JP2002373237A - Method and system for automatically generating questions of pluralistic questionnaire and unitarily evaluating questionnaire result - Google Patents

Abstract

PROBLEM TO BE SOLVED: To have respective questionnaire answerers answer without burdening them so much even when the total number of evaluation object items prepared in matching with a certain questionnaire purpose is extremely large and to derive significant unitary answer ranking accurately reflecting the mental evaluations of the questionnaire answerers for a lot of the evaluation object items by statistically processing pluralistic answers from a lot of answerers. SOLUTION: In an answer analysis process, n pieces of weighting presentation item sets stored in an answer data base are summed up, n-dimensional pluralistic ranking by the weighting of i pieces of the evaluation object items in each set is united on the basis of connection relations among n pieces of the presentation item sets and the unitary answer ranking is imparted to m pieces of the evaluation object items.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and a system for automatically generating a questionnaire of a pluralistic questionnaire and integrally evaluating a questionnaire result, and is applied to, for example, a marketing survey for measuring a brand value of a company. It is about technology.

[0002]

BACKGROUND OF THE INVENTION For example, it is assumed that a consumer who is a smoker decides on what kind of cigarette to purchase from a variety of cigarettes through a questionnaire. Department. In other words, a questionnaire is given to a large number of smokers, so that they can know what is important and what is considered as a motive for selecting products.

[0003] The questionnaire implementer lists items that may be motivated to select a product, such as "scent strength", "taste strength", "tar quantity", "CM image", and "package design". Ask the house a questionnaire that states, "Please choose the factors you value the most when choosing cigarettes." Here, items such as “scent intensity” and “taste intensity” that are asked by the respondent to be weighted and evaluated will be referred to as evaluation target items.
If the total number of items to be evaluated is about 10, these items are presented to the respondents in a list, and the most important items, 2
In many cases, the questionnaire objective can be achieved by having the respondents answer the items that matter the most. However, if the total number of evaluation target items prepared by the questionnaire performer for a certain questionnaire purpose exceeds 100, the method of presenting these items to the respondents in a list and having them weighted and evaluated is not The burden on consumers is too great to expect a truly meaningful answer in the general field of marketing research. Also,
Since the act of “scoring” the presented object does not exist in the daily life of consumers, the problem of the survey method itself of scoring the evaluation target item is also the problem in the past in the marketing field. Research.

[0004]

DISCLOSURE OF THE INVENTION Even if the total number of evaluation target items prepared for a certain questionnaire purpose is extremely large, individual questionnaire respondents can respond without too much burden, and multiple respondents can respond. By statistically processing such answers, a meaningful unified answer order that accurately reflects the psychological evaluation of the questionnaire respondents for many evaluation target items is derived. This is the object of the present invention.

The present invention to achieve the above object is a method for automatically generating a questionnaire of a pluralistic questionnaire and evaluating the questionnaire result in a unified manner, characterized by including the following processes. . (1) In the evaluation target item input process, a large number of evaluation target items to be unitarily evaluated by a questionnaire are prepared in the item database. The number of evaluation target items prepared in this process is m. (2) In the questionnaire design process, the number i of evaluation target items presented to the respondent in one unit question, the number n of unit questions forming one questionnaire, and each evaluation in a set of n unit questions The number of appearances j of the target item is appropriately set. (3) In the execution question set generation process, i is selected from the m evaluation target items prepared in the item database.
The number of evaluation target items is selected and set as presentation items of the unit question, and n presentation item sets are generated and prepared in the question database. At this time, n sets are set such that m evaluation target items appear j times in the n presentation item sets. In addition, when both sets in a case where a common item to be evaluated is included in a given presentation item set and another presentation item set are referred to as being in a connection relationship, the n presentation item sets are all linked by this connection relationship. n sets are made. (4) In the questionnaire implementation process, based on the n presented item sets prepared in the question database, a unit question for having the i items to be evaluated in each set weighted is given to the respondent. To present,
A weighted presentation item set is collected as a unit answer to each unit question, and these n unit answers are stored in an answer database. (5) In the answer analysis process, the n weighted presentation item sets accumulated in the answer database are totaled, and the n-dimensional multidimensional order by weighting the i evaluation target items in each set is n. Is unified based on the connection relation between the presented item sets described above, and a unified answer order is assigned to the m evaluation target items. (6) In the questionnaire result reporting process, the unified answer order given by the answer analysis process for the m evaluation target items is expressed and output in an appropriate format.

