JP5212604B2 - Risk detection system, risk detection method and program thereof - Google Patents

Description

  The present invention relates to a risk detection system, a risk detection method, and a program thereof, and in particular, a user can detect a risk that is latent in a project by extracting and presenting information related to project risk from atypical text information such as minutes. The present invention relates to a risk detection system, a risk detection method, and a program thereof.

  In recent years, there are projects of various scales and types, and in order for those projects to succeed, a wealth of knowledge and experience of project managers is required. In particular, the key to success is how early the project manager can detect and take measures against the potential risks in the project. However, in a large-scale and complicated project, there is a problem that it is difficult for the project manager to grasp all the situation of the project and to detect all the risks that exist in the project.

  Therefore, in order to assist the project manager in detecting risks that are latent in the project, a technique for presenting information related to the project risk from the accumulated data on the project has been researched and developed. Conventionally, research and development of project risk detection technology using numerical data related to projects has been the mainstream, but in recent years, technology for detecting risks using various project-related information other than numerical data such as text information is also available. It has been researched and developed. The following patent documents can be cited as techniques for presenting relevant information on project risk to a user using text information related to a project.

In the invention of Patent Document 1, document data is classified into a risk case document and a general document, and related words in each document are extracted using the co-occurrence intensity. Next, a related word extracted in each of the risk case document and the general document is synthesized to create a word network. When the user inputs a document or keyword describing the contents of a project for which risk detection is desired, the system displays a group of keywords related to the input document or keyword.
JP 2006-23789 A

  However, the above-described problem of the prior art is that it is difficult to determine whether or not the keyword group represents risk information depending on the type of risk only by listing related keywords. The reason is that conventionally, a group of words having a high co-occurrence strength is only presented as risk-related words of the input word, and risk information and information that is not so are not classified.

  Therefore, the present invention has been invented in view of the above problems, and its purpose is to provide a project risk detection system, a project risk detection method, and a program capable of presenting the risk information of the project from the data related to the project to the user. Is to provide.

The present invention that solves the above-described problem is a risk detection system, and is an intention determination unit that determines an intention of a statement included in a text sentence related to a project based on a predetermined relationship between a text expression and an intention. And a risk sentence extraction condition comprising a topic determination means for determining a topic of a speech included in the text sentence based on a predetermined text expression and a topic, and a plurality of intentions of speech and topic appearance patterns Is stored in the extraction condition storage means, the intention determined by the intention determination means and the topic determined by the topic determination means, and the risk sentence extraction conditions stored in the extraction condition storage means , remarks specific to identify the utterance corresponding to the intention and topic matching the risk extraction condition as a risk remarks And having a stage.

The present invention for solving the above-mentioned problems is a risk detection system for determining an intention and topic of an input speech, and a risk sentence comprising the determined intention and topic and a plurality of speech intentions and topic appearance patterns. It has a risk judgment means which judges whether the said utterance is a risk utterance from extraction conditions .

The present invention for solving the above problems is a risk detection method, in which a data processing device has information indicating a relationship between a predetermined text expression and an intention, and a relationship between a predetermined text expression and a topic. The intention and topic of the speech included in the input text sentence are determined based on the stored information, and the data processing device determines the determined intention and topic, and the intention of the plurality of speech and the appearance pattern of the topic. And a statement corresponding to the intention and topic that match the extraction condition is identified as a risk statement .

The present invention for solving the above problems is a program for causing an information processing apparatus to detect a risk utterance included in a text sentence , and relates to a project based on a predetermined relationship between text expression and intention. A process for determining an intention of a speech included in a text sentence, a process for determining a topic of a speech included in the text sentence based on a relationship between a predetermined text expression and a topic, an intention of a plurality of comments, and A process for collating the risk sentence extraction condition composed of the appearance pattern of the topic with the determined intention and topic and identifying a statement corresponding to the intention and topic that matches the risk sentence extraction condition as a risk statement The apparatus is characterized by being executed.

  The first effect is that when a text sentence related to a project is input, only the text sentence related to the project risk can be presented as risk information. As a result, it is possible to support the user to detect and monitor the risk in the project. The reason is to determine the intention and topic of the input text sentence, and classify the input text sentence as risk information based on the extraction condition based on the combination thereof.

  The second effect is that when a text sentence related to a project is inputted, the text sentence related to the project risk can be classified and presented for each type of cause of the project risk. As a result, the user can know what type of risk causes the presented risk information is related to, and therefore it is easy to take measures for risk avoidance. The reason is to classify the extracted text sentence for each extraction condition based on the combination of the intention of the text sentence and the topic.

  The best mode for carrying out the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, the first embodiment of the present invention includes a data processing device 1 that operates under program control and a storage device 2 that stores information.

  The storage device 2 includes a project related information storage unit 21, a risk information storage unit 22, an intention expression dictionary storage unit 23, a topic expression dictionary storage unit 24, and a high risk statement specifying rule storage unit 25.

  The project related information storage unit 21 stores information related to a project that is a management target of the user. Information related to a project is often accumulated by a project management tool or the like, and examples thereof include a plan, a work performance table, minutes, bug density, and the like.

  An example of data storage in the project related information storage unit 21 is shown in FIG. In FIG. 2, the utterance of the minutes is taken as an example as a text sentence. For each sentence, a column (field) for assigning the type of intention and topic to be described later is prepared. Initially, no data is entered in this column (FIG. 2A), but the intention and topic information of each statement determined by the statement determination unit 11 and topic determination unit 12 described later is “ It is stored in each column of “Intention type” and “Topic type” ((b) of FIG. 2). Depending on the statement, a plurality of intentions and topics may be assigned to one statement.

  The risk information storage unit 22 stores a statement sentence determined as a statement related to a high risk of the project in the input statement by the high risk specifying unit 13 described later. FIG. 6 shows an example of storage in the risk information storage unit 22. A statement surrounded by a double line is a statement corresponding to a high-risk statement. It should be noted that it is desirable that the high-risk utterances extracted from the minutes are given the target project name, the minutes ID of the extraction source, and the type of matched rule as metadata. This is because the user (project manager) can grasp the basis on which the extracted high-risk remarks are extracted and can analyze the contents of the risk and the situation of the project. If a database as shown in FIG. 6 is used, it is possible to check, for example, high-risk messages that only apply to a certain rule or project, or check the number of high-risk messages between minutes. In addition, speaker information, previous and subsequent speech information, and the like may be provided.

