WO2022220119A1 - 情報処理システム、情報処理方法及びプログラム - Google Patents

情報処理システム、情報処理方法及びプログラム Download PDF

Info

Publication number
WO2022220119A1
WO2022220119A1 PCT/JP2022/015868 JP2022015868W WO2022220119A1 WO 2022220119 A1 WO2022220119 A1 WO 2022220119A1 JP 2022015868 W JP2022015868 W JP 2022015868W WO 2022220119 A1 WO2022220119 A1 WO 2022220119A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
causal
chain
information processing
processing system
Prior art date
Application number
PCT/JP2022/015868
Other languages
English (en)
French (fr)
Inventor
潔 和泉
泰紀 坂地
Original Assignee
国立大学法人 東京大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 国立大学法人 東京大学 filed Critical 国立大学法人 東京大学
Publication of WO2022220119A1 publication Critical patent/WO2022220119A1/ja

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Definitions

  • the present invention relates to an information processing system, an information processing method, and a program.
  • Patent Document 1 discloses a causal relationship extraction system that extracts causal relationships between economic events and corporate performance.
  • Patent Literature 1 can extract a causal relationship, it cannot quantitatively predict an increase or decrease of a parameter.
  • the present invention provides an information processing system that allows users to more easily obtain quantitative prediction results regarding various economic events.
  • an information processing system includes a processor capable of executing a program to perform the following steps.
  • the obtaining step obtains the input information received from the user and preset reference information.
  • the reference information is information containing causes and effects that associate multiple causes and effects.
  • the extracting step extracts a related numerical index based on the acquired input information and reference information.
  • a related numerical index relates to at least one item appearing in a chain of causal relationships starting from a keyword included in the input information.
  • the user can more easily obtain quantitative prediction results regarding various economic events and the like.
  • FIG. 1 is a configuration diagram showing an information processing system 1 according to this embodiment
  • FIG. 2 is a block diagram showing the hardware configuration of the user terminal 2
  • FIG. 3 is a block diagram showing the hardware configuration of the server 3
  • FIG. 3 is a block diagram showing functions realized by a control unit 33 and the like in the server 3.
  • FIG. 10 is an activity diagram showing the flow of information processing regarding generation of reference information IF
  • FIG. 10 is an activity diagram showing the flow of information processing relating to the presentation of causal chain 4 and related numerical indices RI
  • 4 is a conceptual diagram for explaining causal chain 4
  • FIG. 2 is a diagram showing an example of a screen 5 of a display unit 24 visually recognized by a user
  • FIG. 2 is a diagram showing an example of a screen 5 of a display unit 24 visually recognized by a user;
  • FIG. 7A and 7B are diagrams showing another example of a screen 7 of the display unit 24 visually recognized by a user.
  • FIG. 7A and 7B are diagrams showing another example of a screen 7 of the display unit 24 visually recognized by a user.
  • FIG. 7A and 7B are diagrams showing another example of a screen 7 of the display unit 24 visually recognized by a user.
  • the program for realizing the software appearing in this embodiment may be provided as a non-transitory computer-readable medium (Non-Transitory Computer-Readable Medium), or may be downloaded from an external server. It may be provided as possible, or may be provided so that the program is activated on an external computer and the function is realized on the client terminal (so-called cloud computing).
  • the term “unit” may include, for example, a combination of hardware resources implemented by circuits in a broad sense and software information processing that can be specifically realized by these hardware resources.
  • various information is handled in the present embodiment, and these information are, for example, physical values of signal values representing voltage and current, and signal values as binary bit aggregates composed of 0 or 1. It is represented by high and low, or quantum superposition (so-called quantum bit), and communication and operation can be performed on a circuit in a broad sense.
  • a circuit in a broad sense is a circuit realized by at least appropriately combining circuits, circuits, processors, memories, and the like.
  • Application Specific Integrated Circuit ASIC
  • Programmable Logic Device for example, Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), and field It includes a programmable gate array (Field Programmable Gate Array: FPGA)).
  • FIG. 1 is a configuration diagram showing an information processing system 1 according to this embodiment.
  • the information processing system 1 includes a user terminal 2 and a server 3, which are connected through a network 11. FIG. These components are further described.
  • the system exemplified by the information processing system 1 consists of one or more devices or components. Therefore, even the server 3 alone is an example of a system.
  • FIG. 2 is a block diagram showing the hardware configuration of the user terminal 2. As shown in FIG. The user terminal 2 has a communication section 21 , a storage section 22 , a control section 23 , a display section 24 and an input section 25 . electrically connected. Descriptions of the communication unit 21, the storage unit 22, and the control unit 23 are omitted because they are substantially the same as those of the communication unit 31, the storage unit 32, and the control unit 33 in the server 3, which will be described later.
  • the display unit 24 may be included in the housing of the user terminal 2 or may be externally attached.
  • the display unit 24 displays a screen of a graphical user interface (GUI) that can be operated by the user.
  • GUI graphical user interface
  • the display unit 24 will be described as being included in the housing of the user terminal 2 .
  • the input unit 25 may be included in the housing of the user terminal 2 or may be externally attached.
  • the input unit 25 may be integrated with the display unit 24 and implemented as a touch panel. With a touch panel, the user can input a tap operation, a swipe operation, or the like.
  • a switch button, a mouse, a QWERTY keyboard, or the like may be employed instead of the touch panel. That is, the input unit 25 receives an operation input made by the user. The input is transferred as a command signal to the control unit 23 via the communication bus 20, and the control unit 23 can execute predetermined control and calculation as necessary.
  • FIG. 3 is a block diagram showing the hardware configuration of the server 3. As shown in FIG. The server 3 has a communication section 31 , a storage section 32 and a control section 33 , and these constituent elements are electrically connected inside the server 3 via a communication bus 30 . Each component will be further described.
  • the communication unit 31 preferably uses wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), wired LAN network communication, etc., but wireless LAN network communication, mobile communication such as 3G/LTE/5G, BLUETOOTH (registered trademark), etc. Communication and the like may be included as desired. That is, it is more preferable to implement as a set of these communication means. That is, the server 3 communicates various information with the user terminal 2 via the network 11 via the communication unit 31 .
  • wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), wired LAN network communication, etc.
  • wireless LAN network communication mobile communication such as 3G/LTE/5G, BLUETOOTH (registered trademark), etc. Communication and the like may be included as desired. That is, it is more preferable to implement as a set of these communication means. That is, the server 3 communicates various information with the user terminal 2 via the network 11 via the communication unit 31 .
  • the storage unit 32 stores various information defined by the above description. For example, it can be used as a storage device such as a solid state drive (SSD) for storing various programs related to the server 3 executed by the control unit 33, or as a temporary storage device related to program calculation. It can be implemented as a memory such as a random access memory (RAM) that stores various information (arguments, arrays, etc.). A combination of these may also be used. In particular, the storage unit 32 stores various programs related to the server 3 that are executed by the control unit 33 .
  • SSD solid state drive
  • RAM random access memory
  • the control unit 33 processes and controls overall operations related to the server 3 .
  • the control unit 33 is an example of a processor, such as a central processing unit (CPU) (not shown).
  • the control unit 33 implements various functions related to the server 3 by reading a predetermined program stored in the storage unit 32 . That is, information processing by software stored in the storage unit 32 can be specifically realized by the control unit 33 which is an example of hardware, and can be executed as each function unit included in the control unit 33 . These are further detailed in the next section.
  • the control unit 33 is not limited to a single unit, and may be implemented to have a plurality of control units 33 for each function. A combination thereof may also be used.
  • FIG. 4 is a block diagram showing functions realized by the control unit 33 and the like in the server 3.
  • the server 3 as an example of the information processing system 1 includes an acquisition unit 331 , a reference information generation unit 332 , a chain processing unit 333 and a display control unit 334 .
  • the acquisition unit 331 is configured to acquire various information necessary for information processing.
  • the acquisition unit 331 may read various information stored in the storage unit 32 in advance, or acquire various information from an external device via the network 11 and the communication unit 31 .
  • the acquisition unit 331 acquires the input text KW (input information) received from the user and the preset causal information database IF1 (reference information IF). This will be explained in more detail later.
  • the reference information generation unit 332 is configured to generate the reference information IF. More specifically, the reference information generating unit 332 generates a causal information database IF1 (reference information IF) by extracting a plurality of causal factors based on the text data TX. This will be explained in more detail later.
  • a causal information database IF1 reference information IF1
  • the chain processing unit 333 is configured to generate a causal chain 4 that visually indicates a chained causal relationship. This will be explained in more detail later.
  • the display control unit 334 is configured to generate various types of display information and control display content that can be visually recognized by the user.