[0006] In the basic method invention having the above requirements, the following requirements (A) and (B) can be selectively added as needed. (A) In the implementation question set generation process, the item pairs for selecting two evaluation target items from the m evaluation target items appear in the n presented item sets an equal number of times. Make it a set. (B) In the answer analysis process, the unified answer order includes a numerical gap between the ranks.

Further, a second invention is a computer information processing system for carrying out the above method invention, comprising: means for managing the item database, the question database and the answer database; Execution means, execution means for the questionnaire design process, execution means for the implementation question set generation process,
The present invention is characterized by comprising: a means for executing the questionnaire implementation process; an means for executing the answer analysis process; and an means for executing the questionnaire result report process.

[0008] In this computer information processing system, the means for executing the questionnaire conducting process functions as a Web server that communicates with a Web browser device handled by the respondent. It is possible to adopt a method of collecting the n unit responses from the respondent Web browser device while transmitting the information to the device.

In the computer information processing system, the questionnaire design process executing means allows the questionnaire designer to input the number of options i, and based on the known total number m of the remaining two items, Is determined and presented to the questionnaire designer, and numerical values within the range are accepted.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS === Explanation of Terms === Here, the meanings and definitions of terms used in describing the present invention are listed. Evaluation items Items such as "scent intensity", "taste intensity", "tar quantity", and "CM image" in the example of the questionnaire for smokers described above Refers to each one. The evaluation target item can be expressed not only by text data but also by various multimedia information such as image data and audio data. Unit Question This refers to one question that presents some items to be evaluated to respondents and asks them to weight them. 1
The number of evaluation target items presented to the respondent in one unit question is referred to as the number of options i. Presentation item set Refers to a set of i evaluation target items presented to respondents in one unit question. Evaluation target item set Refers to the entire set of evaluation target items prepared for a certain questionnaire purpose. The number of evaluation target items constituting this set is referred to as a total item number m. Implemented question set Refers to a set of multiple unit questions that are actually transmitted to respondents in accordance with a single questionnaire purpose. The number of unit questions constituting this set is referred to as the total number of questions n. Unit answer (weighted presentation item set) Refers to a weighted presentation item set acquired as an answer to one unit question. Answer result set Refers to the entire set of a large number of unit answers obtained from the implementation question set.

=== Overview of Questionnaire Apparatus === FIG. 1 shows an overview of a computer information processing system (hereinafter referred to as a questionnaire apparatus) according to an embodiment of the present invention.
Is shown in the block diagram of FIG. This questionnaire device is constructed by one computer or as a distributed processing system by a plurality of computers. This questionnaire device creates and manages an item database, a question database, an answer database, and a report database.

=== Evaluation Target Item Input Process === A large number of evaluation target items prepared for a certain purpose are accumulated in the item database. Each item is assigned a unique item ID. The number of evaluation target items prepared for one purpose is referred to as a total item number m. In the item database, a plurality of groups of items to be evaluated prepared for a plurality of different questionnaire purposes are separately stored. The questionnaire implementer inputs an evaluation target item set to the questionnaire device in an online method or an offline method and stores the set in the item database. This is the evaluation target item input process.

=== Overview of Questionnaire Design Process ==
= In the questionnaire design process, the number of evaluation target items presented to the respondent in one unit question (this is the number of options i) and the number of unit questions constituting one questionnaire (this is Is determined), and the number of appearances j of each evaluation target item in the set of n unit questions is determined. The questionnaire apparatus prompts a questionnaire implementer or an assistant to select the number of options i, and presents the other two parameter selection candidates to the operator based on the value of the parameter i input by the operator. Then, the final four parameters are determined. It should be noted that a detailed example of this part will be described in the section of “Example of questionnaire design process”.