  The intention expression dictionary storage unit 23 is a dictionary of utterance intention types and features on the surface character string. FIG. 3 shows an example of the intention expression dictionary. Examples of intent types include “warning”, “impossible”, “suggestion”, “question”, “forecast”, “duty”, “request”, “dissatisfaction”, “denial”, “apology”, “ Report "etc. It is a dictionary that focuses mainly on the features of sentence end expressions for each type of intention. The left column indicates the type of intention, and the expressions in the sentence corresponding to each intention are listed in the right column. For example, if the end of a text sentence is one of "I can't", "I can't do it", "I can't do it", "I can't do it", or "I can't do it", the intention of the text sentence is "Illegal" It can be determined that it is possible.

  The topic expression dictionary storage unit 24 is a dictionary of utterance topic types and their features on the surface character string. FIG. 4 shows an example of the topic expression dictionary. The types of topics include “bug”, “schedule”, “work area”, “human resource”, “cost”, “health”, “material”, and the like. The left column of the table in FIG. 4 represents the type of topic, and expressions in the sentence corresponding to each topic are listed in the right column. For example, if there is any of “Team”, “Group”, “Member”, “Assignment”, “Share”, “Factor”, “Human Resource” in the text, the topic of the text is “People” It can be judged that it is related to “resources”.

  The high-risk utterance specific rule storage unit 25 rules (patterns) combinations of intentions and topics common to utterances related to the high risk of the project, and stores the rules. FIG. 5 shows an example of a high-risk statement specifying rule. The high risk in the project is a situation such as “A problem that is difficult to solve” or “A problem that has occurred is still unresolved”. The rule shown in FIG. Is a pattern of intentions and topics commonly seen in. Problems that are difficult to solve in a project are often mixed with a variety of factors, and this rule is a pattern of intentions and topic features found in statements about such problems.

  For example, the rule “R1” represents a rule that a statement whose intention is other than “report” and four or more topics are assigned is a high-risk statement.

  The rule of “R2” is a pattern in which the intention is “warning”, “impossible”, “denial”, and the topic is different between the target utterance and the immediately following utterance. This is a pattern of the intention of the flow of speech and the characteristics of a topic in a situation in which a serious problem has occurred, but has moved to another topic unsolved.

  In the rule of “R3”, the intention is “question”, the intention of the immediately following statement is “warning”, “impossible”, “denial”, and the topic of the preceding and following statements is common. It is. This is a pattern of the intention of the flow of speech and the characteristics of the topic in a situation where there is no unresolved or unanswered question in response to a question (problem) about the subject matter.

  Moreover, you may make it give the priority according to a risk to each rule. For example, in the above-mentioned example, the rule “R2” is a pattern of a situation in which a serious problem has occurred but the topic has been unsolved and has moved to another topic. It is a problem that needs to be discussed. Therefore, the priority of the rule “R2” is set to “1”. Furthermore, the “R3” rule is a pattern of situations where there are no unresolved or unanswered questions for questions (problems) on the subject matter, and some solution must be considered. The priority of the rule “R3” is “2”. Finally, the priority of the rule “R1” is set to “3”. In this way, if priority is given to each rule according to risk, extracted risk statements to be described later can be sorted by risk importance.

  The data processing apparatus 1 includes a statement intention determination unit 11, a statement topic determination unit 12, and a high-risk statement specifying unit 13.

  The statement sentence intention determination means 11 determines the type of statement statement intentions such as a plan document, work performance table, minutes, and bug density stored in the project related information storage unit 21. The determination may be performed by preparing an intention expression dictionary (in this embodiment, an intention expression dictionary stored in the intention expression dictionary storage unit 23) in which characteristic expressions are registered for each intention and matching the dictionary. It is good to prepare training data to which the type of intention of the correct answer is given in advance, and for each intention, obtain a classification function using the surface character string of the word, part of speech, etc. as a feature quantity, and use that function to determine the correct answer. The intention of a sentence that has not been granted may be determined. Moreover, you may perform using the technique of machine learning.

  For example, in the case of the statement of the minutes of the project meeting in FIG. 2 (“Transition to the T3 process on 10/13 is dangerous”), the dictionary of FIG. 3 which is an example of the intention expression dictionary storage unit 23 is used. The word “dangerous” is registered to indicate the intention of “warning”, and since the expression “dangerous” is included in the above statement, the intention of the statement is “warning” by matching with the intention expression dictionary. ". The determined intention information is stored in, for example, the “intention type” column of the project text information storage unit 21 and stored in a format as shown in FIG.

  The topic determination unit 12 of the comment sentence determines the topic type of the comment sentence. The determination may be performed by matching with a topic expression dictionary (in this embodiment, a topic expression dictionary stored in the topic expression dictionary storage unit 24) as shown in FIG. 4 in which characteristic expressions are registered for each topic. It is good to prepare training data in which the types of correct topics are assigned in advance, and for each topic, obtain a classification function using the surface character string of the word, part of speech, etc. as a feature quantity, and use that function to determine the correct answer. You may determine the topic of the sentence which is not given. Moreover, you may perform using the technique of machine learning.

  For example, in the case of the statement of the minutes of the project meeting in Fig. 2 ("Transition to the T3 process on 10/13 is dangerous"), if the topic expression dictionary is used, the statement will be expressed as a date expression "10/13" It can be determined that the topic is “Schedule”. The determined topic information is stored in, for example, the “topic type” column of the project text information storage unit 21, and together with the intention information determined by the intention determination means 11 of speech, the format as shown in FIG. Remember it.

  The high-risk utterance determination unit 13 specifies an utterance related to the high risk of the project from each utterance sentence in which the intention and the topic are determined, and stores the utterance in the risk information storage unit 22. The determination of which intention and topic appearance pattern is related to a high-risk utterance is prepared as a high-risk utterance identification rule as shown in FIG. The determination may be made using a technique such as machine learning.

  As an example in the case of using the high-risk utterance specifying rule, a utterance in which the appearance pattern of the intention and the topic corresponds to each rule in FIG.

  For example, the statement of the minutes of the project meeting in Fig. 2 (3D problem, the J side schedule was delayed by about 2 weeks due to the delay of the J side startup schedule, and the test equipment and evaluation man-hours increased. 5), the intent is “forecast” and the topics are “schedule”, “material”, “personnel”, “cost”, so rule R1 “intention is other than“ report ”” in FIG. Since the topic is “mixed with 4 or more topics”, it can be extracted as a high-risk statement.

  In addition, in the case of the statement of the minutes of the project meeting in Figure 6 (please understand that delays beyond this are considered critical for the project), the intention is “warning” and “request” It contains a “warning”, the topic is “schedule”, and the topic of the immediately following remark (I would like to collect members other than the camera if necessary) changed to “personnel” Therefore, it matches the rule R2 “intention is any one of“ warning ”,“ impossible ”,“ denial ”” in FIG. 5 and “the topic of the next speech changes”. Can be extracted as a risk statement.