  • the display information may be information itself generated in a user-visible manner, such as screens, images, icons, texts, etc., or, for example, displays screens, images, icons, texts, etc. on the display unit 24 of the user terminal 2. It may be rendering information for rendering. More specifically, the display control unit 334 controls to display a screen including the related numerical index RI based on the acquired input text KW (input information) and causal information database IF1 (reference information IF). do. This will be explained in more detail later.
  • FIG. 5 is an activity diagram showing the flow of information processing regarding generation of reference information IF.
  • the reference information IF is information that the server 3 refers to when performing information processing related to generating a causal chain 4 (see FIG. 7) and presenting a related numerical index RI (see FIGS. 8 and 9), which will be described later.
  • the reference information IF may be a causal information database IF1 that includes causalities in which multiple causes and effects are associated.
  • the reference information IF may further include information that associates a plurality of keywords with related numerical indices RI.
  • the relevant numerical index RI associated with the extracted node 4N in the causal chain 4 can be easily presented.
  • the associated information is a learned model IF2 obtained by machine-learning in advance the correlation between a plurality of keywords and the related numerical index RI.
  • the generation of the reference information IF is illustrated as being performed by the server 3.
  • the user may operate the server 3 directly, or may remotely operate the server 3 using the user terminal 2 or the like by various methods represented by remote desktop or the like.
  • the user continuously generates the causal information database IF1, which is an example of the reference information IF, and the learned model IF2 for presenting the related numerical index RI.
  • a user who wants to construct a causal information database IF1 which is an example of reference information IF, selects arbitrary text data TX.
  • the acquisition unit 331 in the server 3 acquires at least one piece of text data TX.
  • the text data TX may be, for example, news, topic articles, academic papers, technical specifications, patent documents, research reports, and the like.
  • the text data TX is not limited to a text file, and may be a word processing software file, a PDF file, or an HTML file on the Internet (activity A001).
  • the control unit 33 in the server 3 reads the dedicated program stored in the storage unit 32 to analyze the text data TX selected in activity A001.
  • causal sets relating to various events are extracted from the text data TX (activity A002).
  • a method for extracting causal pairs is not particularly limited, and an existing natural language processing algorithm may be appropriately adopted.
  • the control unit 33 in the server 3 reads out the dedicated program stored in the storage unit 32 to add the causal information database IF1 to the causal information database IF1 by adding the causal group extracted in the activity A002. Generate (activity A003).
  • the reference information generating unit 332 generates the causal information database IF1 (reference information IF) by extracting a plurality of causal factors based on the text data TX. According to this aspect, it is possible to generate the causal information database IF1 according to the user's preference and the desired usage environment, so that it is possible to realize the prediction of an event with a more limited application.
  • the user can select multiple text data TX. For example, after the causal information database IF1 is generated using the text data TX selected for the first time, another text data TX can be selected for the second and subsequent processes. In other words, by appropriately repeating the activities A001 to A003, it is possible to generate the causal information database IF1 with a larger amount of information.
  • a user who wants to build a learned model IF2 which is an example of reference information IF, selects various keywords and numerical indices that serve as teacher data (activity A004). This may be the selection of the text data TX used to extract the causal pair, or the selection of different data.
  • the user inputs various selected keywords and numerical indices into an existing machine learning algorithm to cause the server 3 to perform machine learning (activity A005).
  • a learned model IF2 is generated (activity A006).
  • the user can update the learned model IF2 by repeating machine learning using additional teacher data. For example, after generating the trained model IF2 using the keywords and numerical indices selected the first time, as the second and subsequent processes, it is possible to perform further machine learning using different teacher data as input. In other words, by repeating activities A004 to A006 as appropriate, it is possible to generate a more accurate learned model IF2.
  • FIG. 6 is an activity diagram showing the flow of information processing regarding the presentation of the causal chain 4 and the associated numerical index RI.
  • the acquisition unit 331 acquires the input text KW (input information) received from the user and the (preset) reference information IF generated by the information processing shown in FIG. Activity A 102). That is, the control unit 33 matches the input text KW received through the communication unit 31 with the causal information database IF1 stored in the storage unit 32, thereby searching for a causal set suitable for the input text KW (activity A103). , determine this (activity A104). Examples of causal sets are described below.
  • control unit 33 inputs the keyword included in the causal pair to the learned model IF2 stored in the storage unit 32, so that the related numerical index RI highly related to the keyword is obtained. proposed. An example of the related numerical index RI will be described later.
  • a screen such as screen 5 is displayed in a manner that allows the user to grasp the set of determined causality and the proposed related numerical index RI (activity A105).
  • the control unit 33 reads out a predetermined program stored in the storage unit 32, and as an extraction step, based on the acquired input text KW (input information) and causal information database IF1 (reference information IF) to extract the related numerical index RI.
  • the display control unit 334 controls to display the screen 5 or the like including the extracted related numerical index RI.
  • the user grasps the related numerical index RI as well as the cause and effect of various events related thereto. can do. That is, the user can obtain a more quantitative prediction result than in the past.
  • the user may optionally enter a different input text KW of interest next.
  • the user can obtain prediction results for a plurality of events.
  • another similar causal set can be searched continuously. Specifically, first, the user selects one of the causal pairs (node 4N) already displayed on the display unit 24 using the input unit 25 (activity A106).
  • control unit 33 reads out the reference information IF (activity A107), and compares the selected causal pair with the causal information database IF1 stored in the storage unit 32 to determine the relationship between the causal pair and the relationship. search for and determine a causal set of (activity A108). Examples of relevant causal sets are described below.
  • control unit 33 inputs a keyword included in the causal pair into the learned model IF2 stored in the storage unit 32, thereby making the keyword highly related to the keyword.
  • a new relevant numerical index RI is proposed (activity A109).
  • screens such as screen 5 are updated in a manner that allows the user to grasp the set of determined causality and the proposed related numerical index RI (activity A110).
  • the user repeats the activities A101 to A105 or A106 to A110 as appropriate to generate the causal chain 4, which is visibly displayed to the user.
  • the chain processing unit 333 which is a functional unit of the control unit 33, can generate the causal chain 4 visually showing the chained causal relationship as a chain processing step. Then, along with the causal chain 4, a relevant numerical index RI associated with each causal set will be proposed.
  • this information processing method includes each step of the information processing system 1, and more specifically, includes the following steps.
  • the acquisition step the input text KW (input information) received from the user and the preset causal information database IF1 (reference information) are acquired.
  • the causal information database IF1 (reference information) is information containing causalities in which multiple causes and effects are associated.
  • the extraction step the related numerical index RI is extracted based on the acquired input text KW (input information) and causal information database IF1 (reference information).
  • the related numerical index RI is related to at least one item appearing in a chain of causal relationships starting from the keyword included in the input text KW (input information).
  • FIG. 7 is a conceptual diagram for explaining the causal chain 4.
  • FIG. 8 and 9 are diagrams showing an example of the screen 5 of the display unit 24 visually recognized by the user.
  • the causal chain 4 is a network structure that hierarchically has nodes 4N representing causal pairs.
  • a causal chain 4 indicates a chain of causal relationships starting from the input text KW entered by the user. More specifically, the causal chain 4 includes multiple nodes 4N that indicate causality in the causal information database IF1.
  • a node 41, a node 42, and a node 43 which are nodes 4N representing a set of three causal effects, are presented based on the causal information database IF1.
  • causal information database IF1 is an example of reference information IF preset based on text data TX.
  • nodes 411, 412 and 413 are presented as nodes 4N following node 41.
  • nodes 421, 422 and 423 are presented as nodes 4N following node 42, related to the causal set shown at node 42, and related to the causal set shown at node 43, node 431 , node 432 and node 433 are presented as node 4 N following node 43 .
  • Such a causal chain 4 makes it possible to form a causal stream of related events.
  • the causal effects shown in the nodes 4N connected to each other are causal effects similar to each other.
  • the plurality of nodes 4N are configured to be able to calculate similarities with each other.
  • the causal chain 4 has a given causal chain 4' containing a plurality of nodes 4N in series. For example, in FIG. 7, input text KW, node 41 and node 411 correspond to a given causal chain 4'.
  • the chain processing unit 333 adds a new node 4N succeeding the terminal node 4N of the predetermined causal chain 4', and the added new node 4N is similar to the terminal node 4N by a predetermined value or more. degree.
  • the concept of similarity is not particularly limited, and the similarity may be determined based on the distance defined in existing natural language processing, for example.
  • the causal chain 4 may be presented on the screen 5 as shown in FIG.