A simple example will be described briefly. For example, it is assumed that a questionnaire is used to determine from what point of view a consumer who is a smoker decides a product to purchase from among many types of cigarettes. The questionnaire implementer enumerates items considered to be motivation for product selection, such as “fragrance intensity”, “taste intensity”, “tar quantity”, “CM image”, “package design”, and uses the questionnaire device. To enter. Assume that the total number m of items in this example is 136.

[0015] Then, as shown in FIG. 2, four evaluation target items are selected from among the unit questions for asking the respondent, "Please select the following items from the following items when you choose cigarettes." Shall be presented as That is, the number of options i is set to four. At this time, the questionnaire apparatus calculates the total number of questions n and the number of appearances j based on the total number of evaluation target items 136 and the number of options 4 (n = 1530, j
= 45) or (n = 3060, j = 90). The meaning is as follows. For example, a total of 1530 unit questions have a total of 153
0 × 4 = 6120 items to be evaluated are included, but 136
The setting is such that any one of the types of evaluation target items appears in 45 unit questions.

=== Working Question Set Generation Process === In the working question set generation process, i evaluation target items out of m evaluation target items prepared in the item database are presented as unit question presentation items. It is selected and set, and n presentation item sets are generated and prepared in the question database. At this time, n evaluation target items appear j times in the n presentation item sets.
Make a set of pieces. In addition, when both sets in a case where a common item to be evaluated is included in a given presentation item set and another presentation item set are referred to as being in a connection relationship, the n presentation item sets are all linked by this connection relationship. n
Make a set of pieces. Furthermore, all pairs (combinations of two items) that can be selected from m kinds of evaluation target items
Cause the same number of times (assumed to be the number of pairings k) to appear in n unit questions. In the system configuration diagram of FIG. 1, the means for executing the implementation question set generation process is expressed as a question generation engine. A specific example of the algorithm executed by the question generation engine will be described mathematically.
This algorithm is shown in the flowchart of FIG.

[0017]

=== Questionnaire Implementation Process === In the questionnaire implementation process, each one based on the n presented item sets prepared in the question database.
In addition to presenting to the respondent a unit question for weighting the i evaluation target items in the set, a presentation item set weighted as a unit answer to each unit question is collected, and these n unit responses are collected. Is stored in the answer database.

In the embodiment shown in FIG. 1, the questionnaire implementation process is executed by the Web engine, Web server, and collection engine in the questionnaire apparatus.
Respondents communicate with the Web server using a browser terminal such as a personal computer. The Web server creates, for example, a unit question Web document as shown in FIG. 2 based on the n presentation item sets stored in the question database, and delivers the Web document to the respondent browser terminal. The respondent looks at the Web document of the unit question displayed on the browser terminal and weights the presented i evaluation target items. In the example of the unit question shown in FIG. 2, four items are presented, and
Select one and mark it and send it back to the web server. Thus, the presentation item set marked with one of the four items is the above-described weighted presentation item set, that is, the unit answer. The web-based engine, web server, and collection engine communicate with one or many respondents' browser terminals, send out n unit questions created based on n presented item sets, and weight the respondents The n unit answers (weighted presentation item sets) are collected and stored in the answer database.

=== Overview of Answer Analysis Process === In the answer analysis process, n weighted presentation item sets accumulated in the answer database are totaled, and i evaluation target items in each set are calculated. The n-dimensional multiple order based on the weighting is unified based on the connection relationship between the n presentation item sets, and the unified answer order is assigned to the m evaluation target items. In the system configuration diagram of FIG. 1, the means for executing the answer analysis process is expressed as an answer analysis engine.

In the example of the unit question shown in FIG. 2, four items to be evaluated are presented to the respondent, and one of them is marked. The four items presented here are A, B,
Assume that C and D are set, and the respondent marks A, for example. This unit response indicates a sequence of A> B, A> C, and A> D (this is referred to as a unit response sequence). Further, it is assumed that four items B, E, F, and G are presented in another unit question, and the respondent has marked B. This unit answer shows the order of B> E, B> F, B> G (this is also the unit answer order). In this way, an n-dimensional multiple response order is obtained from the n unit responses.