  In addition, two statements in the minutes of the project meeting in Figure 6 (“There was a talk that CE is the only certification to be obtained, what kind of system will be promoted in the future?” ”)”), The intent of the previous statement is “question”, the topic is “work”, “personnel”, the intent of the immediately following statement is “impossible”, and the topic is “ Work ". In this case, since the intention of the previous statement is “question” and the intention of the later statement is “impossible”, and the topic of the two statements is common, the rule R3 “during two statements” in FIG. The intention of the previous statement is “Question”, and the intention of the subsequent statement is “Warning”, “Impossible”, or “Negation” ”, and“ The topic of two statements is common ” Therefore, it can be extracted as a high-risk statement.

  In addition, if the priority according to the risk is provided to each rule, it is also possible to sort the extracted risk statement according to the priority according to the risk to each rule.

  Next, the overall operation of the present embodiment will be described in detail with reference to FIG. 1 and FIG.

  The text sentence stored in the project related information storage unit 21 is supplied to the intention determination means 11 of the statement sentence. Take the utterance of the minutes as an example of text. The utterance of the minutes is, for example, a text sentence such as “Transition to the T3 process on 10/13 is dangerous.” The statement intention determination means 11 determines the intention included in each of the supplied statements, assigns the intention, and stores the intention in the project related information storage unit 21 (step S1 in FIG. 7). For the determination of the intention, an intention expression dictionary as shown in FIG. 3 or a classification technique such as machine learning may be used. If the intention expression dictionary of FIG. 3 is used, the intention of the above-described utterance example is determined to be “warning”.

  Next, the topic determination unit 12 determines the topic of each message (step S2 in FIG. 7). For topic determination, a topic expression dictionary as shown in FIG. 4 stored in the topic expression dictionary storage unit may be used, or a classification technique such as machine learning may be used. If the topic expression dictionary of FIG. 4 is used, the above-mentioned utterance example is determined to be “schedule”.

  Next, the high risk statement specifying means 13 determines whether the statement is related to the high risk of the project using the intention and topic information given to each statement (step S3 in FIG. 7). . The determination as to whether or not the input utterance is a high-risk utterance may use the rules shown in FIG. 5 stored in the high-risk utterance specific rule storage unit 25, or one of the utterances corresponding to the high-risk utterance. A machine learning technique for classifying whether or not the input statement is related to a risk may be used by previously learning the part as training data.

  As a result of this determination, if the sentence is a sentence related to risk, it is stored in the risk information storage unit 22 (step S5 in FIG. 7), and if the sentence is not a sentence related to risk, The process ends (step S4 in FIG. 4).

  In the case of determination using the rule of FIG. 5, in the above-described utterance example, the intention is “warning”, the topic is “schedule”, and does not correspond to the high-risk utterance specifying rule in FIG. The process ends without being stored in 22. On the other hand, in the second remark from the top of FIG. 2, the intention is “forecast” and the topics are “schedule”, “material”, “personnel”, and “cost”, which correspond to the rule “R1” in FIG. . Therefore, this statement is stored in the risk information storage unit 22.

  Next, the effect of this embodiment will be described.

  In the present embodiment, an intention and a topic included in a text sentence related to a project such as a statement of a minutes are determined, and a combination thereof is used to determine whether the statement represents a high risk in the project. Therefore, only the text information related to the high risk of the project in the project related text information can be presented to the project manager, and the grounds extracted as the high risk from the combination of the intention and the topic can be presented. As a result, the project manager can confirm and manage, as project risk information, a sentence associated with a high risk among text sentences associated with the project and the basis thereof.

  A second embodiment will be described.

  In the first embodiment described above, the example of determining the risk in sentence units has been described, but in the second example, the risk is determined in paragraph units.

  Referring to FIG. 8, in the second embodiment of the present invention, there is a paragraph cutout unit 100 for dividing information related to a project stored in the project related information storage unit 21, for example, minutes, etc. into paragraphs. , A speaker determination means 101 for determining a speaker of a statement is newly provided. Further, instead of the high risk statement specifying rule storage unit 25, a project risk paragraph rule storage unit 201 that stores rules for extracting project risks in units of paragraphs, and a high risk statement determination unit 13, a project risk paragraph Project risk paragraph determining means 102 for extracting a paragraph with a project risk using the rules stored in the rule storage unit 201, and a project risk paragraph storage unit for storing the extracted project risk paragraph instead of the risk information storage unit 22 202. Since the other components are the same as those in the first embodiment described above, detailed description thereof is omitted.

  The project related information storage unit 21 stores a stored document group of projects described in natural language sentences. Specific examples of documents include electronic project minutes, e-mails, texts of meetings transcribed using voice recognition, chat and instant message logs, etc., recorded as is without being summarized or omitted. It is what has been. An example of data storage in the project related information storage unit 21 is shown in FIG. In FIG. 9, the target text document is the project minutes. The initial state in which data is input includes project name, date (meeting date), and document ID information for each minutes, and ID statements are managed by ID.

  The paragraph cutout unit 100 divides the project minutes as an analysis target document into paragraphs as topic units. Paragraph division may use clue expressions that represent paragraph breaks, such as blank lines and heading symbols (“•” and “■”), or prepare training data for paragraph breaks and prepare the sentence at the beginning of the paragraph. Alternatively, a machine learning method obtained using a classification function may be used. FIG. 10 shows an example of storage after the project minutes are divided into paragraphs which are topical units by the paragraph cutout means 101 and the paragraph IDs are given. Sentences belonging to the same paragraph are assigned the same paragraph ID.

  The intention determination unit 11, the topic determination unit 12, and the speaker determination unit 101 read a document registered in the project related information storage unit 21 one by one, and retrieve information such as a statement intention, a topic, and a speaker. The extracted information is registered in the project related information storage unit 21.

  As described in the first embodiment, the speech intent and topic extraction method can be realized by matching with an expression dictionary in which a surface character string characteristic of each intention and topic is converted into a dictionary. . It can also be realized by using a machine learning method for determining the intention and topic of each utterance by obtaining a classification function of each intention and topic from the training data.

  In addition, the speaker determination unit 101 can extract speaker information by using expressions that are clues representing the speaker, such as parenthesis expression and arrow expression.

  Hereinafter, an implementation example of extraction of intention, topic, and speaker information using an expression dictionary will be described. In the expression dictionary, sentence end expressions such as “can't (impossible)” and “probably (guess)”, intention expression dictionary centering on adverbs, noun phrases such as “bug (quality)” and “[numerical value] month (schedule)” Prepare a topic expression dictionary centered on. Also, extraction patterns such as “(“ speaker name ”)” and “→“ speaker name ”” are prepared as a speaker extraction clue dictionary. For example, if there is a statement “The bug of module A cannot be reproduced. (Saito)”, the intention of this statement is “impossible” from the expression “I cannot”, and this statement from the expression “bug” The topic is extracted as “quality”. Further, it is extracted from the expression “(Saito)” that the speaker of this statement is “Saito”, and the information is registered in the project related information storage unit 21.

