A system for determining the risk of communication failures.

Abstract

This system provides a mechanism to determine the potential for communication gaps that arise when the assumptions or magnitude of impact in an expression of intent become unintended, without the user realizing it. [Solution] A risk assessment system has the function of detecting whether a string contains words that have a risk of causing communication problems, and displaying and warning about the resulting communication risks. When two or more words that have a risk of causing communication problems are detected, the system uses a database in which word combinations are registered to display the communication risks caused by the combination of those two or more words, and the type of communication risk is classified into categories including six categories: factual agreement, typographical error, misidentification, coincidence, agreement due to circumstances, and mismatch due to circumstances.

Description

Technical Field

[0001] The present invention relates to a risk determination system and a risk determination method for communication languages in language communication performed by two or more people.

Background Art

[0002] In communication between two or more people, the intention may not always be conveyed as intended. The side making a statement of intention (for example, information transmission in a chat) may inadvertently issue a statement of intention that indicates a different intention from the intention they want to convey to the side receiving the statement of intention. There are various factors that can cause a misunderstanding of the received statement of intention. These factors occur because the recipient of the statement of intention does not notice, for example, differences in underlying facts, differences in language or terminology, etc. As general processing by artificial intelligence develops, research is also being conducted on the computer's ability to grasp and process emotional expressions. However, a communication gap may occur due to the recipient's failure to recognize the facts underlying the emotions that are the basis for emotional expressions. Computer processing may not be able to fully process ambiguous languages such as emotional expressions. This may also be because the facts underlying the emotions that are the basis for emotional expressions have not been fully incorporated into the computer system, taking into account their importance weighting, etc. Of course, it is also possible that more abstract concepts, or thinking itself, contains a lot of physical and experiential elements, and the underlying facts to be incorporated depending on the situation may spread infinitely. First, logical documents (technical papers, patent specifications, news, etc.) that do not contain many emotional expressions are more suitable for computer processing, and since these applications have high business value, it is also conceivable to limit the applications to these, or to perform processing considering only elements with a certain level of weighting or higher. Furthermore, one approach to solving the above problems is to detect and warn about this communication gap by designing computer functions in a way that is easy for humans to understand, improving the human interface to assist and utilize human comprehension. This approach is also considered effective in addressing the problem of AI and computer processing becoming black boxes. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2019-045953

[0004] Patent Document 1 describes an invention relating to a thesaurus processing device and a program, and discloses how to process a program by understanding the meaning of words and phrases. However, there is no disclosure regarding how administrators can detect words that pose a risk of communication breakdown by combining elements that reflect the reality of human communication, including response times and word intervals, as well as other facts that form the basis for generating emotions. [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] In interpersonal communication, communication gaps can arise due to a failure to recognize underlying assumptions. This is particularly common in communication between different languages ​​and cultures. However, such communication gaps are not limited to communication between different languages ​​and cultures; they can frequently occur in situations where differences in underlying facts, language, or terminology are easily overlooked, such as communication between experts and non-experts, between people of different ages, between men and women, or between doctors and patients. Furthermore, in a more limited context, communication through machine translation presented a challenge: because machine translation software cannot grasp the meaning of the target language, it can produce translations that are completely nonsensical. [Means for solving the problem]

[0006] The communication language risk assessment system of the present invention checks for typographical errors, omissions, fragmented sentences, dangerous words, unclear or wide ranges in quantities, unknown units, and easily confused units in the communication language, and displays the potential for communication gaps resulting therefrom by methods such as displaying a list of risk factors, and issues a warning. More specifically, the present invention may have, but is not limited to, the following configurations (1) to (5). (1) The system detects if a string contains one or more words that pose a risk of communication failure, including typographical errors, omissions, fragmented sentences, dangerous words, unclear quantities or ranges in numerical values, unknown units, or easily confused units. A risk assessment system having a function to display and warn about the resulting communication risks, When two or more words that pose a risk of causing communication problems are detected, the system uses a database containing registered word combinations to perform the following actions: This displays the communication risks arising from combinations of two or more of those words. A risk assessment system that classifies communication risks into six categories: factual agreement, typographical error, misrepresentation, coincidence, circumstantial agreement, and circumstantial discrepancy. (2) The risk assessment system described in (1) above, which has a function to detect, display, and warn about communication risks by combining the time element between receiving a message from an information sender and sending a message from the information receiver after the information receiver has received the message. (3) The risk assessment system according to (1) or (2) above, which can determine the risk by adding the number of characters and / or words between words in at least one set of two detected words to the criteria. (4) Furthermore, if they are different, whether or not they are sentences in messages sent by the same information sender. The risk assessment system described in (2) or (3) above, which can determine whether a word used earlier is a sentence in a message sent by an information recipient other than the information sender who sent it, after the recipient has received it. (5) When determining whether a word is one that poses a risk of causing poor communication, Words registered in the database, The types of communication risks related to words included within 100 characters before and after that phrase, A risk assessment system as described in (1) or (2) above, which uses a database in which words detected for assessment are registered based on information about the type of communication risk of the words to be assessed. (6) The risk assessment system described in (1) or (2) above, which has a dictionary expansion function that expands one word into related words and has a function to detect the expanded words. [Effects of the Invention]