The two presented item sets (A, B,
C, D) and the presentation item set (B, E, F, G) are connected by a common evaluation target item B. Based on this connection relationship and the obtained answer order A> B, A> E,
A consistent order of A> F, B> G can be derived for the time being. The order derived indirectly from the plurality of unit answer orders directly obtained from the respondent in this way is referred to as a guide order.

According to the above principle, when deriving a guidance sequence from a unit response sequence based on the connection relation of n weighted presentation item sets (unit responses), a guidance sequence derived from a certain unit response sequence group and another May be inconsistent with the guidance sequence derived from the group of unit response sequences. In general questionnaires, such inconsistencies occur at a considerable frequency. For example, a guide sequence of A> E is derived from a certain group of unit response sequences, and E> A from another unit response sequence group.
May be derived. When contradictions occur frequently in this way, it is recognized that the distance between the item A and the item E (the distance between the unitary orders described above) is small. If there is no inconsistency in the order of A> B, it is recognized that the distance between item A and item B is large. In this way, the unified answer rank is given to the m evaluation target items, and the distance between each rank in the rank is quantified.

=== Details of the Answer Analysis Process === (A) The principle of giving a unified answer order can be explained as follows. As described above, when n weighted presentation item sets (unit answers) are stacked,
Often includes inconsistencies due to inversion of choice. This contradiction is caused by a change in the score of each item in the respondent's recognition process. At this time, a quantitative calculation of the psychological score is performed using the fact that it is possible to determine that the psychological scores of the evaluation target items are closer to each other as the contradiction is more likely to occur.

(A) In performing this quantitative calculation, the change in the score of the evaluation target item in the process of recognizing the respondent is as follows.
The change based on the psychological score of the respondent is set as a precondition. For example, the variation may be based on the normal distribution centered on the psychological score of the respondent, may be based on a Gaussian distribution, or the like. From a psychological point of view, it is common to assume that the distribution is based on a normal distribution.

(C) Based on the premise of the change of the score in the process of recognizing the respondent described in (a) above, based on the answer result set created from the weighted presentation item set by the respondent, [Appearance probability of comparison information between evaluation target items in the group] is [Appearance frequency of the comparison information]
The most probable psychological score that maximizes the “likelihood of the answer result set” obtained by multiplying the factorial by the total comparison information between the evaluation target items is inversely calculated.

(D) As a calculation method, a simulation calculation as described below is used instead of an analysis operation in order to make quick calculation by utilizing the performance of a computer. FIG. 4 shows the algorithm of this simulation calculation. The answer analysis engine in FIG. 1 executes this algorithm. Set up (z + 1) (generally, z = about 10) initial coordinates that form coordinates in an m-dimensional space. The coordinates of the least likely point among the (z + 1) coordinates are symmetrically moved around the center of gravity of the other coordinates. If the coordinates after the symmetric movement are most likely compared to the other z coordinates, the more likely coordinates among the coordinates and the coordinates obtained by further symmetrically moving the center of gravity around the coordinates are determined. New coordinates are set, and coordinates (original points) that are least likely to be removed are removed to obtain new (z + 1) points. If the coordinates after the symmetric movement are less likely than the most likely coordinates among the other z coordinates, and are more likely than the least likely coordinates, the symmetrically moved coordinates are regarded as new coordinates; The least likely coordinates (original coordinates) are removed to obtain new (z + 1) coordinates. If the function value derived by the coordinates after the symmetrical movement is still less likely than any of the other z coordinates, the function value is calculated based on a predetermined constant coefficient until it is no longer the least likely coordinate. Move toward the center of gravity of the z coordinates. The position at the time when the coordinates are no longer the least likely is set as new coordinates, and the least likely coordinates (original coordinates) are removed to obtain a new (z + 1) combination of coordinates.

By repeating the above steps, (z +
1) The least likely coordinates and the most likely coordinates among the coordinates are repeated until the coordinates fall within a preset allowable error range condition. When the two coordinates fall within the allowable error range condition, the most likely coordinate is set as the optimum coordinate.