  • the user is inputting "infectious disease” into the input field 50 as the input text KW while referring to the screen 5 .
  • a node 4N representing a set of causes and effects consisting of a first cause 51 and a first result 52 is displayed in the region R1.
  • the first cause 51 includes the keywords "world situation”, “infectious disease” and “influence”, and the first results 52 resulting from these are "Japanese”, “Departure company”, “Year-on-year change” and the keyword “many people”.
  • the region R2 further displays a node 4N indicating a causal set consisting of the second cause 61 and the second effect 62.
  • the second causes 61 include the keywords “Japanese,” “guests,” and “decrease,” and the resulting second results 62 are "supply and demand,” “mitigation,” “direction,” and “guest rooms.” ” and “unit price” are included.
  • each time the node 4N is added a new area and the corresponding node 4N should be displayed one after another under the area R2.
  • the method of selecting the terminal node 4N when adding the succeeding node 4N is not particularly limited. , a new node 4N is displayed in the region R2, or the like.
  • a first cause 51 and its result, a first result 52 are presented.
  • a second result 62 which is a further result thereof, is presented. That is, the causal chain 4 is presented in the direction of pursuing the effect from the cause.
  • the associated numerical index RI associated with second result 62 is presented along with causal chain 4. That is, the related numerical index RI is related to at least one item that appears in a chain of causal relationships starting from the input text KW. At least one item corresponds to a "guest room" here. More specifically, the associated numerical index RI is associated with at least one node 4N appearing in the causal chain 4; Further, in generating such a screen 5, the display control unit 334 controls to display a screen including the causal chain 4 and the related numerical index RI. According to this aspect, since the related numerical index RI is displayed on the screen together with the causal chain 4 in a glanceable manner, the user can comprehend the flow of causality and the related numerical index RI from a bird's-eye view. can do.
  • the trained model IF2 can be referred to when presenting the related numerical index RI.
  • a method for determining the degree of association is not particularly limited, but more specifically, it may be determined based on a modified odds ratio as shown in Equation 1, for example.
  • the degree of unrelevance may be determined and the numerical relevance index RI may be presented.
  • the keywords included in the set of causality indicated by the node 4N include keywords indicating a large or small increase or decrease, such as "decrease”, “increase”, “significant”, or “minor”.
  • a node 4N of may have a polarity representing a numerical increase or decrease, and the associated numerical index RI may be determined based on the polarity. In predicting economic events, etc., increasing or decreasing changes are extremely important. Considering the directionality of the change in polarity along the causal chain 4, it is possible to present a preferable related numerical index RI desired by the user, which is a user-friendly specification. In addition to presenting the related numerical index RI, it may be implemented so as to obtain a predicted value of the numerical value.
  • the information processing system 1 searches for causal relationships such as economic events in a chained manner from the causal information database IF1 extracted from the text data TX, and presents the related numerical index RI related to the ripple effect. Proposed. According to such an information processing method by the information processing system 1, the user can more easily obtain quantitative prediction results regarding various economic phenomena and the like.
  • a program that causes a computer to execute each step of the information processing system 1 may be provided.
  • Such a program may be configured to be executable offline on a stand-alone computer separated from the network 11 .
  • the related numerical index RI When presenting the related numerical index RI, it may be implemented not only to present the index itself, but also to obtain a predicted value of time change, such as an upward change in the value after X days or within X months.
  • At least one text data TX selected for generating reference information such as causal information database IF1 may be a plurality of text data TX written in different languages.
  • the first time, text data TX written in Japanese is used to extract causal pairs
  • the second time, text data TX written in a foreign language such as English or Chinese is used to extract causal pairs.
  • a new node 4N related to the terminal node 4N is searched for and determined based on the causal information database IF1.
  • the causal information database IF1 may be implemented.
  • other nodes 4N located in parallel in the same hierarchy may be referred to.
  • not only the immediately preceding node 4N but also a new node 4N that follows the previous causal flow is proposed, so that the user can make more accurate predictions.
  • the causal chain 4 is presented in the direction of pursuing the effect from the cause, but the user is prompted to enter the input text KW regarding the result, and the causal chain is shown in the direction of pursuing the cause from the result. It may be implemented to present chain 4.
  • 10 and 11 are diagrams showing another example of the screen 7 of the display unit 24 visually recognized by the user.
  • FIG. 10 the user is inputting "infectious disease" into the input field 70 as the input text KW while referring to the screen 7.
  • nodes 81, 82, 83, and a plurality of nodes 4N are presented on the right side of the input field 70.
  • the screen 7 is provided with a selection button 76 and a selection button 77 below the input field 70 .
  • the select button 76 When the user selects the select button 76, the causal chain 4 can be displayed in the direction of pursuing the effect from the cause.
  • the selection button 77 when the user selects the selection button 77, the causal chain 4 can be displayed in the direction of pursuing the cause from the effect.
  • one item causes a plurality of results, it is considered to be a forward problem to present the direction to pursue the result from the cause.
  • constraint conditions are not particularly limited, and may be assumed as appropriate, such as changing the weighting value, limiting the subsequent node 4N, and the like.
  • the causal chain 4 can theoretically be added almost endlessly as long as the information in the causal information database IF1 is sufficient. It is assumed that the relationship with KW will become weaker. To prevent this, it may be implemented to limit the hierarchy of nodes 4N that can be included in one causal chain 4.
  • FIG. Although the number of layers to be restricted is not particularly limited, specifically, for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 layers, and may be in the range between any two of the values exemplified herein. It is more preferably 1 to 5 layers, and particularly preferably 3 layers.
  • the causal chain 4 and the related numerical index RI are generated with the input text KW as the starting point, and an example in which these are presented has been described. It may be implemented such that a causal chain 4 containing 4N is presented.
  • the acquisition unit 331 acquires the numerical index (related numerical index RI) received from the user and preset reference information IF as an acquisition step.
  • the reference information IF is information that associates a plurality of keywords with numerical indices, and is, for example, a learned model IF2.
  • the control unit 33 extracts the causal chain 4, which is a chained causal relationship including the keyword, based on the acquired numerical index and reference information.
  • the information processing system is configured to further execute a display control step, wherein the display control step controls to display a screen including the extracted related numerical index.
  • the reference information further includes information that associates a plurality of keywords with the related numerical index.
  • the associated information is a learned model obtained by previously machine-learning the correlation between the plurality of keywords and the associated numerical index.
  • the information processing system further comprising a chain processing step, wherein the chain processing step generates a causal chain visually indicating a chained causal relationship, wherein the causal chain indicates causality in the reference information including a plurality of nodes, wherein the display control step controls display of a screen including the causal chain and the related numerical index, the related numerical index being related to at least one node appearing in the causal chain ,thing.
  • the plurality of nodes are configured to be able to calculate similarity with each other
  • the causal chain includes a predetermined causal chain including a plurality of serially connected nodes
  • the chain processing step includes the predetermined adding a new node succeeding the terminal node of the causal chain of , wherein the added new node has similarity with the terminal node equal to or greater than a predetermined value.
  • a plurality of nodes in the predetermined causal chain are weighted, and new subsequent nodes are determined based on the weighted evaluation values.
  • the plurality of nodes have polarities representing numerical increases or decreases, and the associated numerical indices are determined based on the polarities.
  • the information processing system further includes a reference information generating step, wherein the obtaining step obtains at least one piece of text data, and the reference information generating step extracts a plurality of causal factors based on the text data, Generating said reference information.
  • the information processing system further comprises a chain processing step, wherein the at least one text data is a plurality of text data written in different languages, and the chain processing step visually indicates chain causal relationships. generating a causal chain, wherein said causal chain includes a plurality of nodes each representing causality described in said different languages.
  • An information processing system comprising a processor capable of executing a program so as to perform the following steps, in an obtaining step obtaining a numerical index received from a user and preset reference information, wherein The reference information is information that associates a plurality of keywords with the numerical index, and in the extracting step, a chain causal relationship including the keyword is extracted based on the obtained numerical index and the reference information. ,thing.
  • a method of information processing comprising steps of the information processing system.
  • a program that causes a computer to perform each step of the information processing system. Of course, this is not the only case.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