[0007] In this invention, linguistic communication is streamlined by: 1) confirming and outputting the language of the person making the declaration of intent as is; 2) outputting a summary translation or interpretation that reflects the parts that are not understood, even if the general meaning is clear; and 3) displaying a message indicating that the meaning is not understood and assisting in human translation (arranging a translator, requesting necessary additional information from the person making the declaration of intent including language, and returning schedule management and budget estimates). This makes machine translation more efficient and allows for more effective use of machine translation, which is becoming possible in multiple languages. Furthermore, the above classifications 1) to 3) can be applied to all forms of linguistic communication, and risk information can be classified and assigned accordingly. Furthermore, by performing language sorting in the above-mentioned language communication and assigning communication risk information, the system of the present invention can achieve at least one of the following: prevention of misunderstandings and troubles, danger detection, detection and provision of opportunities for emergency calls, discovery and provision of opportunities for fraudulent reports, improved operational efficiency, and improved efficiency of recording and post-incident processing. [Brief explanation of the drawing]

[0008] [Figure 1] A diagram illustrating verbal communication between people. [Figure 2] A diagram illustrating that communication language and the resulting text are formed from a collection of ideas demarcated by boundaries that can change depending on the context. [Modes for carrying out the invention]

[0009] Embodiments of the present invention are illustrated below. The present invention is not limited to the following unless its essence is changed. The communication language risk assessment system of the present invention determines the possibility of communication gaps occurring in communication language based on elements such as typographical errors, fragmented sentences, dangerous words, unclear quantities, numerical ranges, unknown units, and easily confused units. Furthermore, the present invention provides a risk assessment method for that purpose.

[0010] In person-to-person communication, communication gaps can arise when the underlying conditions are not shared between the individuals expressing their intentions. (Figure 1) As shown in Figure 2, communicative language and the texts it produces are formed from a collection of ideas demarcated by boundaries that can change depending on the context. In translation, this collection of ideas is expressed in another language. Each concept that forms a collection of concepts may be defined by being described in a language such as Japanese, English, etc. Even if there is no term in the translated language to express the concept, and even if such a concept itself is not culturally possessed, it suffices to define it using another language in such a case. This can be done, for example, by defining a term that is a dictionary, or by using the method when defining a patent invention in claims. A database of replacing terms with synonyms is also useful for this.

[0011] In the present invention, there is a risk associated with collecting communication risk information itself. As a method for avoiding it, for example, performing necessary distributed management by mediating through a human (physical blocking, forgetting of the database), or using AI (management that is black-boxed to a certain extent as variable data, etc.) can be mentioned. Also, there is a risk that needs to be avoided by entrusting interpretation to a machine. As a method for avoiding it, for example, having a system that leaves it to an advisor and always requires human confirmation for the final decision, performing sufficient confirmation for use in legally valid decision-making, and limiting it to use by experts can be mentioned.

[0012] [Terms and Definitions in the Specification] In the present invention, the communication language is at least a language used in daily conversations, papers, reports, etc. in a certain country or region, such as Japanese, English, French, Chinese, Russian, Arabic, German, etc., and includes languages within the range called natural languages. A fragmentary text is a text that can be interpreted or understood even if it is not described in normal grammar, such as having omissions or including coined words. Examples of fragmentary texts include poetic texts such as prose poems (aphorisms), prose texts, slogan phrases, and coined words with a certain regularity and interpretability. For example, abbreviations like "drive recorder" to "dora reko", coined words from "job hunting" like "job hunting preparation end", etc. can be assumed. In the system of the present invention, the intention expression of the information sender is always made by a method including characters. The intention expression may include information transmission other than characters, but it is desirable that it does not change significantly, such as canceling the information transmitted by characters. The chat function is mainly a function for exchanging relatively short messages between two or more than three persons. In the chat function, in the communication between two persons, one of the two persons becomes the information sender, and the information sender transmits a message to the information recipient, who is the other one of the two persons, thereby performing a communication function by characters. Here, either of the two persons performing the communication can become both the information sender and the information recipient. Also, in a specific application, in the chat function of two or more persons, there may be a person who only receives information or a person who only sends information. The mail function is mainly a function for exchanging messages longer than the chat function between two persons or a large number of persons. Since the message is long, it is useful to detect words that may cause a communication risk during one information transmission and reduce the risk.

