JP2004086350A - Text information analysis system and presentation method of analysis result - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a text information analysis system used for presenting a plurality of results by analysis techniques different from one another by systematically combining them with one another. <P>SOLUTION: A knowledge analysis part 12 executes, for text information stored in a knowledge database 13, an analysis by a clustering part 121 and an analysis by a text mining part 122, and stores the results thereof (a clustering result 141 and a text mining result 142) in an analysis result storing database 14. A user interface part 11 receives, from a user, selections of a cluster of the clustering result 141 disposed on the longitudinal axis and the lateral axis, and of a category of the text mining result 142 by an analysis axis selection part 111, and carries out counting of the analysis results based on the selections by an analysis result counting part 112 to present the counting result to the user. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a text information analysis system for analyzing text information such as questionnaires and daily reports collected and stored via a LAN (Local Area Network) or an intranet, and a method of presenting analysis results applied to the system.
[0002]
[Prior art]
2. Description of the Related Art In recent years, an increasing number of companies have laid a LAN or an intranet to collect and accumulate information held by each employee, for example, irregular information such as a questionnaire and a daily report generated during work, across departments. This collected and accumulated information is generally subjected to various analyzes in order to be shared and utilized as knowledge of all employees. At present, clustering analysis and text mining analysis are well known as the analysis methods.
[0003]
In the clustering analysis, as described in, for example, JP-A-2002-149670, information collected and stored is classified according to the appearance frequency of each word and the degree of association between a plurality of words. Here, the degree of relevance between a plurality of words refers to the presence or absence of co-occurrence. For example, there is a lot of text information including both “A” and “B” such as “I purchased A and B.” In this case, it is determined that “A” and “B” have co-occurrence.
[0004]
As a result, not only information with a high frequency of appearance of the word "A" is treated as belonging to the same cluster, but also information with a high frequency of appearance of the word "B" is treated as belonging to the same cluster. Is performed automatically and the classification with high accuracy is performed automatically.
[0005]
On the other hand, the text mining analysis classifies the collected and accumulated information into categories desired by the user, as described in, for example, JP-A-2001-147937. For example, when it is desired to group information on “C”, “D”, and “F” products into respective categories, the user may determine what kind of description is included in the category and determine whether the information belongs to each category. Specify conditions.
[0006]
According to the clustering analysis and the text mining analysis, knowledge sharing and utilization can be effectively achieved, for example, it is possible to grasp a certain tendency from a large amount of information collected and accumulated in a random manner.
[0007]
[Problems to be solved by the invention]
By the way, both the clustering analysis and the text mining analysis described above have a disadvantage that a large amount of information that does not belong to any cluster or category is generated. Therefore, no matter which analysis method is employed, extremely important information cannot be extracted, and may be buried in a large number of other information.
[0008]
Even if both analysis methods are provided, simply referring to the analysis results individually, for example, recognizes, in the other analysis, a tendency that targets only the information buried in one analysis. It's difficult, and you can't expect so-called synergy.
[0009]
The present invention has been made in view of such circumstances, and a text information analysis system that organically combines and presents a plurality of analysis results obtained by different analysis techniques and an analysis result analysis method applied to the text information analysis system. The purpose is to provide a presentation method.
[0010]
[Means for Solving the Problems]
In order to achieve the above-mentioned object, the present invention provides a text information analysis system for analyzing a large amount of text information collected and stored, based on the frequency of appearance of each word and the degree of association between a plurality of words. Clustering analysis means for analyzing a large amount of accumulated text information; text mining analysis means for analyzing the large amount of collected and accumulated text information based on arbitrarily specified conditions; and An analysis result presentation unit that organically combines and presents the analysis result of the clustering analysis unit and the analysis result of the text mining analysis unit is provided.
[0011]
In the text information analysis system according to the present invention, the results of the two analysis are presented, for example, by presenting the result of the clustering analysis and the result of the text mining analysis in the form of a two-dimensional array assigned to the vertical and horizontal axes, respectively. Are presented organically bound. As a result, for example, it is possible to easily grasp, using the other analysis results, the tendency for only the information that has been buried in one analysis result. Realize the presentation.