(E) The estimated maximum value calculation coordinates are used as an initial value, and are repeated a preset number of times (about 10 times) to execute the maximum value calculation process (d). The m-dimensional coordinate values of the final optimal coordinates are used as the psychological scores of the subject group for the m evaluation target items. The coordinates are output and stored in the report database.

=== Questionnaire Result Reporting Process === In the questionnaire result reporting process, a unified response sequence (stored in the report database) assigned to the m evaluation target items by the response analysis process.
Is expressed in an appropriate format and output. FIG. 5 exemplifies the format of the report document of the questionnaire result. FIG. 5A is a representation of a list of unit response ranks, and FIG.
Is a report document of the unified answer rank given by the answer analysis process. This report output may be performed by a display or a printer of the questionnaire apparatus, or a computer of a remote questionnaire requester and the questionnaire apparatus may communicate via the Internet or the like and report using a Web mechanism.

=== Example of Questionnaire Design Process =
== A specific algorithm and mathematical method of the questionnaire design process will be described with reference to the flowchart of FIG. The operator first inputs an evaluation target item to the questionnaire device. The questionnaire apparatus records the input evaluation target items, and records the total number of evaluation target items as a parameter m. Further, the operator is prompted to input the number of options (parameter i). The operator sets the parameter i
Is input, based on the conditions described in FIG. 6, (n, i,
m, j) to the user and final (n, j)
i, m, j). The user determines and inputs (n, i, m, j) based on the presented candidates. The conditions described in FIG. 6 are examples of conditions required for (n, i, m, j) when creating a presentation item set that also satisfies requirement (A) under the conditions of (m, i). is there.

[0032]

The present invention relates to a method and a system for automatically generating a questionnaire of a multiple questionnaire and integrally evaluating the questionnaire result, and can be applied to a very wide range of specific applications. In explaining the effect of the present invention, a use of measuring a brand value of a company will be described. A company's brand value is determined by its reputation in the minds of its intended customers. Accurately measuring customer sentiment is fundamentally very difficult, so there is no absolutely superior way to measure a company's brand value, and there are a variety of methods. Since the brand value is determined within the psychology of the customer, it is common for the company to know the psychological score for the brand through questioning of the target customers and calculate the brand value. In particular, if you want to quantitatively measure psychological scores for various qualitative factors that make up brand value,
Present all the items to be compared to the subject at the same time, or present one item at a time in order, and let them freely answer the "degree of importance" or "rank" or select from the prepared score group Studies have found it a quantitative measure of the qualitative factor of brand power.

However, with this method, it is difficult to quantitatively detect a difference in psychological scores among survey items that are relatively important to the subject among given items. this is,
This is because it is difficult for the subject to be aware of the quantitative difference in the psychological score between the items that are emphasized more than a certain degree when comparing the items. From a psychological standpoint, from the psychological point of view, even if each survey item is evaluated as a psychological score having a certain value with respect to the evaluation axis in the subject's internal psychology, these psychological scores are not Is not always recognized as a stable value, and is considered to be caused by the fact that the value is changed and recognized with a certain width.
In other words, the evaluation score on the relevant evaluation axis for each survey item, which is the basis for comparison by the subject when taking the comparative test, is a result of a change in the psychological score inside the subject.

Therefore, when it is desired to accurately analyze the results of the comparative survey, the obtained answer results are based on the evaluation scores of the results that fluctuate within the psychology of the subject, and often the comparison results are reversed. It should be considered that there are also. This fact does not pose a significant problem when the difference between the subject's evaluations for a plurality of survey items is large or when the purpose is to measure rough evaluations, because the influence of the fluctuation is reduced. However, this is a problem to be solved when it is desired to detect / measure differences in evaluation of a plurality of survey items in which the difference in evaluation between subjects is not so large.

Measuring the difference in psychological scores between qualitative factors that are relatively important to customers for a company's brand strategy is very important in planning a company's brand strategy, which requires narrowing down the appeal. Importantly, there was a need for a method and apparatus for appropriately measuring psychological scores. The method and system of the present invention can meet such a demand.