【課題】ユーザがより容易に、種々の経済事象等に関する定量的な予測結果を得ることができる情報処理システム等を提供すること。 【解決手段】本発明の一態様によれば、情報処理システムが提供される。この情報処理システムでは、次の各ステップがなされるようにプログラムを実行可能なプロセッサを備える。取得ステップでは、ユーザから受け付けた入力情報と、予め設定された参照情報とを取得する。参照情報は、複数の、原因と結果とを関連付けた因果を含む情報である。抽出ステップでは、取得された、入力情報及び参照情報に基づいて、関連数値指標を抽出する。関連数値指標は、入力情報に含まれるキーワードを起点とした、連鎖的な因果関係に登場する少なくとも1つの事項に関連する。

Description

情報処理システム、情報処理方法及びプログラム
 本発明は、情報処理システム、情報処理方法及びプログラムに関する。
 特許文献1には、経済事象と企業業績との因果関係の抽出を行う因果関係抽出システムが開示されている。
特開2020-24689号公報
 ところで近年では、情報処理技術を用いて、種々の経済事象等をより定量的に予測することが求められている。しかしながら、特許文献1に開示される技術は、因果関係を抽出することはできるものの、パラメータの増加や減少といった定量的な予測をすることができない。
 本発明では上記事情に鑑み、ユーザがより容易に、種々の経済事象等に関する定量的な予測結果を得ることができる情報処理システム等を提供することとした。
 本発明の一態様によれば、情報処理システムが提供される。この情報処理システムでは、次の各ステップがなされるようにプログラムを実行可能なプロセッサを備える。取得ステップでは、ユーザから受け付けた入力情報と、予め設定された参照情報とを取得する。参照情報は、複数の、原因と結果とを関連付けた因果を含む情報である。抽出ステップでは、取得された、入力情報及び参照情報に基づいて、関連数値指標を抽出する。関連数値指標は、入力情報に含まれるキーワードを起点とした、連鎖的な因果関係に登場する少なくとも1つの事項に関連する。
 このような一態様によれば、ユーザがより容易に、種々の経済事象等に関する定量的な予測結果を得ることができる。
本実施形態に係る情報処理システム1を表す構成図である。 ユーザ端末2のハードウェア構成を示すブロック図である。 サーバ3のハードウェア構成を示すブロック図である。 サーバ3における制御部33等によって実現される機能を示すブロック図である。 参照情報IFの生成に関する情報処理の流れを示すアクティビティ図である。 因果チェーン4及び関連数値指標RIの提示に関する情報処理の流れを示すアクティビティ図である。 因果チェーン4を説明するための概念図である。 ユーザに視認される表示部24の画面5の態様一例を示した図である。 ユーザに視認される表示部24の画面5の態様一例を示した図である。 ユーザに視認される表示部24の画面7の態様別例を示した図である。 ユーザに視認される表示部24の画面7の態様別例を示した図である。
 以下、図面を用いて本発明の実施形態について説明する。以下に示す実施形態中で示した各種特徴事項は、互いに組み合わせ可能である。
 ところで、本実施形態に登場するソフトウェアを実現するためのプログラムは、コンピュータが読み取り可能な非一時的な記録媒体(Non-Transitory Computer-Readable Medium)として提供されてもよいし、外部のサーバからダウンロード可能に提供されてもよいし、外部のコンピュータで当該プログラムを起動させてクライアント端末でその機能を実現(いわゆるクラウドコンピューティング)するように提供されてもよい。
 また、本実施形態において「部」とは、例えば、広義の回路によって実施されるハードウェア資源と、これらのハードウェア資源によって具体的に実現されうるソフトウェアの情報処理とを合わせたものも含みうる。また、本実施形態においては様々な情報を取り扱うが、これら情報は、例えば電圧・電流を表す信号値の物理的な値、0又は1で構成される2進数のビット集合体としての信号値の高低、又は量子的な重ね合わせ(いわゆる量子ビット)によって表され、広義の回路上で通信・演算が実行されうる。
 また、広義の回路とは、回路(Circuit)、回路類(Circuitry)、プロセッサ(Processor)、及びメモリ(Memory)等を少なくとも適当に組み合わせることによって実現される回路である。すなわち、特定用途向け集積回路(Application Specific Integrated Circuit:ASIC)、プログラマブル論理デバイス(例えば、単純プログラマブル論理デバイス(Simple Programmable Logic Device:SPLD)、複合プログラマブル論理デバイス(Complex Programmable Logic Device:CPLD)、及びフィールドプログラマブルゲートアレイ(Field Programmable Gate Array:FPGA))等を含むものである。
1.ハードウェア構成
 本節では、本実施形態のハードウェア構成について説明する。
1.1 情報処理システム1
 図1は、本実施形態に係る情報処理システム1を表す構成図である。情報処理システム1はユーザ端末2と、サーバ3とを備え、これらがネットワーク11を通じて接続されている。これらの構成要素についてさらに説明する。ここで、情報処理システム1に例示されるシステムとは、1つ又はそれ以上の装置又は構成要素からなるものである。したがって、サーバ3単体であってもシステムの一例となる。
1.2 ユーザ端末2
 図2は、ユーザ端末2のハードウェア構成を示すブロック図である。ユーザ端末2は、通信部21と、記憶部22と、制御部23と、表示部24と、入力部25とを有し、これらの構成要素がユーザ端末2の内部において通信バス20を介して電気的に接続されている。通信部21、記憶部22及び制御部23の説明は、後述の、サーバ3における通信部31、記憶部32及び制御部33と略同様のため省略する。
 表示部24は、例えば、ユーザ端末2の筐体に含まれるものであってもよいし、外付けされるものであってもよい。表示部24は、ユーザが操作可能なグラフィカルユーザインターフェース(Graphical User Interface:GUI)の画面を表示する。これは例えば、CRTディスプレイ、液晶ディスプレイ、有機ELディスプレイ及びプラズマディスプレイ等の表示デバイスを、ユーザ端末2の種類に応じて使い分けて実施することが好ましい。ここでは、表示部24は、ユーザ端末2の筐体に含まれるものとして説明する。
 入力部25は、ユーザ端末2の筐体に含まれるものであってもよいし、外付けされるものであってもよい。例えば、入力部25は、表示部24と一体となってタッチパネルとして実施されてもよい。タッチパネルであれば、ユーザは、タップ操作、スワイプ操作等を入力することができる。もちろん、タッチパネルに代えて、スイッチボタン、マウス、QWERTYキーボード等を採用してもよい。すなわち、入力部25がユーザによってなされた操作入力を受け付ける。当該入力が命令信号として、通信バス20を介して制御部23に転送され、制御部23が必要に応じて所定の制御や演算を実行しうる。
1.3 サーバ3
 図3は、サーバ3のハードウェア構成を示すブロック図である。サーバ3は、通信部31と、記憶部32と、制御部33とを有し、これらの構成要素がサーバ3の内部において通信バス30を介して電気的に接続されている。各構成要素についてさらに説明する。
 通信部31は、USB、IEEE1394、Thunderbolt(登録商標)、有線LANネットワーク通信等といった有線型の通信手段が好ましいものの、無線LANネットワーク通信、3G/LTE/5G等のモバイル通信、BLUETOOTH(登録商標)通信等を必要に応じて含めてもよい。すなわち、これら複数の通信手段の集合として実施することがより好ましい。すなわち、サーバ3は、通信部31を介して、ユーザ端末2とネットワーク11を介して種々の情報を通信する。
 記憶部32は、前述の記載により定義される様々な情報を記憶する。これは、例えば、制御部33によって実行されるサーバ3に係る種々のプログラム等を記憶するソリッドステートドライブ(Solid State Drive:SSD)等のストレージデバイスとして、あるいは、プログラムの演算に係る一時的に必要な情報(引数、配列等)を記憶するランダムアクセスメモリ(Random Access Memory:RAM)等のメモリとして実施されうる。また、これらの組合せであってもよい。特に、記憶部32は、制御部33によって実行されるサーバ3に係る種々のプログラム等を記憶している。
 制御部33は、サーバ3に関連する全体動作の処理・制御を行う。制御部33は、プロセッサの一例であり、例えば不図示の中央処理装置(Central Processing Unit:CPU)である。制御部33は、記憶部32に記憶された所定のプログラムを読み出すことによって、サーバ3に係る種々の機能を実現する。すなわち、記憶部32に記憶されているソフトウェアによる情報処理が、ハードウェアの一例である制御部33によって具体的に実現されることで、制御部33に含まれる各機能部として実行されうる。これらについては、次節においてさらに詳述する。なお、制御部33は単一であることに限定されず、機能ごとに複数の制御部33を有するように実施してもよい。またそれらの組合せであってもよい。
2.機能構成
 本節では、本実施形態の機能構成について説明する。前述の通り、記憶部32に記憶されているソフトウェアによる情報処理がハードウェアの一例である制御部33によって具体的に実現されることで、制御部33に含まれる各機能部として実行されうる。