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0013]
FIG. 1 is a diagram showing a network configuration of a knowledge analysis system according to an embodiment of the present invention.
[0014]
The knowledge analysis system 1 is built on a high-performance computer called a server or the like, and is connected to a plurality of client computers 2 via a network 3 such as a LAN or an intranet. Then, the knowledge analysis system 1 receives an analysis request from the client computer 2 and returns an analysis result based on the request.
[0015]
FIG. 2 is a diagram showing functional blocks of the knowledge analysis system 1. As shown in FIG. 2, the knowledge analysis system 1 has a processing unit of a user interface unit 11 and a knowledge analysis unit 12, and a data unit of a knowledge database 13 and an analysis result storage database 14. The processing unit is configured by a program that describes the operation procedure of a CPU mounted on a computer on which the knowledge analysis system 1 is constructed, and the data unit is a storage unit such as a magnetic disk device included in the computer. It is configured on a medium.
[0016]
The user interface unit 11 serves as a contact point for the user of the client computer 2, and has an analysis axis selection unit 111 and an analysis result totaling unit 112. The analysis axis selection unit 111 receives a part of the instruction from the client computer 2, and the details will be described later. On the other hand, the analysis result totaling unit 112 totals the analysis results based on the instruction received by the analysis axis selecting unit 111 and returns the result to the client computer 2. This will be described later in detail.
[0017]
The knowledge analysis unit 12 analyzes a large amount of text information accumulated in the knowledge database 13 such as a questionnaire and a daily report generated in business, and stores the analysis result in the analysis result storage database 14. The clustering unit 121 And a text mining unit 142. The clustering unit 121 classifies the text information of the knowledge database 13 into clusters based on the appearance frequency of each word and the degree of association between a plurality of words (clustering analysis). The clustering result 141 obtained by this is analyzed as a result of analysis. It is stored in the storage database 14. On the other hand, the text mining unit 142 classifies the text information in the knowledge database 13 into categories desired by the user based on the conditions specified by the user (text mining analysis), and obtains the text mining result obtained by this. 142 is stored in the analysis result storage database 14.
[0018]
Here, an outline of a method of presenting an analysis result, which is a feature of the knowledge analysis system 1, will be described with reference to FIGS.
[0019]
Now, it is assumed that a large amount of text information such as a questionnaire, a daily report, and an e-mail is stored in the knowledge database 13 (A in FIG. 3). Then, on the one hand, the clustering unit 121 performs a clustering analysis on the same text information group to obtain a clustering result 141 (B2 in FIG. 3), and on the other hand, the text mining unit 122 performs the text mining analysis. It is assumed that the text mining result 142 is obtained by the execution (B1 in FIG. 3).
[0020]
First, paying attention to the clustering result 141, the text information group is classified as C1, C2, C3,..., But a large amount of text information that does not belong to any of them is generated. Similarly, focusing on the text mining result 142, the text information group is classified as T1, T2, T3,..., But a large amount of text information that does not belong to any of them is generated. Therefore, as it is, text information that does not belong to any of them will be simply buried together with a large number of "other" text information.
[0021]
Therefore, in the knowledge analysis system 1, the two analysis results are organically connected, and more specifically, for example, a two-dimensional array in which the clustering result 141 is assigned to the vertical axis and the text mining result 142 is assigned to the horizontal axis. The data is tabulated and presented to the user (C in FIG. 3). In the figure, n11 indicates the number of pieces of text information that belong to the cluster C1 by the clustering analysis by the clustering unit 121 and belong to the category T1 by the text mining analysis by the text mining unit 122.
