to the degree And a process of selecting data contents to be provided.

  According to this invention, the degree of seriousness, which is the degree of seriousness when the user writes the sentence, is obtained from the text data of the questionnaire text, and the content of the reward is selected corresponding to the degree of seriousness, that is, the degree of seriousness. As a result, it is possible to easily perform a process of giving a better reward to a user who has a higher degree of seriousness than a user who has a low degree of seriousness.

It is a conceptual diagram which shows the structural example of the questionnaire evaluation system 1 by the 1st Embodiment of this invention. It is a figure which shows an example of the questionnaire screen displayed on the display screen of the user terminal 200 of the user who accessed the questionnaire evaluation system. 3 is a diagram illustrating an example of a configuration of a user table stored in a storage unit 15. FIG. It is a figure which shows an example of the screen of point provision in case the reward in this embodiment is a point. It is a flowchart which shows the operation example of the selection process of the reward by the text of the questionnaire which the questionnaire evaluation system 1 performs. It is a conceptual diagram which shows the structural example of 1 A of questionnaire evaluation systems by the 2nd Embodiment of this invention.

<First Embodiment>
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a conceptual diagram showing a configuration example of a questionnaire evaluation system 1 according to the first embodiment of the present invention. In FIG. 1, the questionnaire evaluation system 1 includes an input control unit 11, a language analysis unit 12, a seriousness analysis unit 13, a process selection unit 14, and a storage unit 15. The questionnaire evaluation system 1 is installed in, for example, an information center that aggregates company questionnaires.
The questionnaire evaluation system 1 is connected to a plurality of user terminals 200 through an information communication line 100 including the Internet. In the present embodiment, for example, a user is connected to the questionnaire evaluation system 1 via the information communication line 100 by the user terminal 200.

FIG. 2 is a diagram illustrating an example of a questionnaire screen displayed on the display screen of the user terminal 200 of the user who has accessed the questionnaire evaluation system 1.
FIG. 2A shows an example of a questionnaire start display screen displayed on the screen of the user terminal 200 when the user accesses the questionnaire evaluation system. It describes that a questionnaire is evaluated and a reward (reward) for the questionnaire is given based on the evaluation result. For example, in the present embodiment, the character string “The answer to this questionnaire is analyzed by artificial intelligence. If you give a better answer, you will receive a high reward.” Is done. As a result, since the artificial intelligence evaluates the questionnaire by reading the description of the above character string, the content of the sentence in the questionnaire is evaluated fairly, and if it is written seriously (seriously), a higher reward is obtained. Consciously encouraged to write a questionnaire seriously. Clicking the button with the character string “start questionnaire” in FIG. 2A with a mouse or the like displays a questionnaire input screen.

FIG. 2B shows an example of a questionnaire input screen. Check boxes 511, 512 and 513 are provided as check boxes for entering gender. Check box 511 is checked when the respondent of the questionnaire is male, check box 512 is checked (selected) when the respondent is female, and check box 513 is checked when the respondent does not want to answer gender. Is done. In addition, check boxes 515 and 516 corresponding to the answer of age are provided. When the user describes his / her age, the check box 515 is checked and the age is described in the age entry column 521. When the user does not want to answer the age (does not want to answer), the check box 516 is checked.
In the area 530, “What do you think about XX? (Free description)” is described as the question content of the questionnaire. As a result, the user describes an answer for XX in the text box 550.

  Returning to FIG. 1, the input control unit 11 stores the data in the check boxes 511, 512, 513, 515, 516, the age entry column 521 and the text box 550 input on the input screen shown in FIG. Write to 15 user tables and store.

FIG. 3 is a diagram illustrating an example of a configuration of the user table stored in the storage unit 15. The user table shown in FIG. 3 is provided with columns of user identification information, attribute information, text data index, seriousness, and grant points for each record.
The user identification information is identification information for identifying a user who has answered a questionnaire, and is, for example, a user's mail address. The attribute information is data such as gender and age input in FIG. The text data index is an address in the storage unit 15 in which text data written in the text box 550 by the user is stored. The seriousness is a degree indicating whether or not the answer is described below seriously, calculated based on the text data. The granted point is a point as a reward to be given to the user, determined in accordance with the seriousness.

Returning to FIG. 1, the language analysis unit 12 reads text data from the storage unit 15 by referring to the text index of the table in the storage unit 15. The language analysis unit 12 performs natural language processing on the text data and obtains an evaluation value for the text data.
The seriousness analysis unit 13 obtains a seriousness degree indicating the degree of seriousness of the user describing the text data based on the evaluation value.
The process selection unit 14 selects process contents to be provided to the user who has written the text data in accordance with the seriousness. The processing contents to be provided include rewards such as reward points, and are rewards such as winning probabilities in games, information contents to be provided, discount coupons for products, and the like.