[0013] In the chat function and the mail function, in a system where the terminal used by the information sender has a function of receiving information and the information sender can also become the information recipient, when two or more words corresponding to words having a risk of communication failure are detected, the number of characters and / or the number of words between the words in at least one set of two words among those words, and the relationship between the sentences in which the words in the two-word set exist are also considered. The relationship between the sentences in this case means whether two words are in the same sentence, whether the two words are in sentences in a message transmitted by the same information sender, how many exchanges have occurred, the number of characters and words between the two words, and the time from the transmission of one word until the other word is input and transmitted may also be considered. By considering the number of characters, the strength of the relationship between two words can be estimated. Furthermore, by considering the time between the transmission of one word and the input and transmission of the other word, it is possible to estimate implied meaning, that is, the existence of an interpretation that should be assumed from the context and surrounding situation. For example, if someone is asked something that should be obvious and replies without any pause of about a second, "Why are you asking that?", it is reasonable to interpret it literally. However, if they think for 10 seconds before replying, "Why are you asking that?", it may be necessary to consider, as needed, the possibility that the person being asked may be misunderstanding something basic, or that they have lost motivation, are tired, etc. In email functionality, the risk assessment system of the present invention may be used to determine the risk of communication failures in conjunction with pre-sending spell checks, in order to send messages containing many sentences in a single transmission. This system detects potentially risky words, issues warnings, and prompts users to add or revise the message.

[0014] [Classification of information transmission states] When the intended message conveyed by the information sender matches the facts, and the written representation of the information also matches the facts, this is called a "fact-matching" state. When the intended message of the information sender matches the facts, but the written representation of the information differs from the facts, this is called a "typographical error." When the intended message of an information provider differs from the facts, and the written representation of that information also differs from the facts, this is called a "misrepresentation." In theory, it is conceivable that, even if the intended message of the information sender is originally different from the facts, the written representation of the transmitted information may, for some reason or by chance, match the facts. In this specification, a match due to some reason will be referred to as a "match due to circumstances," and a match by chance will be referred to as a "match by chance." If the intended message of the information provider, which was originally different from the facts, ends up matching the facts for some reason, it could be because, for example, a typographical error was corrected by a software correction function, or because someone else who composes or checks the written text inserted an unconventional process to correct the text, resulting in it matching the facts. A coincidence could occur, for example, in a communication software where an information sender and an information receiver view each other's images, and while viewing the images, the sender sends a text message via voice or chat saying "raising the right hand," but the intended message is "raising the left hand." In this case, mirroring processing is applied to the communication image, making it appear to the information receiver that "the right hand is raised." Regardless of whether either the sender or receiver, or both, are aware of the mirroring processing, the fact that the image shows "raising the right hand" may justify to the information receiver that it is consistent with the facts. Furthermore, whether the information sender is in a state of typographical error, misidentification, or accidental coincidence may also depend on whether the information sender sent the text message "Right hand raised" taking into account the effects of mirroring processing so that the image of the information recipient shows an image of "raising the right hand." In this case, if, when an information sender sends a text message, a certain time lag in communication is tolerated, and a checker inserts an unconventional process to correct the textual representation, and as a result, it becomes reasonable to interpret that the text is consistent with the facts, then for the information sender, this may constitute a coincidence due to circumstances.

[0015] In this case, even if the information sender sent a text message that was factually accurate, if the person checking it made a mistake, resulting in a situation that should not be interpreted as factually accurate, then a situation that should be classified as "inconsistency due to circumstances" may arise. Based on the above, the situations can be classified into six categories: factual agreement, typographical error, misidentification, coincidence, agreement due to circumstances, and discrepancy due to circumstances. Furthermore, in cases where discrepancies occur accidentally, these can be classified as either typographical errors, misinterpretations, or discrepancies due to circumstances, depending on the nature of the accidental occurrence. Therefore, the system of the present invention can be operated without independently classifying situations involving accidental discrepancies. The six categories above are defined separately for the information sender and the information receiver. If an information sender sends the opposite of one of two options with the intention of deceiving the recipient, and the information sender themselves mistakenly believes the opposite of the two options to be true, then the message, including the wording, is a misunderstanding, not a typographical error, for the information sender, and a misunderstanding, but a coincidence, for the information receiver.