  FIG. 11 shows an example of registration of an intention, a topic, and a speaker by the intention determination unit 11, the topic determination unit 12, and the speaker determination unit 101.

The project risk paragraph rule storage unit 202 stores a rule for extracting a paragraph including a project risk. In the present invention, the following six examples are shown as project risk situations occurring in scenes such as meetings, e-mails, chats, and instant messages.
(1) Conflicts between members and discrepancies have occurred (2) No concrete response method has been found (3) Clear decision making has not been made (4) Difficult problems have occurred (5) The work to be executed is not executed, the hand is not turned (6) The project member does not have enough knowledge and information about the work content The extraction rule of the project risk paragraph rule storage unit 202 is: This is a rule for extracting a paragraph portion where any one of the above six situations has occurred from a document such as a minutes. The extraction rule here is a rule having attributes such as appearance series, frequency, distribution, topic type, speaker appearance series, etc. of intentions of continuous utterances.

  FIG. 12 shows an example of rules stored in the project risk paragraph rule storage unit 201. In the project risk paragraph rule with ID (A), if any intention of “denial”, “impossible”, or “warning” appears more than M times in the statement intention included in paragraph P, paragraph P is The rule is that it is a project risk paragraph. The project risk paragraph rule whose ID is (B) is a rule that if the intention of “question” appears in the statement intention included in the paragraph P more than N times consecutively, the paragraph P is a project risk paragraph.

For example, a paragraph composed of the following sentences corresponds to the project risk paragraph rule of ID (b).
“Did you submit a report for Bug Y?” [Question, quality, quality risk]
"Is there an explanation of function Z?" [Question, work, work risk]
“Is the response completed?” [Question, work, work risk]
“Please make more information public.” [Request, work, work risk]

  FIG. 13 shows details of the project risk paragraph rule. This is a specific example in the project risk paragraph rules and applicable situations, and the corresponding minutes. The rules are written in natural language for explanation. Hereinafter, specific examples of (A), (B), (C), (D), and (E) in FIG. 13 will be described.

(1) In the case of rule ID (A) A rule of ID (A) is a text sentence in a paragraph, and if the intention of denial, impossible or warning appears more than M times, the paragraph is (1) Paragraph (project risk paragraph) including the project risk corresponding to the situation of (4) (5).

A specific example of ID (A) in FIG. 13 is as follows.
“Is the delivery date ahead of schedule ahead?” [Question]
“I have not heard of that.” [Negative]
“I should have reported it.” [Guess]
“But it ’s not communicated.” [Negation]
“There is also a problem with the transmission efficiency after the report.” [Warning]
In the above-mentioned specific example, many negative statements that are considered to be inconsistent opinions and situation grasps are made on a topic in the meeting, and the project risk situation in which there is a conflict or misunderstanding among members can be read.

(2) In the case of rule ID (B) If the rule of ID (B) includes a text sentence of a question sentence in which a certain paragraph continues N times or more, the paragraph is a project risk paragraph.

A specific example of ID (B) in FIG. 13 is shown below.
“Did you submit a report for Bug Y?” [Question]
“Is there an explanation of function Z?” [Question]
“Is the response completed?” [Question]
“Please make more information public.” [Request]
In the above-mentioned specific example, the question of the members continues to be folded, and the project risk situation in which there is a conflict or misunderstanding among the members can be read.

(3) In the case of rule ID (C) The rule of ID (C) is a paragraph in which the intention of the last statement (text sentence) in the paragraph is one of denial, impossible, warning, obligation, or question. is there.

A specific example of ID (C) in FIG. 13 is shown below.
“It is necessary to select Tool A or Tool B as soon as possible.” [Not applicable]
“But it is not yet a selection stage.” [No]
In the specific example described above, the discussion on the raised problem ends with a statement indicating denial, and it is possible to read the project risk situation in which no specific countermeasures have been found for the problem or no clear decision has been made.

(4) In the case of ID (D) of a rule The rule of ID (D) is a paragraph including a utterance (text sentence) of continuous L times or more of one speaker in the paragraph.

A specific example of ID (D) in FIG. 13 is shown below.
“It's a different story from the customer.” [Mr. A]
“I think they are confused.” [Mr. A]
"Tool B is to be developed" [Mr. A]
“It should be cut out” [Mr. A]
In the above specific example, the remarks of a certain member (Mr. A) continues continuously, and it is possible to read the project risk situation in which the frustration that information sharing is not successful, the conflict between members, and the difference in recognition occur. .

(5) In the case of rule ID (E) The rule of ID (E) is a statement of any of denial, impossible, warning, duty, request, request, or question with no more than K types of topics in the paragraph. It is a paragraph containing (text sentence).

  A specific example of ID (E) in FIG. 13 is shown below.

“Due to the delay of the sub-team start-up schedule, it is necessary to review the delivery date, test equipment and personnel plan, and review the budget.”
[Intention: Obligation, Topic: Schedule, Cost, Personnel, Material] (K = 4)
In the above-described specific example, a large number of topics are mixed in remarks explaining one problem, and a project risk situation in which a problem that is difficult to solve can be read.

  The project risk paragraph extraction means 102 determines whether each paragraph in the document corresponds to any one of the project risk paragraph rules using the information in the project related information storage unit 21 and the project risk paragraph rule storage unit 201. The result is stored in the project risk paragraph storage unit. For example, it is determined by the project risk paragraph extraction means 102 that the paragraph of ID (B) corresponds to the project risk paragraph rule (B), the paragraph of ID (B) is the project risk paragraph, and the corresponding rule is (B ) Is stored in the project risk paragraph storage unit 202.

  The project risk paragraph storage unit 202 stores the project risk paragraph extracted by the project risk paragraph extraction means 102. FIG. 14 shows a data storage example of the project risk paragraph storage unit 202 in the present embodiment. The ID of the corresponding project risk paragraph extraction rule is stored for each ID managed paragraph. By using such a storage method, it is possible to store information on whether or not each paragraph has a project risk, and if so, by which rule the project risk is extracted.

  Next, the operation of the second embodiment for carrying out the present invention will be described in detail with reference to FIGS.