As described above, an expensive reward having higher information is set as the seriousness is higher. FIG. 4 is a diagram illustrating an example of a point grant screen when the reward in this embodiment is a point.
FIG. 4A shows that, for example, 500 points are given to a user having a high degree of seriousness.
On the other hand, Fig.4 (a) has shown that it is 200 points as a point provided with respect to a user with low seriousness.
As shown in FIG. 4, the process selection unit 14 gives points set for each of cases where the degree of seriousness is high or low to each user who answers the questionnaire corresponding to the degree of seriousness. For example, the process selection unit 14 determines that the seriousness is high when the seriousness is equal to or higher than a preset threshold value, and determines that the seriousness is low when the seriousness is less than the threshold. In addition, as the evaluation of seriousness, a numerical value range of seriousness is set in multiple stages, and rewards (number of points to be awarded) are ordered in descending order of seriousness for each range from high seriousness to low seriousness. ) Level may be changed.

Moreover, the process selection part 14 is good also as a structure which changes the winning probability in the game which can be performed after answering a questionnaire according to a seriousness degree instead of performing the division of the point to provide. For example, the process selection unit 14 may be configured to set the winning probability of a slot game or the like higher as the seriousness increases.
In addition, the process selection unit 14 has different contents of information depending on the degree of seriousness of the transition destination of the Web page to which the user can connect, that is, the higher the degree of seriousness, the more information content is obtained, and the content that is beneficial to the user The configuration may be changed to a Web page.

Moreover, it is good also as a structure which changes the content of information, such as a product to provide, corresponding to seriousness. For example, if the degree of seriousness is high, thank you for answering the questionnaire and displaying a display screen that provides information such as pre-sale of products that are scheduled to be released, while if the degree of seriousness is low, The process selection unit 14 may be configured to display a display screen that gives thanks for answering the questionnaire.
Further, according to the seriousness, the volume of information such as the above-mentioned pre-sale may be adjusted between several lines to several pages, and the amount of information given to the user may be adjusted.
Also, if the seriousness is low, only important information (if the product, for example, when to start selling) is provided. If the seriousness is high, in addition to the important information, detailed information ( For example, it may be configured to provide information such as detailed product specifications and effective usage.

  Moreover, the process selection part 14 is good also as a structure which changes the discount rate of the discount coupon of a product corresponding to seriousness. For example, if the degree of seriousness is high, the discount rate of the discount coupon that discounts the price when purchasing the product is higher than the case of low degree of seriousness, while if the degree of seriousness is low, the discount rate of the discount coupon is increased. The processing selection unit 14 may be configured so as to be given lower (or not give a discount coupon) than when the seriousness is high. That is, the process selection part 14 is good also as a structure which sets high the discount rate of a discount coupon as seriousness becomes high.

Hereinafter, an evaluation value obtained by the language analysis unit 12 and a method for calculating the seriousness of the seriousness analysis unit 13 will be described.
・ Configuration # 1
The language analysis unit 12 performs morphological analysis of text data using a corpus (information such as a grammar of a target language or a word part of speech called a dictionary) as natural language processing, extracts words constituting the text data, and extracts them. The number of received words is used as the evaluation value.
For example, the seriousness analysis unit 13 divides the number of words extracted from the text data by the number of words that can be input into the text box 550, and calculates the division result as the seriousness.
In this configuration, it is assumed that the sentence as text data is longer and seriously written as the number of words is larger.

・ Configuration # 2
The language analysis unit 12 may be configured to perform dependency analysis (syntax analysis) between each word after performing morphological analysis of text data as natural language processing, and output the number of dependency as an evaluation value.
The seriousness analysis unit 13 extracts the maximum value from the number of dependency of all the questionnaires in a predetermined questionnaire period, divides the dependency number in each text data by this maximum value, and normalizes the result. The seriousness of each user.
In this configuration, it is assumed that the sentence of the text data can be considered as the dependency between words increases, and is seriously described.

・ Configuration # 3
The language analysis unit 12 performs the semantic analysis of the text data text as natural language processing, and uses the accuracy of the text obtained by the semantic analysis (evaluation level of meaning connection) as an evaluation value.
The seriousness analysis unit 13 extracts the maximum value from the above-described accuracy of all the questionnaire sentences in a predetermined questionnaire period, and normalizes the accuracy by dividing the accuracy of each text data by the maximum value. The result is obtained as seriousness.
Further, in the anaphora analysis, the evaluation value may be obtained corresponding to the ratio obtained by dividing the noun phrase determined to be omitted and complemented by the total number of words in the text data. In this case, the seriousness increases as the evaluation value decreases.