According to the present invention, even if the total number of evaluation target items prepared for a certain questionnaire purpose is extremely large, individual questionnaire respondents can answer without much burden, By statistically processing the multiple responses from, it is possible to derive a meaningful and unified response rank that accurately reflects the psychological evaluation of the questionnaire respondents for a large number of evaluation target items.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a computer information processing system according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a unit question Web document according to the present invention.

FIG. 3 is a flowchart illustrating an example of an algorithm of an implementation question set generation process according to the present invention.

FIG. 4 is a flowchart illustrating an example of an algorithm of a response analysis process according to the present invention.

FIG. 5 is a diagram showing an example of a report document of a questionnaire result according to the present invention.

FIG. 6 is a flowchart illustrating an example of an algorithm of a questionnaire design process according to the present invention.

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Masahiro Aoyagi 3-2-8 Mita, Minato-ku, Tokyo Net2 Mita Building 5F Inside Mindshare F-term (reference) 5B056 BB71 HH00 5B082 GA08 HA08

Claims (6)

[Claims] 1. A method of automatically generating a question of a multiple questionnaire and evaluating a questionnaire result in a centralized manner, characterized by including the following processes. (1) In the evaluation target item input process, a large number of evaluation target items to be unitarily evaluated by a questionnaire are prepared in the item database. The number of evaluation target items prepared in this process is m. (2) In the questionnaire design process, the number i of evaluation target items presented to the respondent in one unit question, the number n of unit questions forming one questionnaire, and each evaluation in a set of n unit questions The number of appearances j of the target item is appropriately set. (3) In the execution question set generation process, i is selected from the m evaluation target items prepared in the item database.
The number of evaluation target items is selected and set as presentation items of the unit question, and n presentation item sets are generated and prepared in the question database. At this time, n sets are set such that m evaluation target items appear j times in the n presentation item sets. In addition, when both sets in a case where a common item to be evaluated is included in a given presentation item set and another presentation item set are referred to as being in a connection relationship, the n presentation item sets are all linked by this connection relationship. n sets are made. (4) In the questionnaire implementation process, based on the n presented item sets prepared in the question database, a unit question for having the i items to be evaluated in each set weighted is given to the respondent. To present,
A weighted presentation item set is collected as a unit answer to each unit question, and these n unit answers are stored in an answer database. (5) In the answer analysis process, the n weighted presentation item sets accumulated in the answer database are totaled, and the n-dimensional multidimensional order by weighting the i evaluation target items in each set is n. Is unified based on the connection relation between the presented item sets described above, and a unified answer order is assigned to the m evaluation target items. (6) In the questionnaire result reporting process, the unified answer order given by the answer analysis process for the m evaluation target items is expressed and output in an appropriate format. 2. The method according to claim 1, wherein in the execution question set generation process, all item pairs that select two evaluation target items from m evaluation target items are n presentation items. It is characterized in that the above-mentioned setting is performed so that it appears in the item set an equal number of times. 3. The method according to claim 1, wherein the answer analysis process includes a numerical gap between the ranks as the unified answer order. 4. A computer information processing system for implementing the method according to claim 1, wherein said item database, said question database, and said answer database are managed, and said evaluation item is provided. Means for executing an input process, means for executing the questionnaire design process, means for executing the question set generation process, means for executing the questionnaire process, means for executing the answer analysis process, and the questionnaire result report process And execution means. 5. The computer information processing system according to claim 4, wherein the means for executing the questionnaire implementation process functions as a Web server that communicates with a Web browser apparatus handled by the respondent, and transmits n unit questions. We
It is characterized in that it is transmitted to the respondent Web browser device in the form of a b-document, and that n unit responses are collected from the respondent Web browser device. 6. The computer information processing system according to claim 4, wherein the means for executing the questionnaire design process allows the questionnaire designer to input the number i of options, and based on the known total number m of items. , The remaining two
It is characterized in that a settable range of two numerical values is obtained and presented to a questionnaire designer, and a numerical value input within the range is accepted.