  The document stored in the project related information storage unit 21 is divided into paragraphs by the paragraph cutout unit 100 (S1 in FIG. 15). Then, for each divided paragraph, the intention determination unit 11, the topic determination unit 12, and the speaker determination unit 101 extract the intention, topic, speaker, and the like for each statement sentence in the paragraph (see FIG. 15 S2).

  Next, for each paragraph, the project risk paragraph extraction means 102 determines whether there is a project risk using the project related information storage unit 21 and the project risk paragraph rule storage unit 201 (S3 in FIG. 15). If there is a project risk, it is registered which rule corresponds (S4 in FIG. 15).

  The content of the paragraph Z which consists of a sentence of the above content is demonstrated as a specific example.

Paragraph Z
"Did you submit a report for Bug Y? (Mr. X)"
"Is there an explanation of function Z? (Mr. X)"
"Is the response completed? (Mr. Y)"
“Please release more information. (Mr. Y)”
Assume that the paragraph cutout unit 100 cuts out the paragraph. This paragraph X includes four remarks. The intention determination unit 11, the topic determination unit 12, and the speaker determination unit 101 extract the intention, topic, and speaker information of the statements from these statements (S2). Examples of extraction results are shown below as [Intent, Topic, Speaker] after the text of the example.
“Did you submit a report for Bug Y? (Mr. X)” [Question, Quality, Ms. X]
"Is there an explanation of function Z? (Mr. X)" [Question, work, Mr. X]
"Is the response completed? (Mr. Y)" [Question, work, Mr. Y]
“Please release more information. (Mr. Y)” [Request, work, Mr. Y]
Next, the project risk paragraph extraction means 102 extracts the project risk type of each paragraph using the extraction result and the information in the project risk paragraph rule storage unit 201 (S3). In the above example, the paragraph from the project risk paragraph rule “ID (B) in FIG. 12, if the intention of“ question ”appears more than N consecutive times in the statement intention included in paragraph P, paragraph P is a project risk paragraph”. X is determined to be a project risk paragraph corresponding to the rule (B), and the information is stored in the project risk paragraph storage unit 202 (S4).

  Next, the effect of this embodiment will be described.

  In this embodiment, a paragraph with project risk is expressed by a rule with attributes of the intention and topic of the statement sentence constituting the paragraph, and the speaker's series pattern. It can be extracted in paragraphs.

  Next, a third embodiment of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 16, in the second embodiment of the present invention, in addition to the second embodiment of the present invention, a project related word storage unit 304 for storing words related to a project, and the storage unit Referring to the project related word extracting means 404 for extracting the project related words included in the project risk paragraph, the project related information storage unit 21, the project risk paragraph rule storage unit 201, and the project risk paragraph storage unit 202 to the user. Project risk information display means 405 for displaying information on the project risk.

  FIG. 17 shows an example of data in the project risk paragraph rule storage unit 201 in the present embodiment. The difference from the project risk paragraph rule storage unit 201 in the second embodiment is that the corresponding project risk situation is stored in addition to the rules.

  For example, the project risk status corresponding to the rule ID (A) is (1) conflict between members or recognition difference, (4) difficult problem to solve, (5) work to be performed The information that is not executed or the hand is not turned is stored. In addition, the project risk situation corresponding to rule ID (B) is that (1) there is a conflict or misunderstanding among members, and (6) the member does not have sufficient knowledge and information about the work contents. Information is stored. Further, the project risk situation corresponding to the rule ID (C) stores information that (2) a specific countermeasure for the problem is not found, and (3) a clear decision is not made. Further, the project risk situation corresponding to the rule ID (D) stores (1) information that conflicts and recognition gaps occur between members. The project risk status corresponding to the rule ID (E) stores information that (4) a problem that is difficult to solve occurs.

  The project related word storage unit 304 stores words characteristic of the project. Project-related terms may be prepared as prior knowledge by the user, or may be automatically extracted from the project minutes to be analyzed. The following describes how to automatically extract the minutes from the project minutes. First, the included words are extracted from the minutes of the project using the part of speech as a clue. The word is preferably a noun phrase or a series of unknown words.

  Next, a document unrelated to the project such as a newspaper article is prepared as a filtering document, and contained words are extracted from the document using the part of speech as a clue. Words extracted from the above two documents are compared, and words that appear in the project minutes but do not appear in documents unrelated to the project, such as newspaper articles, can be extracted as project-related words. For example, if words such as “Release”, “Recognition” and “Module” are extracted from the minutes of a project, and words such as “Recognition”, “Economy” and “Economy” are extracted from newspaper articles, two words “Release” and “Module” Are words that appear only in the project minutes, so these are extracted as project-related words.

  The project related word extraction unit 404 extracts the project related words included in the project risk paragraph with reference to the project related word storage unit and stores them in the project risk paragraph storage unit 202.

  FIG. 18 shows a data storage example of the project risk paragraph storage unit 202 in the present embodiment. The difference from the project risk paragraph storage unit 202 (FIG. 14) of the second embodiment is that the project risk paragraph extracted by the project risk paragraph extraction means and the project related words included in the paragraph are stored. is there.

  The project risk display means 405 classifies and displays the project risk paragraphs based on project-related words and topic axes included in the corresponding paragraphs. When the user designates a project to be analyzed, the project risk paragraph of the designated project is designated as a project related word and topic using the project related information storage unit 21, the project risk paragraph rule storage unit 201, and the project risk paragraph storage unit 202. Are classified as a classification axis, and paragraphs with the same classification axis are displayed in association with each other.

  In the example of FIG. 18, the paragraph P01 includes the project related word “module A”, and referring to FIG. 2, the topics corresponding to P01 are “quality” and “schedule”. Classify. Then, the appearance frequency of the paragraph corresponding to each classification axis is calculated and displayed in order from the classification axis with the highest appearance frequency. Instead of a topic, a project risk management item that is a combination of an intention and a topic may be used. Project risk management items are information given to individual statements, and are determined by a combination of intention and topic. For example, the project risk management item of a statement including the intention of “impossible” and the topic of “quality” is “insufficient quality control”. The advantage of using project risk management items is that elements that are more likely to be related to project risk can be used as the classification axis for the project risk paragraph.

  Hereinafter, specific display examples of the project risk display unit 405 will be described with reference to FIGS. 19 and 20. FIG. 19 shows a screen example in which the user sets analysis conditions. Select the project name to be analyzed from the screen using the pull-down menu. After selecting the project, press the “Analyze” button in the center to display the analysis results on a separate screen.

  FIG. 20 shows an example of an analysis result display screen. The upper part of the screen is a part for designating a period to be analyzed. The period is selected using a pull-down menu. When nothing is selected, the period from “Project start” to “Point of analysis” is selected. The analysis result after specifying the target period is displayed by pressing the “Redisplay” button. By using such a screen, the user can interactively obtain information necessary for project risk management while viewing the analysis result. On the left side of the screen, the classification axis corresponding to the project risk paragraph is listed. The number in parentheses immediately after the classification axis is the number of applicable project risk paragraphs. The order of the classification axes is determined by the number of applicable project risk paragraphs.