[0016] [Dictionary Expansion Function] In the dictionary expansion function, words and short strings of characters that do not have a specific meaning, which have been referred to as "words" in this specification, are treated as "words" that can be registered in the dictionary table. Words registered in system tables for use in the system, such as dangerous words, warning words, and caution words, are expanded from the following perspectives. The expanded words are registered in the same or different tables and used for searching, extracting, etc., various words. If a word derived from a cautionary word becomes a dangerous word or a warning word, the surrounding words may be reclassified as dangerous words or warning words mechanically using certain criteria and processing routines, or with the intervention of human judgment. A string of characters that is in the form of a sentence is broken down into words (in the case of Japanese, words according to Japanese grammar) through morphological analysis. Furthermore, multivariate analysis can be used to analyze the probability of another dangerous word appearing and the trend of the probability of determining a communication risk, based on the character spacing, distance and relationship between a word with one specific attribute, such as a warning word, and another word with a specific attribute, such as another warning word, as well as the time interval between appearances. Artificial intelligence (AI) can then be used to speed up the processing of determining the probability of another dangerous word appearing and the presence of a communication risk, as well as improve accuracy. In this case, the predictions made by multivariate analysis may be recorded and stored in a database separately from the operation of the system of the present invention, and the administrator may perform maintenance and checks as needed. By measuring the interval between words using the number of characters, or alternatively, the number of characters plus the number of words, and using this as one of the parameters in the multivariate analysis described above, it becomes possible to more accurately determine the attributes of words such as dangerous words, warning words, and caution words that appear in a text.

[0017] In the dictionary expansion function, a word with a specific attribute, such as a "dangerous word," is treated together with the expanded word derived from it, allowing for a more realistic risk assessment. The expanded word may include, for example, a word expanded from the following five elements: Element 1: Sound This refers to a relationship between words that, due to their similar sounds, can lead to mishearing or misinterpretation. It can be described as a relationship between words that can cause mishearing or confusion. For example, similar-sounding words (trademarks) are described in the reference "Changes in Phonological Sense in Judging Similarity of Pronunciation Evidence" - As Seen in Decisions Concerning Alphabetical Trademarks - (Suga, Fusao, Patent 2014 Vol 67, No. 6). By expanding and registering names that are prone to confusion in the database related to the present invention, processing speed can be improved. Element 2: Appearance - Bear and camphor tree, or bread and bun, etc. Element 3: Concept - Cancer, and the negative consequences of it worsening and growing if left untreated, etc. Element 4: Association (including euphemism) Examples include expressing continuous losses as "bleeding that won't stop" or using "people who don't come back" to refer to those who have died or retired. Element 5: Situation Meanings arising from the situation. For example, when words are regarded as difficult to understand, such as incantations, sutras, Greek words, etc. The following will illustrate expansion examples using words listed in national language dictionaries and specialized term dictionaries as examples. Words listed in dictionaries used in compulsory education such as junior high school dictionaries may be defined and treated as everyday language. However, since many of these dictionaries also list words that are rarely used in daily life, they may be used appropriately according to the vocabulary, national language ability, and knowledge level of the information sender and receiver who need to judge risks in communication. Moreover, for words that are known, depending on factors such as age, occupation, gender, place of origin, and specialization, and further depending on the situation, there may be words that cannot be immediately grasped because they do not fit the situation even when suddenly received. Words not listed in those dictionaries but listed in comprehensive national language dictionaries such as the "Kōjien" may be treated as advanced or difficult words. Specialized terms and words within a group exist depending on each industry and the attributes of the users. There are words specified to give an exact definition, as well as slang and buzzwords used within a group of a specific attribute, which can be factors contributing to communication risks. Therefore, even these words that are not generally used may be beneficial to register in the database depending on the situation.