 図4は、サーバ3における制御部33等によって実現される機能を示すブロック図である。具体的には、情報処理システム1の一例であるサーバ3は、取得部331と、参照情報生成部332と、チェーン処理部333と、表示制御部334とを備える。
 取得部331は、情報処理に必要な種々の情報を取得するように構成される。例えば、取得部331は、予め記憶部32に記憶された種々の情報を読み出してもよいし、ネットワーク11及び通信部31を介して外部の機器から種々の情報を取得してもよい。特に、取得部331は、ユーザから受け付けた入力テキストKW(入力情報)と、予め設定された因果情報データベースIF1(参照情報IF)とを取得する。これについては後に更に詳述する。
 参照情報生成部332は、参照情報IFを生成するように構成される。より具体的には、参照情報生成部332は、テキストデータTXに基づいて複数の因果を抽出することで、因果情報データベースIF1(参照情報IF)を生成する。これについては後に更に詳述する。
 チェーン処理部333は、連鎖的な因果関係を視覚的に示した因果チェーン4を生成するように構成される。これについては後に更に詳述する。
 表示制御部334は、種々の表示情報を生成して、ユーザが視認可能な表示内容を制御するように構成される。表示情報とは、画面、画像、アイコン、テキスト等といった、ユーザが視認可能な態様で生成された情報そのものでもよいし、例えばユーザ端末2の表示部24に画面、画像、アイコン、テキスト等を表示させるためのレンダリング情報であってもよい。より具体的には、表示制御部334は、取得された、入力テキストKW(入力情報)及び因果情報データベースIF1(参照情報IF)に基づいて、関連数値指標RIを含む画面を表示させるように制御する。これについては後に更に詳述する。
3.情報処理方法
 本節では、前述した情報処理システム1の情報処理方法について説明する。
3.1 参照情報IFの生成
 まず、参照情報IFの生成に関する情報処理ついて説明する。図5は、参照情報IFの生成に関する情報処理の流れを示すアクティビティ図である。参照情報IFは、後述の因果チェーン4(図7参照)の生成や関連数値指標RI(図8及び図9参照)の提示に係る情報処理に際して、サーバ3が参照する情報である。
 例えば参照情報IFは、複数の、原因と結果とを関連付けた因果を含む因果情報データベースIF1であるとよい。あるいは、参照情報IFは、複数のキーワードと関連数値指標RIとを関連付けた情報をさらに含むものであるとよい。このような態様によれば、因果チェーン4における抽出されたノード4Nに関連する関連数値指標RIを容易に提示することができる。好ましくは、関連付けた情報は、複数のキーワードと、関連数値指標RIとの相関性を予め機械学習させた学習済モデルIF2であるとよい。
 また、参照情報IFの生成は、サーバ3によって行われることを例示する。この際ユーザは、サーバ3を直接操作してもよいし、ユーザ端末2等を用いて、リモートデスクトップ等に代表される種々の手法で遠隔でサーバ3を操作してもよい。以下では、ユーザが、参照情報IFの一例である因果情報データベースIF1と、関連数値指標RIを提示するための学習済モデルIF2とを連続して生成する場合を説明する。
 まず、参照情報IFの一例である因果情報データベースIF1を構築したいユーザは、任意のテキストデータTXを選択する。換言すると、サーバ3における取得部331が、少なくとも1つのテキストデータTXを取得している。テキストデータTXは例えば、ニュース、トピック記事、学術論文、技術仕様書、特許文書、調査レポート等であるとよい。また、テキストデータTXは、テキストファイルに限定されるものではなく、ワープロソフトのファイルでもよいし、PDFファイルでもよいし、インターネット上のHTMLファイルでもよい(アクティビティA001)。
 続いて、サーバ3における制御部33が、記憶部32に記憶した専用プログラムを読み出すことで、アクティビティA001にて選択されたテキストデータTXを解析する。この結果、テキストデータTXから、種々の事象に関する因果の組が抽出される(アクティビティA002)。因果の組の抽出手法は特に限定されず、既存の自然言語処理のアルゴリズムが適宜採用されるとよい。
 続いて、サーバ3における制御部33が、記憶部32に記憶した専用プログラムを読み出すことで、アクティビティA002にて抽出された因果の組を因果情報データベースIF1に追加することで、因果情報データベースIF1を生成する(アクティビティA003)。換言すると、参照情報生成部332は、テキストデータTXに基づいて複数の因果を抽出することで、因果情報データベースIF1(参照情報IF)を生成している。このような態様によれば、ユーザの好みや、求める使用環境に応じて因果情報データベースIF1を生成することができるため、より用途等を限定した事象の予測を実現することができる。
 なお必要に応じて、ユーザは、複数のテキストデータTXを選択することができる。例えば、1回目に選択したテキストデータTXを用いて、因果情報データベースIF1の生成した後に、2回目以降の処理として、別のテキストデータTXを選択することができる。換言すると、アクティビティA001~A003を適宜繰り返すことで、より情報量の高い因果情報データベースIF1を生成することができる。
 続いて、参照情報IFの一例である学習済モデルIF2を構築したいユーザは、教師データとなる、種々のキーワードと、数値指標とを選択する(アクティビティA004)。これは、因果の組の抽出に用いたテキストデータTXの選択でもよいし、異なるデータの選択でもよい。続いて、ユーザは、選択された種々のキーワードと、数値指標とを、既存の機械学習のアルゴリズムに入力し、サーバ3に機械学習を実行させる(アクティビティA005)。その結果、学習済モデルIF2が生成される(アクティビティA006)。このように機械学習を行うことで、ルックアップテーブル等を使った手法よりも、さらに精度高く、関係性のある関連数値指標RIを提示することができる。
 なお必要に応じて、ユーザは、追加の教師データを用いて機械学習を繰り返すことで、学習済モデルIF2を更新することができる。例えば、1回目に選択したキーワード及び数値指標を用いて、学習済モデルIF2の生成した後に、2回目以降の処理として、別の教師データを入力としてさらなる機械学習を行うことができる。換言すると、アクティビティA004~A006を適宜繰り返すことで、より精度の高い学習済モデルIF2を生成することができる。
3.2 因果チェーン4及び関連数値指標RIの提示
 次に、因果チェーン4及び関連数値指標RIの提示に関する情報処理について説明する。図6は、因果チェーン4及び関連数値指標RIの提示に関する情報処理の流れを示すアクティビティ図である。
(流れの概説)
 以下、図6の各アクティビティに沿って、情報処理の流れを概説する。ここでのユーザは、情報処理システム1が提供するサービスを用いて経済事象を予測したい者であるとする。ユーザは、ユーザ端末2における表示部24に表示された画面5を参照しながら、入力部25を用いて、自身が予測したい事象に関する入力テキストKW(入力情報)を入力する(アクティビティA101)。この入力テキストKWは、ネットワーク11を介してサーバ3に送信される。
 続いて、取得部331は、取得ステップとして、ユーザから受け付けた入力テキストKW(入力情報)と、図5に示される情報処理により生成された(予め設定された)参照情報IFとを取得する(アクティビティA102)。すなわち、制御部33は、通信部31を通じて受信した入力テキストKWを記憶部32に記憶された因果情報データベースIF1に突合させることで、入力テキストKWに適した因果の組を探索し(アクティビティA103)、これを決定する(アクティビティA104)。因果の組の例は後述する。
 また、因果の組が決定すると、制御部33が因果の組に含まれるキーワードを記憶部32に記憶された学習済モデルIF2に入力することで、かかるキーワードに関連性の高い関連数値指標RIが提案される。関連数値指標RIの例は後述する。
 続いて、決定された因果の組及び提案された関連数値指標RIをユーザが把握可能な態様で、画面5等の画面が表示される(アクティビティA105)。換言すると、制御部33が、記憶部32に記憶された所定のプログラムを読み出すことで、抽出ステップとして、取得された、入力テキストKW(入力情報)及び因果情報データベースIF1(参照情報IF)に基づいて、関連数値指標RIを抽出する。そして、表示制御部334は、表示制御ステップとして、抽出された関連数値指標RIを含む画面5等を表示させるように制御する。このような態様によれば、ユーザは、自身が関心のある事象に関する入力テキストKWを入力することで、これに関連する種々の事象の因果とともに、関連する数値指標である関連数値指標RIを把握することができる。すなわち、ユーザが従来に比してより定量的な予測結果を得ることができる。
 なお、ユーザは、適宜次に関心のある異なる入力テキストKWを入力してもよい。換言すると、アクティビティA101~A105を適宜繰り返すことで、ユーザは、複数の事象について、予測結果を得ることができる。
 また、アクティビティA103において決定された因果の組に含まれるキーワードに基づいて、これに類似する別の因果の組を続けて探索することができる。具体的にはまず、表示部24にすでに表示されている因果の組(ノード4N)のうちの1つを、ユーザが入力部25を用いて選択する(アクティビティA106)。
 続いて、制御部33は、参照情報IFを読み出す(アクティビティA107)とともに、選択された因果の組を記憶部32に記憶された因果情報データベースIF1に突合させることで、かかる因果の組と関連性のある因果の組を探索し、これを決定する(アクティビティA108)。関連性のある因果の組の例は後述する。