  For example, in FIG. 18, two project risk paragraphs corresponding to the classification axis of “module A, quality, schedule” are P01 and P03, and one of the project risk paragraphs P03 corresponding to the classification axis of “tool B, quality, schedule”. Therefore, the former is a project risk that should be displayed more preferentially, and is displayed on the screen in order from “module A, quality, schedule”. When the user clicks a classification axis whose details are to be confirmed from the displayed list of classification axes, the corresponding detailed information is displayed on the right side. There are two types of detailed information. One is an item representing the status of the project risk paragraph corresponding to the selected category axis, and the other is the original text of the project risk paragraph corresponding to the selected category axis. When there are multiple paragraphs with the same classification axis, detailed information of all the corresponding paragraphs is displayed.

  The status of the project risk paragraph is displayed using the information in the project risk paragraph rule storage unit 201 as the status represented by the pattern corresponding to the paragraph. For example, the project risk represented by the classification axis of “module A, quality, human resources” has the corresponding pattern (A) (G), so the corresponding project risk status is (1) conflicts between members There are deviations, (4) problems that are difficult to solve, (5) work to be performed is not being performed, and hands are not turning. The text excerpt results will be displayed under the reason for the project risk status item. In the screen example, paragraphs corresponding to the “module A, quality, human resources” classification axis are displayed in the minutes. The original text is displayed in chronological order.

  As a modification of the above display example, there is a method of displaying risk information of a plurality of projects in parallel. When an analysis target project is selected, other projects are recognized as comparative projects. Projects are identified by project name. There is a display method in which the risk information classification axis of the project to be analyzed is listed on the left and the risk information classification axis of the comparison project is simply listed on the right. There is also a method of displaying only those that are not listed as risk information classification axes of the analysis target project. With the latter method, the system can effectively present risk information that was not noticed only by the risk information of the analysis target project to the project manager.

  Next, the operation of the present embodiment will be described in detail with reference to the drawings.

  The operations of the paragraph cutout unit 100, the intention determination unit 11, the topic determination unit 12, the speaker determination unit 101, and the project risk paragraph determination unit 102 in the present embodiment indicated by S1 to S4 in FIG. 21 are the second implementation. Since the operation is the same as that of each of the units 201, 202, and 203, the description is omitted.

  When the project risk paragraph is extracted from the project minutes by the project risk paragraph determination means 102 and registered in the project risk paragraph storage unit 202 (S4), the project related word extraction means 404 includes the project risk paragraph storage unit 202 and the project related word. Referring to the storage unit 104, project related words included in the project risk paragraph are extracted and registered in the project risk paragraph storage unit 202 (S5).

  Finally, the project risk information display means 405 refers to the information in the project related information storage unit 21, the project risk paragraph rule storage unit 201, and the project risk paragraph storage unit 202, and extracts the extracted project risk paragraph as a project related word. Classification / arrangement is performed on the axis composed of topics, etc., and the classification axis and the corresponding paragraph text are linked and displayed to the user (S6).

  Next, effects of the third exemplary embodiment of the present invention will be described.

  In the third embodiment of the present invention, words characteristic of the project included in the extracted project risk paragraph are extracted as project-related words, and the corresponding project risk paragraph is classified / organized by those and topics. By displaying to the user, the risk information of the project to be analyzed can be displayed to the user in an easy-to-understand manner. In addition, by comparing the risk information of other projects with the classification axis as a clue, it is possible to present the project risk information that the user has not noticed.

  INDUSTRIAL APPLICABILITY According to the present invention, the present invention can be applied to uses such as a project risk management / detection support device that detects a project risk from project-related information, and a device that supports the planning of risk avoidance measures by investigating the cause of an existing risk . Further, the present invention can also be applied to an application such as an information sharing device for sharing the project status among the parties concerned with the project.

It is a block diagram which shows the structure of the 1st Embodiment of this invention. It is a figure which shows the example of a data storage of the project relevant information storage part. It is a figure which shows the example of the intention expression dictionary used by the intention determination means 11 of the text sentence of 1st Embodiment. It is a figure which shows the example of the topic expression dictionary used with the topic determination means 12 of the text sentence of 1st Embodiment. It is a figure which shows the pattern example of the intention and topic of the high risk statement used with the high risk statement determination means 13 of 1st Embodiment. It is a figure which shows the example of the format of the risk information storage part of 1st Embodiment. It is a flowchart of operation | movement of 1st Embodiment. It is a block diagram showing the structure of 2nd Embodiment. It is a figure showing the example of data storage in the project relevant-information storage part 21 in 2nd Embodiment. It is a figure showing the example of data storage in the project relevant-information storage part 21 in 2nd Embodiment. It is a figure showing the example of data storage in the project relevant-information storage part 21 in 2nd Embodiment. It is a figure for demonstrating the detail of the project risk paragraph rule in 2nd Embodiment. It is a figure for demonstrating the detail of the project risk paragraph rule in 2nd Embodiment. It is a figure showing the example of data storage in the project risk paragraph memory | storage part 202 in 2nd Embodiment. It is a flowchart showing operation | movement of 2nd Embodiment. It is a block diagram showing the structure of 3rd Embodiment. The example of data of the project risk paragraph rule storage part 201 in 3rd Embodiment is shown. The example of data storage of the project risk paragraph memory | storage part 202 in 3rd Embodiment is shown. It is an example of the analysis condition setting screen of the project risk information display means 405 in 3rd Embodiment. It is an example of the project risk information display screen of the project risk information display means 405 in 3rd Embodiment. It is a flowchart showing operation | movement of 3rd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Data processing apparatus 2 Storage apparatus 11 Statement intention determination means 12 Statement sentence topic determination means 13 High risk statement determination means 21 Project related information storage section 22 Risk information storage section 23 Intention expression dictionary storage section 24 Topic expression dictionary storage section 25 High Risk Statement Specific Rule Storage Unit

Claims (44)