[0022]
Thereby, for example, the text information group grouped as “others” in the clustering result 141 can be referred to by the classification of T1, T2, T3,... Of the text mining result 142 (nx1, nx2, nx3,. Similarly, in the text mining result 142, the text information group summarized as “others” can be referred to by the classification of C1, C2, C3,... In the clustering result 142 (n1y, n2y, n3y,. ). By organically combining two analysis results from different viewpoints, a so-called synergistic effect can be expected, such as prompting a new discovery that cannot be obtained from only one analysis result.
[0023]
The clustering result 141 obtained by the clustering analysis of the clustering unit 121 and the text mining result 142 obtained by the text mining analysis of the text mining unit 142 classify the text information group into multi-level clusters or categories. Is common. FIG. 5 shows an example of a text mining result 142 classified into a multi-layer category. Therefore, in this knowledge analysis system 1, the cluster and category hierarchies of the clustering result 141 and the text mining result 142 arranged as items on the vertical axis and the horizontal axis can be moved up and down for each axis according to the user's instruction. I did it.
[0024]
For example, in the table (C) shown in FIG. 3, in response to a request to refer to the text mining result 142 in more detail by focusing on T1, the knowledge analysis system 1 uses the horizontal axis as shown in FIG. The category of the category of the text mining result 142 arranged as an item is re-aggregated and presented to be moved down one level. This movement is possible as long as the hierarchy continues, and conversely, movement from bottom to top is also possible.
[0025]
Next, with reference to FIGS. 6 to 10, an operation principle when the knowledge analysis system 1 presents an analysis result will be described.
[0026]
When providing the knowledge analysis service to the client computer 2 connected via the network 3, the user interface unit 11 first transmits screen data for displaying the screen shown in FIG. The screen includes a button a1 for instructing execution of text mining analysis, a button a2 for instructing execution of clustering analysis, a button a3 for shifting to an analysis axis selection operation, and an analysis result after selecting the analysis axis. And a button a4 for starting tallying. The user who has received this screen operates a pointing device such as a mouse provided in the client computer 2 and selects a desired button.
[0027]
When the selection of the button a1 is notified, the user interface unit 11 instructs the knowledge analysis unit 12 to execute the text mining analysis. On the other hand, in response to the instruction, the knowledge analysis unit 12 performs the text mining analysis on the latest text information group accumulated in the knowledge database 13 by the text mining unit 122, and the analysis result, that is, the text mining result 142 Is stored in the analysis result storage database 14.
[0028]
Similarly, when the selection of the button a2 is notified, the user interface unit 11 instructs the knowledge analysis unit 12 to execute the clustering analysis. On the other hand, in response to the instruction, the knowledge analysis unit 12 performs the text mining analysis on the latest text information group stored in the knowledge database 13 by the clustering unit 121, and analyzes the analysis result, that is, the clustering result 141. The result is stored in the result storage database 14.
[0029]
Further, when the selection of the button a3 is notified, the user interface unit 11 transmits screen data for displaying the screen shown in FIG. On this screen, a button b1 for shifting to a selection operation on the vertical axis of the table to which the clustering result 141 is assigned, and a button b2 for shifting to a selection operation on the horizontal axis of the table to which the text mining result 142 is assigned. It is additionally arranged. Then, when the selection of the button b1 or the button b2 is notified, the user interface unit 11 starts the notified analysis axis selection processing.
[0030]
Now, assuming that the selection of the button b1 has been notified, the user interface unit 11 causes the analysis axis selection unit 111 to acquire the hierarchical structure of the cluster in the clustering result 141 stored in the analysis result storage database 14. Then, the user interface unit 11 creates and transmits screen data for displaying a screen showing the hierarchical structure of the acquired cluster. FIG. 8 illustrates a screen presented to the user at this time.
[0031]
In the example of FIG. 8, in the hierarchical structure of the cluster in the clustering result 141, C1, C2,... Exist in the uppermost layer, and C11, C12,. Further, C111, C112, C113, C114, and C115 exist in a layer immediately below C11. Then, when the user who has been presented with this screen selects C11 from among them, the user operates a pointing device such as a mouse provided in the client computer 2 and selects the button c1 with C11 selected. On the other hand, the user interface unit 11 notified of the selection of C11 edits and transmits screen data for displaying the screen shown in FIG. As shown in FIG. 9, it is indicated that C11 selected by the user has been selected as the vertical axis of the table (d1 in FIG. 9).