・ Configuration # 4
The language analysis unit 12 uses a text mining technique as natural language processing, divides text data into words and phrases, appearance frequency of these words and phrases, co-occurrence correlation, appearance tendency (covariance or correlation), Time series etc. are analyzed, and any one of these or a combination or all of them are used as evaluation values.
The seriousness analysis unit 13 generates an evaluation vector for each text data in an evaluation space with each evaluation value as an axis, and uses the size of the evaluation vector as the seriousness. Each axis is normalized by the maximum evaluation value of that axis.

・ Configuration # 5
The language analysis unit 12 uses a deep learning (deep learning) technique as natural language processing, and classifies input text data into corresponding labels by a deep learning prediction model. In this case, text data as teacher data is input in advance, and each of the text data is output as a label of a rank evaluated by a plurality of evaluators, and the prediction (or estimation) model is learned.
The seriousness analysis unit 13 outputs the seriousness set corresponding to the rank of the label as the seriousness of the user describing each text data.

・ Configuration # 6
The language analysis unit 12 uses a prediction model that uses a neural network to estimate that the prediction model sub is estimated when the composition and expression of the Japanese sentence in the sentence of the text data is accurate. 0 "to" 1 ") is output as the evaluation value. Here, the estimated value is, for example, a numerical value between “0” and “1”, and is a value that the sentence is better as it is closer to “1”. In this case, a plurality of feature values for presuming that the sentence is good are extracted from the text data, the feature value in the text data as the teacher data is input, and it is determined that the sentence of the text data is seriously described. The prediction model is trained in advance so that the output is “1” when it is present and the output is “0” when it is determined that it is not seriously described. The text data used as teacher data in advance is determined by a plurality of people as to whether it is a good sentence or a bad sentence.
Then, the seriousness analysis unit 13 outputs the estimated value as the seriousness.

FIG. 5 is a flowchart showing an operation example of a reward selection process based on questionnaire text performed by the questionnaire evaluation system 1. In the following description, the evaluation value of the language analysis unit 12 is used as a dependency number, and the processing selection unit 14 is described with a configuration in which points are given as rewards.
Step S11: The user accesses the questionnaire evaluation system 1 through the user terminal 200. Then, the user checks each of the check boxes on the questionnaire input screen shown in FIG. 2B displayed on the display screen of the user terminal 200, and then writes a sentence such as an impression and an opinion on the text box 550. Describe.

Step S12: The input control unit 11 extracts the check box and text data of the text input by the user on the questionnaire input screen, and writes and stores them in the user table of the storage unit 15.
Then, the language analysis unit 12 reads the text data of the sentence described by the user from the storage unit 15, and analyzes the read text data by the analysis method already described to obtain an evaluation value. That is, after performing morphological analysis on the text data, the language analysis unit 12 performs a dependency analysis of words in the text data, and outputs the number of the obtained dependency as an evaluation value.

Step S13: The seriousness analysis unit 13 obtains the seriousness corresponding to the evaluation value obtained by the language analysis unit 12 by the method for calculating the seriousness described above.
Here, when the dependency number of words in the text data is supplied from the language analysis unit 12, the seriousness analysis unit 13 compares the dependency number with a preset threshold value.
Moreover, it is good also as a structure which classify | categorizes a seriousness degree into multiple steps, determines which seriousness level the dependency number is included in, and gives a reward set for each step.

  Step S14: The seriousness analysis unit 13 determines that the seriousness is high when the dependency number is higher than the threshold, and advances the processing to step S15. On the other hand, the seriousness degree analysis unit 13 determines that the seriousness degree is low when the dependency number is less than the threshold value, and advances the processing to step S16. In addition, the seriousness analysis unit 13 writes and stores the determined seriousness in the seriousness column in the user table of the storage unit 15.

Step S15:
The process selection unit 14 reads out a preset number of points to be given when it is determined with reference to the storage unit 15 that the degree of seriousness set in advance is high.
And the process selection part 14 writes and memorize | stores the read point number in the user table of the memory | storage part 15, and displays the display screen of Fig.4 (a) with respect to the display screen of the user terminal 200. FIG.

Step S16:
The process selection unit 14 refers to the storage unit 15 and reads a preset number of points to be given when it is determined that the preset seriousness is low.
And the process selection part 14 writes and memorize | stores the number of points read to the user table of the memory | storage part 15, and displays the display screen of FIG.4 (b) with respect to the display screen of the user terminal 200. FIG.