 また、関連性のある因果の組が決定すると、制御部33が因果の組に含まれるキーワードを記憶部32に記憶された学習済モデルIF2に入力することで、かかるキーワードに関連性の高い、新たな関連数値指標RIが提案される(アクティビティA109)。
 続いて、決定された因果の組及び提案された関連数値指標RIをユーザが把握可能な態様で、画面5等の画面が更新される(アクティビティA110)。
 すなわち、アクティビティA101~A105又はA106~A110を、ユーザが適宜繰り返すことで、因果チェーン4が生成され、これがユーザに視認可能に表示される。換言すると、制御部33の機能部であるチェーン処理部333は、チェーン処理ステップとして、連鎖的な因果関係を視覚的に示した因果チェーン4を生成することができる。そして、因果チェーン4とともに、各因果の組に関連する関連数値指標RIが提案されることとなる。
 以上をまとめると、この情報処理方法は、情報処理システム1の各ステップを備え、より具体的には、次の各ステップを備える。取得ステップでは、ユーザから受け付けた入力テキストKW(入力情報)と、予め設定された因果情報データベースIF1(参照情報)とを取得する。因果情報データベースIF1(参照情報)は、複数の、原因と結果とを関連付けた因果を含む情報である。抽出ステップでは、取得された、入力テキストKW(入力情報)及び因果情報データベースIF1(参照情報)に基づいて、関連数値指標RIを抽出する。関連数値指標RIは、入力テキストKW(入力情報)に含まれるキーワードを起点とした、連鎖的な因果関係に登場する少なくとも1つの事項に関連する。
(情報処理の詳細)
 続いて、図6で概説された情報処理の詳細部分を説明する。図7は、因果チェーン4を説明するための概念図である。図8及び図9は、ユーザに視認される表示部24の画面5の態様一例を示した図である。
 図7に示されるように、因果チェーン4は、因果の組を表すノード4Nを階層的に有するネットワーク構造体である。因果チェーン4は、ユーザが入力した入力テキストKWを起点とした連鎖的な因果関係を示している。より具体的には、因果チェーン4は、因果情報データベースIF1における因果を示す複数のノード4Nを含む。この例では、ユーザが入力テキストKWを入力すると、因果情報データベースIF1に基づいて、3つの因果の組を表すノード4Nである、ノード41、ノード42及びノード43が提示されている。先に説明したように、因果情報データベースIF1は、テキストデータTXに基づいて予め設定されている参照情報IFの一例である。
 また、ノード41に示される因果の組に関連し、ノード411、ノード412及びノード413が、ノード41に後続するノード4Nとして提示されている。同様に、ノード42に示される因果の組に関連し、ノード421、ノード422及びノード423が、ノード42に後続するノード4Nとして提示され、ノード43に示される因果の組に関連し、ノード431、ノード432及びノード433が、ノード43に後続するノード4Nとして提示されている。このような因果チェーン4によれば、関連する事象の因果の流れを形成することができる。
 ここで、互いに接続されたノード4Nに示される因果それぞれは、互いに類似する因果であることが好ましい。より具体的には、複数のノード4Nは、互いに類似度を算出可能に構成される。因果チェーン4は、直列された複数のノード4Nを含む所定の因果チェーン4'を有する。例えば、図7では、入力テキストKW、ノード41及びノード411が所定の因果チェーン4'に相当する。そして、チェーン処理部333は、所定の因果チェーン4'の末端のノード4Nに後続する新たなノード4Nを追加し、追加された新たなノード4Nは、末端のノード4Nと、所定値以上の類似度を有するとよい。類似度の概念は特に限定されるものではなく、例えば既存の自然言語処理において規定された距離に基づいて、類似度を判断すればよい。
 さらに具体的には、図8に示されるような画面5において、因果チェーン4が提示されてもよい。図8の例では、ユーザは、画面5を参照しながら、入力テキストKWとして、「感染症」を入力欄50に入力している。この状態で、ユーザが検索アイコン501を押下すると、領域R1に、第1の原因51及び第1の結果52からなる因果の組を示すノード4Nが表示される。第1の原因51は、「世界情勢」、「感染症」及び「影響」というキーワードを含み、これらに起因する第1の結果52は、「日本人」、「出国社」、「前年比」及び「万人」というキーワードを含んでいる。
 さらにユーザが、領域R1に対応するノード4Nにおけるキーワード「日本人」521を選択すると、領域R2に、第2の原因61及び第2の結果62からなる因果の組を示すノード4Nがさらに表示される。第2の原因61は、「日本人」、「宿泊客」及び「減少」というキーワードを含み、これらに起因する第2の結果62は、「需給」、「緩和」、「方向」、「客室」及び「単価」というキーワードを含んでいる。
 なお、不図示ではあるが、ノード4Nを追加するたびに、領域R2の下に次々と新たな領域と、対応するノード4Nが表示されるとよい。また、後続のノード4Nを追加する際の、末端のノード4Nの選択の仕方は特に限定されず、例えば、領域R1におけるキーワード「日本人」521をユーザが選択した状態で、ボタン55を押下すると、領域R2に新たなノード4Nが表示される等、適宜実施すればよい。また、図8の例では、入力テキストKWを起点として、第1の原因51とその結果である第1の結果52が提示され、さらに第1の結果52に関連するキーワードを第2の原因61として、そのさらなる結果である第2の結果62が提示されている。つまり、原因から結果を追求する方向に因果チェーン4が提示されている。
 さらに、領域R2に対応するノード4Nを選択すると、第2の結果62に関連する関連数値指標RIが、因果チェーン4とともに提示されている。つまり、関連数値指標RIは、入力テキストKWを起点とした連鎖的な因果関係に登場する少なくとも1つの事項に関連している。少なくとも1つの事項とは、ここでは「客室」が相当する。より詳細には、関連数値指標RIは、因果チェーン4に登場する少なくとも1つのノード4Nに関連している。また、このような画面5の生成にあたって、表示制御部334は、因果チェーン4及び関連数値指標RIを含む画面を表示させるように制御している。このような態様によれば、因果チェーン4とともに関連数値指標RIが画面にて一覧性を有して表示されるため、ユーザが、因果の流れと関連する関連数値指標RIとを俯瞰的に把握することができる。
 ここで、関連する関連数値指標RIを提示するにあたって、学習済モデルIF2が参照されうる。関連の度合いの判定方法は特に限定されるものではないが、より具体的には例えば、数1に示されるような修正オッズ比に基づいて判定されるとよい。また、関連の度合いに代えて又はこれとともに、関連しない度合いを判定して、関連数値指標RIを提示するように実施してもよい。
Figure JPOXMLDOC01-appb-M000001
 ところで好ましくは、ノード4Nに示される因果の組に含まれるキーワードには、「減少」、「増加」、「大幅」、「小幅」等の大小増減を示すものが含まれ、さらに好ましくは、複数のノード4Nは、数値的増加又は減少を表す極性を有し、関連数値指標RIは、極性に基づいて決定されるとよい。経済事象等を予測するにあたっては、増加する変化や減少する変化等が極めて重要であるため、このような態様によれば、ユーザが入力した入力テキストKW(入力情報)から関連数値指標RIまでの因果チェーン4に沿った極性の変化の方向性を考慮して、ユーザが求める好ましい関連数値指標RIを提示することができ、ユーザフレンドリな仕様となっている。さらに関連数値指標RIを提示するだけではなく、その数値の予測値を得られるように実施してもよい。
 さらに、ユーザが画面5を参照しながら、関連数値指標RIの一例である関連数値指標「客室稼働率」63を選択すると、図9に示されるような関連数値指標RIのグラフを含む画面5'に遷移することが好ましい。画面5'には、関連数値指標「客室稼働率」63に関するグラフが表示されている。このようなグラフをユーザに参照させることで、より直感的にユーザが予測結果を把握することができる。
 このように本実施形態では、経済事象等の因果関係を、テキストデータTXから抽出した因果情報データベースIF1から連鎖的に検索し、波及効果に関連する関連数値指標RIを提示する情報処理システム1が提案されている。このような情報処理システム1による情報処理方法によれば、ユーザがより容易に、種々の経済事象等に関する定量的な予測結果を得ることができる。
4.その他
 本実施形態に係る情報処理システム1に関して、以下のような態様を採用してもよい。
 以上の実施形態では、情報処理システム1の構成として説明したが、コンピュータに情報処理システム1の各ステップを実行させるプログラムが提供されてもよい。このようなプログラムが、ネットワーク11とは切り離されたスタンドアロンのコンピュータにおいて、オフラインで実行可能に構成されてもよい。
 関連数値指標RIを提示するにあたって、指標そのものの提示だけではなく、X日後又はX月以内に、値が上向きに変化するといった、時間変化の予測値が得られるように実施してもよい。
 因果情報データベースIF1等の参照情報を生成する上で選択される、少なくとも1つのテキストデータTXは、異なる言語で記載された複数のテキストデータTXであってもよい。例えば、1回目には、日本語で記載されたテキストデータTXを用いて因果の組の抽出を行い、2回目には、英語や中国等の外国語で記載されたテキストデータTXを用いて因果の組の抽出を行ってもよい。かかる場合、チェーン処理部333は、連鎖的な因果関係を視覚的に示した因果チェーン4を生成し、因果チェーン4は、異なる言語によってそれぞれ記載された因果を示す複数のノード4Nを含むこととなる。すなわち、異なる言語であっても、共通する概念や、同じ意味のキーワードを抽出することで、それぞれの異なる言語において抽出されたノード4Nの類似度を比較することができる。このような態様によれば、言語の壁を超えた、より普遍的な因果チェーン4を生成することができる。