[0032]
Similarly, the user selects the button a3 and the button b2, and performs the operation of selecting the horizontal axis of the table to which the text mining result 142 is assigned. Then, after the work is completed, the user selects the button a4 to start counting the analysis results.
[0033]
When notified of the selection of the button a4, the user interface unit 11 causes the analysis result totaling unit 112 to perform totalizing for organically combining the clustering result 141 and the text mining result 142.
[0034]
The clustering result 141 and the text mining result 142 stored in the analysis result storage database 14 include information for identifying which text information belongs to each cluster and each category. Therefore, by matching this information, the number of text information belonging to both the arbitrary cluster of the clustering result 141 and the arbitrary category of the text mining result 142 can be totaled. By performing such matching, the analysis result totaling unit 112 totalizes the analysis results in which the clustering result 141 is assigned to the vertical axis and the text mining result 142 is assigned to the horizontal axis. Then, the user interface unit 11 creates and transmits screen data for displaying a screen that presents the analysis results compiled by the analysis result compilation unit 112. FIG. 10 illustrates a screen presented to the user at this time.
[0035]
As shown in FIG. 10, the vertical and horizontal clusters and categories selected by the user are displayed at the top of the screen (e1). Here, the cluster C1 is selected on the vertical axis of the table, and the category T11 is selected on the horizontal axis of the table. Based on this selection, clusters C11, C12, C13,... Of the next lower layer of the cluster C1 are arranged in the center of the screen as items on the vertical axis, and the category of the next lower layer of the category T11 is arranged. The clustering result 141 and the text mining result 142 totaled in a table format in which T111, T112, T113, T114, T115,... Are arranged as items on the horizontal axis are displayed (E2). This table can be scrolled downward and to the right, respectively. At the end, the number of text information items that do not belong to any of the clusters and categories is totaled and displayed.
[0036]
When one of the clusters arranged as the items on the vertical axis and the categories arranged as the items on the horizontal axis is selected, the cluster or category in the hierarchy one level below the selected cluster or category is displayed on each axis. , The clustering result 141 and the text mining result 142 are re-aggregated and displayed (drill-down). For example, if cluster C12 is selected, the vertical axis is replaced with clusters C121, C122,... And the number of records in the table is updated.
[0037]
Further, at the bottom of the screen, a button for moving (drilling up) a cluster arranged as an item on the vertical axis or a hierarchy of a category arranged as an item on the horizontal axis to the next higher cluster or category is arranged. (E3). For example, when it is instructed to drill up the horizontal axis of the table in the state of FIG. 10, the categories arranged as the items of the horizontal axis are replaced with categories T11, T12, T13,... And the number of cases in the table is also updated. Is done.
[0038]
FIG. 11 is a flowchart showing an operation procedure when the knowledge analysis system 1 presents an analysis result.
[0039]
First, the user interface unit 11 transmits screen data for displaying a task selection screen for the user of the client computer 2 to select an operation (step A1). Next, when the user who has received the display of this screen selects “selection of analysis axis” (YES in step A2), the user interface unit 11 makes a selection for selecting one of the vertical axis and the horizontal axis. The image data for displaying the screen is transmitted (step A3). Then, when the user who receives the display of this screen selects “vertical axis” (YES in step A4), the user interface unit 11 executes a cluster selection process using the analysis axis selection unit 111 (step S4). A5) If “horizontal axis” is selected (NO in step A4), a category selection process is performed using the analysis axis selection unit 111 (step A6).
[0040]
When “start analysis” is selected (NO in step A2, YES in step A7), the user interface unit 11 uses the analysis result totaling unit 112 to perform a totaling process using the selected cluster and category as an analysis axis. Is executed (step A8), and the image data for displaying the analysis result screen presenting the tally result is transmitted (step A9).