[0018] (Example 1) Cancer Everyday language Sound Homophones, words with similar sounds. Example: cancer, wild goose, wish Appearance When it can be read as "kan" from the appearance, such as can, official, etc. Concept Severe illness. Bird. Wish made within the company. Or the English "GUN", gun. Association (including euphemism) Growing bad things. Scary things. Situation As "fake", the opposite of "true" for counterfeit goods. (Example 2) This is Homophones, similar-sounding words. Examples: leaves, a woman's name. Exterior (nothing in particular) Concept: Something light. Association (including euphemisms): Something light and worthless. Something that disappears quickly. Situation: A symbolic element remaining as a last hope in the final scene of the novel. (Example 3) chopsticks Homophones, similar-sounding words. Examples: bridge, edge, bridge, fuzzy, hajii (embarrassing). Exterior (nothing in particular) Concepts; something that bridges the gap; something that is not central. Association (including euphemisms): trivial things. (Like the edge of a bento box) The situation is one of being cornered and broken; abandoned; or pushed to the sidelines.

[0019] (Example 4) Glass transition temperature Technical terms Homophones, similar-sounding words. Example: glass point ion degree Appearance: Inspired by the transition temperature of crows Concept: The melting point of glass (because it is the point at which glass becomes soft). The temperature at which glass forms (incorrect according to the technical definition), the temperature at which transparency is achieved (incorrect according to the technical definition). Situation (when simply understood as difficult words): spells, sutras, Greek, Latin. (Example 5) Phase transition Technical terms Homophones, words with similar sounds. Examples: shifting the point of contention, focus I, laughing point. Appearance: Many phase changes Concepts: vaporization, liquefaction, solidification, major changes. Association example: Worsening of a disease, such as cancer metastasis. Situation (when simply understood as difficult words): Spells, obscure theoretical physics terms.

[0020] (Example sentence 1) Do not cross this bridge. Sound: Do not cross this bridge, do not cross this edge, do not cross these chopsticks (meaningless) Appearance: An eerie text consisting of a string of hiragana characters. Conceptually, a slightly unconventional way of writing in a scene from a folktale, like the story of Ikkyu-san. Association: Konoha (female name) must not cross death (You must not die. You must not do dangerous things.) The situation is that a leaf must not cross death. In the scene from the novel "The Last Leaf," one might be able to draw a parallel between the feelings of a painter drawing a leaf to a patient who believes he will die when the last leaf falls, but the situation in which one can empathize with such a "false connection" is likely to be either an urgent situation like a family member's deathbed, or a state of slight mental exhaustion.

[0021] 1. When sending When an information sender sends text information via chat or email, the content of that message can be classified into the following six categories according to the above classification. These fall into six categories: factual agreement, typographical error, misidentification, coincidence, agreement due to circumstances, and discrepancy due to circumstances. The risk assessment system of the present invention may be a system in which the information sender first creates a message, and then the system determines and evaluates the risk before sending it. In such cases, depending on the purpose of the communication system, if the sender's message poses a high risk, the message may be canceled or converted into a message with less risky words before being sent.

[0022] 2. Upon receiving Even when information recipients send text information via chat or email, the content of those messages can be classified into the following six categories according to the above classification. These will be categorized into six categories: factual agreement, typographical error, misidentification, coincidence, agreement due to circumstances, and discrepancy due to circumstances. However, in the case of receiving information, the range of what can be classified as a misinterpretation is not only inherently wide in terms of the range of interpretation, but it becomes even wider when we include misunderstandings that are more likely to occur on the part of the recipient. However, by adding these words containing misidentification and misunderstanding to the dictionary expansion function mentioned above, and registering and accumulating them as expansion words in a database, it is believed that the function of reducing communication risks will be further improved. From the perspective of the information recipient, it may suffice to simply know that the information from the information sender is inaccurate, regardless of whether it is a typographical error, misinterpretation, or inconsistency due to circumstances; in such cases, there may be no need to distinguish between them. On the other hand, failing to distinguish between them may lead to a misunderstanding of the information sender's intentions and hinder communication. By using this system, the possible misinterpretations and misunderstandings are displayed, making it easier for the information recipient to understand and grasp both possibilities. Similarly, from the perspective of the information recipient, it may suffice if the information from the information sender is factually accurate, coincidentally accurate, or coincidentally accurate, as long as it is clear that it is factually accurate, and there may be no need to distinguish between them. On the other hand, failing to distinguish between them may lead to a lack of understanding of the information sender's intentions, resulting in communication problems. From the perspective of the information recipient, it may be more difficult to verify the facts than the information sender, perhaps due to being in a remote location, so a wider range of possibilities must be considered. In this case as well, the determination system of the present invention warns of the possibility of misidentification (which can also be called a misunderstanding) and the possibility of agreement or disagreement due to circumstances, so the system of the present invention can provide a method that helps the information recipient understand the situation.