 本実施形態では、末端のノード4Nに関連する新たなノード4Nを因果情報データベースIF1に基づいて探索及び決定する旨を説明したが、末端のノード4Nだけでなく、他のノード4Nを参照するように実施してもよい。例えば、ノード412から見たノード411又はノード413の如く、同じ階層に並列に位置する他のノード4Nを参照してもよいが、特に好ましくは、チェーン処理部333が、所定の因果チェーン4'における複数のノード4Nに対して重み付けを行い、この重み付けされた評価値に基づいて、後続する新たなノード4Nを決定するとよい。このような態様によれば、直前のノード4Nだけではなく、これまでの因果の流れを踏襲した新たなノード4Nが提案されるため、ユーザは、より精度の高い予測を行うことができる。
 本実施形態、特に、図8においては、原因から結果を追求する方向に因果チェーン4が提示されているが、結果に関する入力テキストKWをユーザに入力させ、結果からその原因を追求する方向に因果チェーン4を提示するように実施してもよい。図10及び図11は、ユーザに視認される表示部24の画面7の態様別例を示した図である。
 図10では、ユーザは、画面7を参照しながら、入力テキストKWとして、「感染症」を入力欄70に入力している。この状態で、ユーザが検索アイコン701を押下すると、入力欄70の右側に、ノード81、ノード82及びノード83と、複数のノード4Nが提示される。さらに、ユーザがボタン751を押下すると、図11に示されるように、ノード821、ノード822及びノード823が、さらに右側に追加される。追加されたノード821、ノード822及びノード823は、ノード82に後続するノード4Nであることが視認可能に、曲線で結ばれている。また、ユーザがボタン752を押下すると、ノード81、ノード82及びノード83と並列関係を有する他のノード4Nが表示されてもよい。
 ここで、図10及び図11に示されるように、画面7には、入力欄70の下方に、選択ボタン76及び選択ボタン77が設けられている。ユーザが選択ボタン76を選択した場合、原因から結果を追求する方向に因果チェーン4を表示させることができる。不図示ではあるが、ユーザが選択ボタン77を選択した場合、結果から原因を追求する方向に因果チェーン4を表示させることができる。なお一般に、1つの事項を原因として複数の結果が生じることとなるため、原因から結果を追求する方向の提示は順問題と考えられる。逆にいえば、結果から原因を追求する方向の提示は逆問題となり、より精度の高い原因予測を行うためには、類似度の判定に加えて、別途制約条件を設けることが好ましい。かかる制約条件は特に限定されず、重み付けの値を変えたり、後続のノード4Nを制限する等が適宜想定されうる。
 本実施形態において、因果チェーン4は、因果情報データベースIF1の情報が充実する限り、理論上は、略際限なく追加することができるが、ノード4Nの数が増えるにつれて、ユーザが当初入力した入力テキストKWとの関連性が薄くなることが想定される。これを防止するために、1つの因果チェーン4に含むことができるノード4Nの階層を制限するように実施してもよい。制限される階層数は、特に限定されるものではないが、具体的には例えば、1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20層であり、ここで例示した数値の何れか2つの間の範囲内であってもよい。より好ましくは、1~5層であり、特に好ましくは、3層であるとよい。
 本実施形態では、入力テキストKWを起点として、因果チェーン4及び関連数値指標RIを生成し、これらが提示される例を説明したが、関連数値指標RIを入力として、これに関連する因果のノード4Nを含む因果チェーン4が提示されるように実施してもよい。換言すると、取得部331は、取得ステップとして、ユーザから受け付けた数値指標(関連数値指標RI)と、予め設定された参照情報IFとを取得する。参照情報IFは、複数のキーワードと数値指標とを関連付けた情報であり、例えば学習済みモデルIF2である。そして、制御部33は、抽出ステップとして、取得された、数値指標及び参照情報に基づいて、キーワードを含む連鎖的な因果関係である因果チェーン4を抽出する。
 さらに、次に記載の各態様で提供されてもよい。
 前記情報処理システムにおいて、表示制御ステップをさらに実行するように構成され、前記表示制御ステップでは、抽出された前記関連数値指標を含む画面を表示させるように制御する、もの。
 前記情報処理システムにおいて、前記参照情報は、複数のキーワードと前記関連数値指標とを関連付けた情報をさらに含む、もの。
 前記情報処理システムにおいて、前記関連付けた情報は、前記複数のキーワードと、前記関連数値指標との相関性を予め機械学習させた学習済モデルである、もの。
 前記情報処理システムにおいて、チェーン処理ステップをさらに備え、前記チェーン処理ステップでは、連鎖的な因果関係を視覚的に示した因果チェーンを生成し、ここで前記因果チェーンは、前記参照情報における因果を示す複数のノードを含み、前記表示制御ステップでは、前記因果チェーン及び前記関連数値指標を含む画面を表示させるように制御し、前記関連数値指標は、前記因果チェーンに登場する少なくとも1つのノードに関連する、もの。
 前記情報処理システムにおいて、前記複数のノードは、互いに類似度を算出可能に構成され、前記因果チェーンは、直列された複数のノードを含む所定の因果チェーンを含み、前記チェーン処理ステップでは、前記所定の因果チェーンの末端のノードに後続する新たなノードを追加し、ここで追加された前記新たなノードは、前記末端のノードと、所定値以上の類似度を有する、もの。
 前記情報処理システムにおいて、前記チェーン処理ステップでは、前記所定の因果チェーンにおける複数のノードに対して重み付けを行い、この重み付けされた評価値に基づいて、後続する新たなノードを決定する、もの。
 前記情報処理システムにおいて、前記複数のノードは、数値的増加又は減少を表す極性を有し、前記関連数値指標は、前記極性に基づいて決定される、もの。
 前記情報処理システムにおいて、参照情報生成ステップをさらに備え、前記取得ステップでは、少なくとも1つのテキストデータを取得し、前記参照情報生成ステップでは、前記テキストデータに基づいて複数の因果を抽出することで、前記参照情報を生成する、もの。
 前記情報処理システムにおいて、チェーン処理ステップをさらに備え、前記少なくとも1つのテキストデータは、異なる言語で記載された複数のテキストデータであり、前記チェーン処理ステップでは、連鎖的な因果関係を視覚的に示した因果チェーンを生成し、ここで前記因果チェーンは、前記異なる言語によってそれぞれ記載された因果を示す複数のノードを含む、もの。
 情報処理システムであって、次の各ステップがなされるようにプログラムを実行可能なプロセッサを備え、取得ステップでは、ユーザから受け付けた数値指標と、予め設定された参照情報とを取得し、ここで前記参照情報は、複数のキーワードと前記数値指標とを関連付けた情報で、抽出ステップでは、取得された、前記数値指標及び前記参照情報に基づいて、前記キーワードを含む連鎖的な因果関係を抽出する、もの。
 情報処理方法であって、前記情報処理システムの各ステップを備える、方法。
 プログラムであって、コンピュータに前記情報処理システムの各ステップを実行させる、もの。
 もちろん、この限りではない。
 最後に、本発明に係る種々の実施形態を説明したが、これらは、例として提示したものであり、発明の範囲を限定することは意図していない。当該新規な実施形態は、その他の様々な形態で実施されることが可能であり、発明の要旨を逸脱しない範囲で、種々の省略、置き換え、変更を行うことができる。当該実施形態やその変形は、発明の範囲や要旨に含まれるとともに、特許請求の範囲に記載された発明とその均等の範囲に含まれるものである。
1    :情報処理システム
11   :ネットワーク
2    :ユーザ端末
20   :通信バス
21   :通信部
22   :記憶部
23   :制御部
24   :表示部
25   :入力部
3    :サーバ
30   :通信バス
31   :通信部
32   :記憶部
33   :制御部
331  :取得部
332  :参照情報生成部
333  :チェーン処理部
334  :表示制御部
4    :因果チェーン
4'   :所定の因果チェーン
4N   :ノード
41   :ノード
411  :ノード
412  :ノード
413  :ノード
42   :ノード
421  :ノード
422  :ノード
423  :ノード
43   :ノード
431  :ノード
432  :ノード
433  :ノード
5    :画面
5'   :画面
50   :入力欄
501  :検索アイコン
51   :第1の原因
52   :第1の結果
521  :キーワード「日本人」
55   :ボタン
61   :第2の原因
62   :第2の結果
63   :関連数値指標「客室稼働率」
7    :画面
70   :入力欄
701  :検索アイコン
751  :ボタン
752  :ボタン
76   :選択ボタン
77   :選択ボタン
81   :ノード
82   :ノード
821  :ノード
822  :ノード
823  :ノード
83   :ノード
IF   :参照情報
IF1  :因果情報データベース
IF2  :学習済モデル
KW   :入力テキスト
R1   :領域
R2   :領域
RI   :関連数値指標
TX   :テキストデータ