[0041]
Further, when a hierarchical movement of the analysis axis is instructed on the analysis result screen (YES in step A10), the user interface unit 11 uses the analysis result totaling unit 112 to set the moved cluster and category as the analysis axis. The counting process is executed again (step A8).
[0042]
According to the above procedure, the knowledge analysis system 1 presents the clustering result 141 of the clustering unit 121 and the text mining result 142 of the text mining unit 122 in an organically combined manner, and furthermore, the hierarchy of the cluster or category to be analyzed. Is moved up and down according to the instruction. This enables the so-called synergistic effect, for example, to allow information in one analysis to grasp trends in the information buried in the other analysis, and to promote new discoveries that cannot be obtained from the results of only one analysis. You can also.
[0043]
Here, as an example of a method of organically combining two analysis results having different viewpoints, an example in which the results are tabulated in a two-dimensional array is shown. However, the present invention is not limited to this, and the present invention is not limited to this. Any form can be applied as long as the relationship can be expressed.
[0044]
Also, here, as shown in FIG. 4, the description has been made on the assumption that the analysis results are arranged in multiple hierarchies. However, this is not essential, and it is necessary to force a plurality of classification viewpoints into one hierarchy. Does not force. A plurality of classification viewpoints can be treated as independent flat classification systems, and can be organically combined using, for example, two axes in a table.
[0045]
That is, the present invention is not limited to the above-described embodiment, and can be variously modified in an implementation stage without departing from the gist of the invention. Furthermore, the embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some components are deleted from all the components shown in the embodiment, the problem described in the column of the problem to be solved by the invention can be solved, and the effects described in the column of the effect of the invention can be solved. Is obtained, a configuration from which this configuration requirement is deleted can be extracted as an invention.
[0046]
【The invention's effect】
As described above, according to the present invention, it is possible to provide a text information analysis system that organically combines and presents a plurality of analysis results obtained by different analysis methods and a method of presenting the analysis results applied to the system. It becomes possible.
[Brief description of the drawings]
FIG. 1 is a diagram showing a network configuration of a knowledge analysis system according to an embodiment of the present invention.
FIG. 2 is an exemplary view showing functional blocks of the knowledge analysis system according to the embodiment;
FIG. 3 is an exemplary first view for explaining an outline of a method of presenting an analysis result executed by the knowledge analysis system of the embodiment.
FIG. 4 is an exemplary second diagram for describing an outline of a method of presenting an analysis result executed by the knowledge analysis system of the embodiment.
FIG. 5 is an exemplary third diagram illustrating an outline of a method of presenting an analysis result executed by the knowledge analysis system according to the embodiment;
FIG. 6 is an exemplary first diagram illustrating a screen displayed by the knowledge analysis system according to the embodiment;
FIG. 7 is an exemplary second diagram illustrating a screen displayed by the knowledge analysis system according to the embodiment;
FIG. 8 is an exemplary third diagram illustrating a screen displayed by the knowledge analysis system according to the embodiment;
FIG. 9 is an exemplary fourth diagram illustrating a screen displayed by the knowledge analysis system according to the embodiment;
FIG. 10 is an exemplary fifth diagram illustrating a screen displayed by the knowledge analysis system of the embodiment.
FIG. 11 is an exemplary flowchart illustrating the operation procedure when the knowledge analysis system according to the embodiment presents an analysis result.