According to this embodiment, when writing a sentence freely described by a user in a questionnaire, the seriousness indicating the degree of seriousness (seriously) described by the user is calculated based on the evaluation value obtained by processing the user text in natural language. . For this reason, according to the present embodiment, it is possible to obtain the seriousness from the evaluation result having fairness for each user, and the user who has a high degree of seriousness in the distribution of the reward to the user using this seriousness. On the other hand, it is possible to easily perform a process of giving a better reward as compared to a user with low seriousness. That is, it is possible to give a better reward to a user who has written a sentence that can be effectively reflected in product development or improvement, etc. (seriously described the sentence) compared to other users.
In addition, according to the embodiment, except for the questionnaire results of sentences with low seriousness, the questionnaire results of sentences with high seriousness that can be effectively reflected in product development or improvement are extracted, and the questionnaire is conducted. Can be provided to the organization that requested

<Second Embodiment>
The second embodiment of the present invention will be described below with reference to the drawings. FIG. 6 is a conceptual diagram showing a configuration example of a questionnaire evaluation system 1A according to the second embodiment of the present invention. In FIG. 6, the questionnaire evaluation system 1 </ b> A includes an input control unit 11, a language analysis unit 12, a seriousness analysis unit 13, a process selection unit 14, a storage unit 15, and a positive / negative classification unit 16. The questionnaire evaluation system 1A is installed, for example, in an information center that collects company questionnaires. The questionnaire evaluation system 1A is connected to a plurality of user terminals 200 through an information communication line 100 including the Internet. In the present embodiment, for example, a user is connected to the questionnaire evaluation system 1A via the information communication line 100 by the user terminal 200.
This embodiment is different from the first embodiment in that a positive / negative classification unit 16 is provided. Hereinafter, configurations and operations different from those of the first embodiment will be described.

After the language analysis unit 12 performs morphological analysis of the text data, the positive / negative classification unit 16 determines whether a sentence composed of each word has a negative meaning (negative for a product) or a positive (positive for a product). Classification) is performed using a classifier.
As this classifier, for example, a naive Bayes classifier is used. In addition, the classifier is made to learn in advance using a plurality of text data for teachers, assuming that the category is a positive category or a negative category.
Thereby, in this embodiment, it is determined whether the text of the text data has a positive content or a negative content.

In the first embodiment, it is determined whether a sentence of text data is seriously described. In other words, the questionnaire data can be used effectively for product development if the text data is written seriously, whether the content of the text is negative or positive for the product. There is no need to classify as much as possible.
However, when reflecting in product development and improvement, whether a sentence is negative or positive has an important meaning. For this reason, in this embodiment, in addition to the seriousness, it is determined whether the content described in the sentence is negative or positive.

According to the present embodiment, each of the questionnaire responses (sentences in the text box 550) determined to have a high degree of seriousness based on the text data is positive in the text content or negative in the text content. It is possible to classify the information, and when reflecting it in product development or improvement, etc., the content of positive text is used for new product development, and the information that uses negative text for product improvement is easily used properly. Can do.
Thereby, according to this embodiment, only the questionnaire result of the positive sentence is provided together with the attribute information of the user describing the questionnaire from the questionnaire result of the seriously written sentence, or the questionnaire result of the negative sentence Can be easily provided to the organization that requested the execution of the questionnaire together with the attribute information of the user describing the questionnaire.

  Further, a program for realizing each function of the questionnaire evaluation system 1 in FIG. 1 and the questionnaire evaluation system 1A in FIG. 6 is recorded in a computer-readable recording medium, and the program recorded in the recording medium is stored in the computer system. By reading and executing, the text data of the sentence written by the user is analyzed, the seriousness when the user's sentence is described is obtained from the analysis result, and the reward corresponding to this seriousness is given May be. Here, the “computer system” includes an OS and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design and the like within a scope not departing from the gist of the present invention.

DESCRIPTION OF SYMBOLS 1,1A ... Questionnaire evaluation system 11 ... Input control part 12 ... Language analysis part 13 ... Seriousness analysis part 14 ... Process selection part 15 ... Memory | storage part 16 ... Positive negative classification | category part 100 ... Information communication line 200 ... User terminal

Claims (2)

A seriousness analysis unit that performs natural language processing of text data of sentences described by a user in a questionnaire, and evaluates the degree of seriousness of the user in answering the questionnaire;
A questionnaire selection system comprising: a process selection unit that selects a process content to be provided to the user according to the degree. A process of performing natural language processing of text data of a sentence described by a user in a questionnaire, and evaluating a degree of seriousness of the user of an answer to the questionnaire;
And a process of selecting a processing content to be provided to the user in correspondence with the degree.

Images