Claims (13)

  1. 情報処理システムであって、
     次の各ステップがなされるようにプログラムを実行可能なプロセッサを備え、
     取得ステップでは、ユーザから受け付けた入力情報と、予め設定された参照情報とを取得し、ここで前記参照情報は、複数の、原因と結果とを関連付けた因果を含む情報で、
     抽出ステップでは、取得された、前記入力情報及び前記参照情報に基づいて、関連数値指標を抽出し、ここで前記関連数値指標は、前記入力情報に含まれるキーワードを起点とした、連鎖的な因果関係に登場する少なくとも1つの事項に関連する、もの。
  2. 請求項1に記載の情報処理システムにおいて、
     表示制御ステップをさらに実行するように構成され、
     前記表示制御ステップでは、抽出された前記関連数値指標を含む画面を表示させるように制御する、もの。
  3. 請求項1又は請求項2に記載の情報処理システムにおいて、
     前記参照情報は、複数のキーワードと前記関連数値指標とを関連付けた情報をさらに含む、もの。
  4. 請求項3に記載の情報処理システムにおいて、
     前記関連付けた情報は、前記複数のキーワードと、前記関連数値指標との相関性を予め機械学習させた学習済モデルである、もの。
  5. 請求項2~請求項4の何れか1つに記載の情報処理システムにおいて、
     チェーン処理ステップをさらに備え、
     前記チェーン処理ステップでは、連鎖的な因果関係を視覚的に示した因果チェーンを生成し、ここで前記因果チェーンは、前記参照情報における因果を示す複数のノードを含み、
     前記表示制御ステップでは、前記因果チェーン及び前記関連数値指標を含む画面を表示させるように制御し、
     前記関連数値指標は、前記因果チェーンに登場する少なくとも1つのノードに関連する、もの。
  6. 請求項5の何れか1つに記載の情報処理システムにおいて、
     前記複数のノードは、互いに類似度を算出可能に構成され、
     前記因果チェーンは、直列された複数のノードを含む所定の因果チェーンを含み、
     前記チェーン処理ステップでは、前記所定の因果チェーンの末端のノードに後続する新たなノードを追加し、ここで追加された前記新たなノードは、前記末端のノードと、所定値以上の類似度を有する、もの。
  7. 請求項6に記載の情報処理システムにおいて、
     前記チェーン処理ステップでは、前記所定の因果チェーンにおける複数のノードに対して重み付けを行い、この重み付けされた評価値に基づいて、後続する新たなノードを決定する、もの。
  8. 請求項5~請求項7の何れか1つに記載の情報処理システムにおいて、
     前記複数のノードは、数値的増加又は減少を表す極性を有し、
     前記関連数値指標は、前記極性に基づいて決定される、もの。
  9. 請求項1~請求項8の何れか1つに記載の情報処理システムにおいて、
     参照情報生成ステップをさらに備え、
     前記取得ステップでは、少なくとも1つのテキストデータを取得し、
     前記参照情報生成ステップでは、前記テキストデータに基づいて複数の因果を抽出することで、前記参照情報を生成する、もの。
  10. 請求項9に記載の情報処理システムにおいて、
     チェーン処理ステップをさらに備え、
     前記少なくとも1つのテキストデータは、異なる言語で記載された複数のテキストデータであり、
     前記チェーン処理ステップでは、連鎖的な因果関係を視覚的に示した因果チェーンを生成し、ここで前記因果チェーンは、前記異なる言語によってそれぞれ記載された因果を示す複数のノードを含む、もの。
  11. 情報処理システムであって、
     次の各ステップがなされるようにプログラムを実行可能なプロセッサを備え、
     取得ステップでは、ユーザから受け付けた数値指標と、予め設定された参照情報とを取得し、ここで前記参照情報は、複数のキーワードと前記数値指標とを関連付けた情報で、
     抽出ステップでは、取得された、前記数値指標及び前記参照情報に基づいて、前記キーワードを含む連鎖的な因果関係を抽出する、もの。
  12. 情報処理方法であって、
     請求項1~請求項11の何れか1つに記載の情報処理システムの各ステップを備える、方法。
  13. プログラムであって、
     コンピュータに請求項1~請求項11の何れか1つに記載の情報処理システムの各ステップを実行させる、もの。
PCT/JP2022/015868 2021-04-12 2022-03-30 情報処理システム、情報処理方法及びプログラム WO2022220119A1 (ja)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2021067163A JP2022162364A (ja) 2021-04-12 2021-04-12 情報処理システム、情報処理方法及びプログラム
JP2021-067163 2021-04-12

Publications (1)

Publication Number Publication Date
WO2022220119A1 true WO2022220119A1 (ja) 2022-10-20

Family

ID=83640666

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/015868 WO2022220119A1 (ja) 2021-04-12 2022-03-30 情報処理システム、情報処理方法及びプログラム

Country Status (2)

Country Link
JP (1) JP2022162364A (ja)
WO (1) WO2022220119A1 (ja)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017026303A1 (ja) * 2015-08-12 2017-02-16 国立研究開発法人情報通信研究機構 未来シナリオ生成装置及び方法、並びにコンピュータプログラム

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017026303A1 (ja) * 2015-08-12 2017-02-16 国立研究開発法人情報通信研究機構 未来シナリオ生成装置及び方法、並びにコンピュータプログラム

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
KIYOSHI IZUMI, HIROKI SAKAJI: "Economic Causal Chain Search System and its Application", 22ND SIG-FIN CONFERENCE OF THE JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE (SIG-FIN); [ONLINE], vol. 22, no. 022-15, 3 March 2019 (2019-03-03), pages 1 - 4, XP009540313 *

Also Published As

Publication number Publication date
JP2022162364A (ja) 2022-10-24

Similar Documents

Publication Publication Date Title
US11500524B2 (en) Automatic generation of preferred views for personal content collections
CN106897350B (zh) 引导渐进搜索系统和方法
US20210157977A1 (en) Display system, program, and storage medium
WO2012116287A1 (en) Methods for electronic document searching and graphically representing electronic document searches
US20200034738A1 (en) Computer System and Method of Presenting Information Related to Basis of Predicted Value Output by Predictor
CN110114755A (zh) 示例编程中的行为特征使用
JP7281024B1 (ja) 求職者検索システム、情報処理方法及びプログラム
EP3683744A1 (en) Computer system and method of presenting information useful for achieving purposes related to object
WO2016141045A2 (en) Detection and visualization of temporal events in a large-scale patient database
CN108604332A (zh) 用于生成预测的系统和方法
JPWO2011001584A1 (ja) 情報分類装置、情報分類方法及び情報分類プログラム
JP2007011604A (ja) 不具合診断システム及びプログラム
US20110055758A1 (en) Smart navigator for productivity software
Solainayagi et al. Trust discovery and information retrieval using artificial intelligence tools from multiple conflicting sources of web cloud computing and e-commerce users
WO2022220119A1 (ja) 情報処理システム、情報処理方法及びプログラム
KR20210043281A (ko) 소셜미디어 빅데이터 분석을 통한 커스텀 위젯의 시각적 제공 방법 및 이를 수행하는 서버
JP6647704B2 (ja) 検索候補語表示装置、検索候補語表示方法及び検索候補語表示プログラム
US10296623B2 (en) Document curation
KR101985014B1 (ko) 탐색적 데이터 시각화 시스템 및 그 방법
CN112783410B (zh) 信息处理方法、介质、装置和计算设备
US20120016890A1 (en) Assigning visual characteristics to records
JPWO2012101702A1 (ja) UI(UserInterface)作成支援装置、UI作成支援方法及びプログラム
EP3942499A1 (en) Analyzing resumes and highlighting non-traditional resumes
JP7492488B2 (ja) 提供装置、提供方法及び提供プログラム
US20240152554A1 (en) System and methods for generating contextual graphs and integrated uses thereof

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22788044

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22788044

Country of ref document: EP

Kind code of ref document: A1