[Explanation of symbols]
1 ... Knowledge analysis system 2 ... Client computer 3 ... Network 11 ... User interface 12 ... Knowledge analysis unit 13 ... Knowledge database 14 ... Analysis result storage database 111 ... Analysis axis selection unit 112 ... Analysis result totaling unit 121 ... Clustering unit 122 ... Text Mining unit 141 Clustering result 142 Text mining result

Claims (15)

In a text information analysis system that analyzes a large amount of collected and stored text information,
First and second analysis means;
A text information analysis system, comprising: an analysis result presenting means for organically combining and presenting two analysis results of the first and second analysis means for the same text information group. In a text information analysis system that analyzes a large amount of collected and stored text information,
Clustering analysis means for analyzing the collected and accumulated large amount of text information based on the appearance frequency of each word and the degree of association between the plurality of words;
Text mining analysis means for analyzing the large amount of collected and accumulated text information based on arbitrarily specified conditions,
A text information analysis system, comprising: an analysis result presentation unit that organically combines and presents the analysis result of the clustering analysis unit and the analysis result of the text mining analysis unit for the same text information group. The text information analysis system according to claim 2, wherein the condition specified by the text mining analysis means is a condition for classifying the text information group into a desired category. The analysis result presentation means presents the analysis result of the clustering analysis means and the analysis result of the text mining analysis means in a two-dimensional array tabular form assigned to a vertical axis and a horizontal axis, respectively. Item 2. The text information analysis system according to Item 2. The clustering analysis means and the text mining analysis means classify the text information group into multi-level clusters and categories, and the analysis result presenting means includes a 5. The text information analysis system according to claim 4, further comprising means for moving the hierarchy up and down for each axis. A computer that operates as a text information analysis system that analyzes a large amount of collected and accumulated text information,
First and second analysis means,
A program for functioning as analysis result presenting means for presenting two analysis results of the first and second analysis means for the same text information group in an organically combined manner. A computer that operates as a text information analysis system that analyzes a large amount of collected and accumulated text information,
Clustering analysis means for analyzing the collected and accumulated large amount of text information based on the appearance frequency of each word and the degree of association between the plurality of words;
Text mining analysis means for analyzing a large amount of collected and accumulated text information based on arbitrarily specified conditions,
A program for functioning as analysis result presenting means for presenting an analysis result of the clustering analysis means and an analysis result of the text mining analysis means for the same text information group in an organically combined manner. The program according to claim 7, wherein the condition specified by the text mining analysis unit is a condition for classifying the text information group into a desired category. The analysis result presentation means presents the analysis result of the clustering analysis means and the analysis result of the text mining analysis means in a two-dimensional array tabular form assigned to a vertical axis and a horizontal axis, respectively. Item 7. The program according to Item 7. The clustering analysis unit and the text mining analysis unit classify the text information group into multi-level clusters and categories,
10. The program according to claim 9, wherein the analysis result presenting means includes means for moving a hierarchy of clusters and categories arranged as items of the vertical axis and the horizontal axis up and down for each axis. A method of presenting an analysis result applied to a text information analysis system for analyzing a large amount of collected and accumulated text information,
First and second analysis steps;
An analysis result presentation step of organically combining and presenting the two analysis results of the first and second analysis steps for the same text information group. A method of presenting an analysis result applied to a text information analysis system for analyzing a large amount of collected and accumulated text information,
A clustering analysis step of analyzing the collected and accumulated large amount of text information based on an appearance frequency of each word and a degree of association between a plurality of words;
Based on arbitrarily specified conditions, a text mining analysis step of analyzing the collected and accumulated large amount of text information,
An analysis result presentation step of organically combining and presenting the analysis result of the clustering analysis step and the analysis result of the text mining analysis step for the same text information group, and presenting the analysis result. . 13. The method according to claim 12, wherein the condition specified in the text mining analysis step is a condition for classifying the text information group into a desired category. The analysis result presenting step presents the analysis result of the clustering analysis step and the analysis result of the text mining analysis step in a two-dimensional array table format assigned to a vertical axis and a horizontal axis, respectively. Item 14. A method for presenting an analysis result according to Item 12. The clustering analysis step and the text mining analysis step classify the text information group into multi-level clusters and categories,
The analysis result presentation step according to claim 14, wherein the analysis result presentation step includes a step of moving up and down a hierarchy of clusters and categories arranged as items of the vertical axis and the horizontal axis for each axis. Method.

Images