An intention determination means for determining an intention of a statement included in a text sentence related to a project based on a relationship between a predetermined text expression and the intention;
Topic determination means for determining a topic of a speech included in the text sentence based on a relationship between a predetermined text expression and a topic;
An extraction condition storage means for storing a risk sentence extraction condition composed of a plurality of intentions of speech and topic appearance patterns;
The intention determined by the intention determination unit and the topic determined by the topic determination unit are compared with the risk sentence extraction condition stored in the extraction condition storage unit, and the intention and the risk sentence extraction condition are matched. A risk detection system comprising: speech specifying means for specifying a speech corresponding to a topic as a risk speech.   The risk detection according to claim 1, wherein the intention is one of proposal, impossible, warning, request or request, prediction, question, obligation, denial, dissatisfaction, apology, report, or a combination thereof. system.   The risk detection system according to claim 1, wherein the topic is any one of bugs, schedules, work ranges, human resources, costs, health, materials, or a combination thereof.   The risk detection system according to any one of claims 1 to 3, wherein the risk sentence extraction condition is that the intention of the statement is other than a report and that there are a plurality of topics of the statement. .   5. The risk sentence extraction condition according to claim 1, wherein the intention of the utterance is warning, impossible, or negative, and the topic is different from the utterance immediately after the utterance. The risk detection system according to any one of the above.   In the risk sentence extraction condition, the intention of the statement is doubtful, the intention of the statement immediately after the statement is warning, impossible or negative, and the topic of the statement and the statement immediately after the statement is common The risk detection system according to any one of claims 1 to 5, wherein   The risk detection system according to claim 1, further comprising: a sorting unit that sorts the specified utterances for each risk extraction condition. Prioritizing the risk sentence extraction conditions,
The risk detection system according to claim 1, further comprising: a sorting unit that sorts the specified utterances according to the priority of the risk extraction condition. It has a paragraph cutout means that divides a group of text sentences related to the project into paragraph units,
Instead of the extraction condition storage means, the intention determination means, the topic determination means, and the statement specifying means,
A risk paragraph extraction condition storage means for storing a risk paragraph extraction condition consisting of a plurality of intentions of each paragraph, a topic, or an appearance pattern of a speaker;
A determination means for determining the intention, topic or speaker of the speech included in the text sentence divided into paragraphs;
The intention, topic or speaker determined by the determination means is compared with the risk paragraph extraction condition stored in the risk paragraph extraction condition storage means, and the paragraph that matches the risk paragraph extraction condition is identified as the risk paragraph The risk detection system according to any one of claims 1 to 3, further comprising: a risk paragraph specifying unit. The risk paragraph Extraction conditions,
(1) Conflicts between members and discrepancies have occurred (2) No concrete response method has been found (3) Clear decision making has not been made (4) Difficult problems have occurred (5) The work to be performed is not being performed, the hands are not working (6) The member does not have sufficient knowledge and information about the work contents, and the intention of multiple statements included in the paragraph The risk detection system according to claim 9, wherein a correspondence relationship with the appearance rule of a topic and a speaker is described.   The risk detection system according to claim 9 or 10, further comprising risk paragraph storage means for storing the risk paragraph specified by the risk paragraph specification means. Project-related word storage means for storing words closely related to a predefined project as project-related words;
Project-related word extraction means for extracting project-related words included in the risk paragraph;
The project information display means for classifying the risk paragraph as a classification axis with a combination of a project-related word included in the paragraph and a topic of a text sentence of the risk paragraph, and displaying it to a user. The risk detection system according to any one of claims 9 to 11.   13. The risk detection system according to claim 12, wherein the project information display means displays a risk paragraph corresponding to the classification axis when the displayed classification axis is selected.   The project information display means classifies a risk paragraph extracted from a project to be analyzed and a risk paragraph extracted from a project other than the analysis target into combinations of project-related words and topics of text sentences of those risk paragraphs, respectively. The risk detection system according to claim 12, wherein the risk detection system is classified as an axis, and risk paragraphs are displayed in parallel for each project.   The project information display means includes a risk paragraph extracted from a project other than the analysis target and a risk paragraph extracted from the analysis target project, a project-related word of each risk paragraph, and a text sentence of each risk paragraph. A combination with a topic is classified as a classification axis, and a classification axis that exists in a risk paragraph of a project to be analyzed, a risk paragraph of a project other than the analysis target but does not exist in a project risk paragraph of the analysis target, and the classification axis The risk detection system according to claim 12, wherein the project risk paragraph is displayed in parallel. The risk paragraph extraction condition is
(1) Among the intentions of statements included in a paragraph, any of “Negation”, “Impossible”, “Warning”, and a paragraph in which those intentions appear M or more times,
(2) Among the intentions of statements included in a paragraph, a paragraph in which the intention of “question” appears N times or more in succession;
(3) A paragraph in which the intention of the last statement in the paragraph is one of denial, impossible, warning, duty, or question,
(4) A paragraph in which the same speaker appears more than L consecutive times among the speakers in the paragraph,
(5) In the paragraph, the topic is at least one of the paragraphs including K or more topics, and the intention includes any statement of denial, impossible, warning, duty, request, request, question. The risk detection system according to claim 9. A risk detection method,
Included in the input text sentence by the data processing device based on information indicating the relationship between the predetermined text expression and the intention and information storing the relationship between the predetermined text expression and the topic Determine the intention and topics
The data processing device collates the determined intention and topic with a risk sentence extraction condition composed of a plurality of intentions and topic appearance patterns, and a statement corresponding to the intention and topic that matches the risk sentence extraction condition A risk detection method characterized by specifying as a risk statement. The risk detection according to claim 17 , wherein the intention is any one of proposal, impossible, warning, request or request, prediction, question, obligation, denial, dissatisfaction, apology, report, or a combination thereof. Method. The risk detection method according to claim 17 or 18 , wherein the topic is any one of bugs, schedules, work ranges, human resources, costs, health, materials, or a combination thereof. The risk detection method according to any one of claims 17 to 19 , wherein the risk sentence extraction condition is that the intention of the statement is other than a report and that there are a plurality of topics of the statement. . The risk sentence extraction conditions, a warning intended to speak, not a negative, and topics of claim 20 claim 17, characterized in that it is different in the speech immediately following the talk The risk detection method according to any one of the above. The risk sentence extraction conditions is a question the intention of the speech, the intention is warning of the text sentence in the remarks immediately after the speech, it is impossible or negative, and, the topic of the remarks immediately after the speech and remarks The risk detection method according to any one of claims 17 to 21 , wherein the risk detection method is common. The data processing device divides a text sentence group into paragraph units,
The data processing device determines the intention of the speech, topics or speaker included in the text sentence divided into paragraphs,
The data processing apparatus collates a risk paragraph extraction condition comprising a plurality of intentions of speech, topics or appearance patterns of speakers for each paragraph with the determined intention, topic or speaker, and extracts the risk paragraphs The risk detection method according to any one of claims 17 to 22 , wherein a paragraph matching the condition is specified as a risk paragraph. The risk paragraph Extraction conditions,
(1) Conflicts between members and discrepancies have occurred (2) No concrete response method has been found (3) Clear decision making has not been made (4) Difficult problems have occurred (5) The work to be performed is not being performed, the hands are not working (6) The member does not have sufficient knowledge and information about the work contents, and the intention of multiple statements included in the paragraph The risk detection method according to claim 23 , wherein a correspondence relationship with the appearance rule of a topic and a speaker is described. The risk detection method according to claim 23 or 24 , wherein the identified risk paragraph is stored. The data processing device extracts project related words closely related to a predefined project from the risk paragraph;
Wherein the data processing apparatus, the risk paragraph, classifies the combination of the text sentence topic projects related word and the risk paragraph included in the paragraph as a classification axis, claim 23, characterized in that display to the user The risk detection method according to claim 25 . 27. The risk detection method according to claim 26 , wherein the data processing device displays a risk paragraph corresponding to the classification axis in accordance with selection of the displayed classification axis. The data processing apparatus classifies a risk paragraph extracted from a project to be analyzed and a risk paragraph extracted from a project other than the analysis target by combining a combination of a project-related word and a topic of a text sentence of those risk paragraphs. 27. The risk detection method according to claim 26 , wherein the risk paragraphs are displayed in parallel for each project. The data processing device includes a risk paragraph extracted from a project other than the analysis target and a risk paragraph extracted from the analysis target project, a project-related word of each risk paragraph, and a topic of a text sentence of each risk paragraph. Are classified as classification axes, and the classification axis that exists in the risk paragraph of the project to be analyzed, the risk paragraph of the project other than the analysis target but does not exist in the project risk paragraph of the analysis target, 27. The risk detection method according to claim 26 , wherein the project risk paragraph is displayed in parallel. The risk paragraph extraction condition is
(1) Among the intentions of statements included in a paragraph, any of “Negation”, “Impossible”, “Warning”, and a paragraph in which those intentions appear M or more times,
(2) Among the intentions of statements included in a paragraph, a paragraph in which the intention of “question” appears N times or more in succession;
(3) A paragraph in which the intention of the last statement in the paragraph is one of denial, impossible, warning, duty, or question,
(4) A paragraph in which the same speaker appears more than L consecutive times among the speakers in the paragraph,
(5) In the paragraph, the topic is at least one of the paragraphs including K or more topics, and the intention includes any statement of denial, impossible, warning, duty, request, request, question. risk detection method according to any one of claims 29 to claim 23. A program for causing an information processing device to detect a risk statement included in a text sentence,
A process for determining an intention of a statement included in a text sentence related to a project based on a relationship between a predetermined text expression and an intention;
A process of determining a topic of a speech included in the text sentence based on a relationship between a predetermined text expression and a topic;
A risk statement extraction condition consisting of multiple statement intentions and topic appearance patterns is compared with the determined intention and topic, and a statement corresponding to the intention and topic that matches the risk statement extraction condition is identified as a risk statement. A program that causes an information processing apparatus to execute processing to be performed. The program according to claim 31 , wherein the intention is any one of proposal, impossible, warning, request or request, prediction, question, obligation, denial, dissatisfaction, apology, report, or a combination thereof. The program according to claim 31 or 32 , wherein the topic is any one of bugs, schedules, work ranges, human resources, costs, health, materials, or a combination thereof. The program according to any one of claims 31 to 33 , wherein the risk sentence extraction condition is that the intention of a statement is other than a report and a plurality of topics of the statement exist. The risk sentence extraction conditions, a warning intended to speak, not a negative, and from the claims 31, wherein the topic is that is different between the speech immediately following the speech of claim 34 The program according to any one. In the risk sentence extraction condition, the intention of the statement is doubtful, the intention of the statement immediately after the statement is warning, impossible or negative, and the topic of the statement and the statement immediately after the statement is common The program according to any one of claims 31 to 35 , wherein: A process of dividing a text sentence group into paragraphs;
A process of determining the intention, topics, or speaker of a statement included in a text sentence divided into paragraphs;
A risk paragraph extraction condition composed of a plurality of intentions of speech, a topic or a speaker appearance pattern for each paragraph is compared with the determined intention, topic or speaker, and a paragraph that matches the risk paragraph extraction condition is determined. 37. The program according to any one of claims 31 to 36 , further comprising: a process specified as a risk paragraph. The risk paragraph Extraction conditions,
(1) Conflicts between members and discrepancies have occurred (2) No concrete response method has been found (3) Clear decision making has not been made (4) Difficult problems have occurred (5) The work to be performed is not being performed, the hands are not working (6) The member does not have sufficient knowledge and information about the work contents, and the intention of multiple statements included in the paragraph 38. The program according to claim 37 , wherein a correspondence relationship with the appearance rules of topics and speakers is described. The risk detection method according to claim 37 or claim 38 , further comprising a process of storing the identified risk paragraph. A process for extracting project-related terms that are closely related to a predefined project from the risk paragraph;
38. The processing according to claim 37 , further comprising: classifying the risk paragraph as a classification axis with a combination of a project-related word included in the paragraph and a topic of the text sentence of the risk paragraph as a classification axis. Item 40. The program according to any one of Items 39 . 41. The program according to claim 40 , further comprising a process of displaying a risk paragraph corresponding to the classification axis when the displayed classification axis is selected. The risk paragraphs extracted from the project to be analyzed and the risk paragraphs extracted from the non-analysis project are classified as combinations of project-related terms and topics in the text sentences of those risk paragraphs as classification axes. 41. The program according to claim 40 , further comprising a process of displaying risk paragraphs in parallel. A classification axis for risk paragraphs extracted from projects other than the analysis target and risk paragraphs extracted from the analysis target projects, and combinations of project-related words in each risk paragraph and topic in the text sentence of each risk paragraph The risk axis of the project to be analyzed, the classification axis that exists in the risk paragraph of the non-analysis project but does not exist in the analysis project risk paragraph, and the project risk paragraph of the classification axis are parallel 41. The program according to claim 40 , further comprising: a process for displaying on the screen. The risk paragraph extraction condition is
(1) Among the intentions of statements included in a paragraph, any of “Negation”, “Impossible”, “Warning”, and a paragraph in which those intentions appear M or more times,
(2) Among the intentions of statements included in a paragraph, a paragraph in which the intention of “question” appears N times or more in succession;
(3) A paragraph in which the intention of the last statement in the paragraph is one of denial, impossible, warning, duty, or question,
(4) A paragraph in which the same speaker appears more than L consecutive times among the speakers in the paragraph,
(5) In the paragraph, the topic is at least one of the paragraphs including K or more topics, and the intention includes any statement of denial, impossible, warning, duty, request, request, question. Item 44. The program according to any one of item 37 to item 43 .

Images