[0023] 3. Intermediate data processing after transmission is complete but before reception. It is also possible to process intermediate data sent but not yet received as if it were to be received later. In this case, the recipient of the information may have already been identified, or it may not have been identified yet, or it may have been roughly identified but not yet confirmed. In this case as well, it is possible to design a data processing flow by classifying the information from the information sender into six categories: factual accuracy, typographical error, misidentification, coincidence, circumstantial accuracy, and circumstantial inconsistency. When determining which category to classify something into, if it could potentially be classified into multiple categories, the criteria for determining which category is safer may be used, or the criteria for determining which category is safer even if it results in poor communication may be used. For example, consider the following scenario: If an information sender sends a message stating they are feeling unwell, the recipient's system, even after considering other circumstances, should consider the possibility of a "coincidence" and implement a processing flow that requires reconfirmation, rather than classifying it as a "misunderstanding." This can prevent an increase in communication risks due to a deterioration of trust. Similarly, if an information sender sends a message stating that "there is a strange smell coming from the plant," it may be safer to treat it as an accident rather than a typographical error or misunderstanding. In such cases, it is acceptable for the information recipient to classify it as a "discrepancy due to circumstances."

[0024] Words of each attribute can be classified, for example, into forbidden words, dangerous words, caution words, danger avoidance words, etc. Depending on the application of the risk assessment system of the present invention, it is possible to either not set or set words that would cause message transmission to stop as "NG words." Examples of words that fit each attribute include the following: Examples of identifying forbidden words and dangerous words This is a forbidden word: kill Dangerous word: 100 watts of energy. Watts are a measure of energy efficiency, so it needs to be checked to see if it's a mistake for watt-hours. This is a dangerous word in the sense that it may be misunderstood. It is also possible to set a separate attribute for dangerous words that may cause emotional distress. While "kill" is a forbidden word for avoiding danger, "kill bacteria" or "sterilize" do not require a warning. However, if it could be interpreted as "sterilize a person," then it becomes a dangerous and forbidden word.

[0025] Examples of warnings output by the risk assessment system of the present invention include the following: Method of issuing warnings: Warnings are displayed in yellow for dangerous words and in red for forbidden words. However, in cases where a warning is not issued at the time but used later for reflection, it may be possible to activate the system in a way that makes it unclear what constitutes a dangerous or forbidden word, and only display and warn when necessary. Normally, these are not warning words, but there may be words that require attention in a different sense, such as words that are likely to cause accidents in specific situations. In such cases, you might ask questions during the message exchange such as, "What do you mean?", "Are you okay?", or "By the way, how many hours did you sleep last night?". Sending messages that could cause an accident, intentionally or unintentionally, or emotionally, while operating machinery also falls under this category.

[0026] The following are specific examples of embodiments of the present invention. Embodiments of the present invention are not limited to those described below. The following embodiment involves verbalizing communication in a place where person-to-person communication takes place and incorporating it into a system such as a computer. The system may be a mobile terminal with an application program installed or built-in. Specifically, the sender of information may manually judge and detect warnings using the risk assessment software of the present invention when inputting text into commonly available chat software or email software, and the receiver of information may do so when receiving text. Alternatively, it may be provided as an additional function, such as being built into the software or as an add-in. It is desirable to allow users to add and use the software at their discretion as a subscription-type application, as this would allow for easy application to a wider range of uses.

[0027] The above describes a system that determines risk for each word in a message. However, it is also possible to output a summary by deleting or replacing words in the text that the computer cannot process. In this case, the summary output process can be performed independently, but by adding it to the process of determining risk for each word in the message, as described above, more effective risk assessment and communication assistance can be achieved. Generally, risk assessment software includes: 1. If the risk assessment and warnings from the risk assessment system can be used as is, output the output results from the risk assessment system. "Person A says, '...(output of the risk assessment system)...'" 2. If the risk assessment and warning system can partially perform risk assessment, the following output will be produced: "Person A is probably saying something along the lines of, '...(output of the risk assessment system)...', but the part about 'the limitations of the risk assessment system that reflect the risk of misjudgment' is unclear. It's probably reasonable to understand it as 'the output reflecting the risk assessment and warnings made by the risk assessment system.'" Advice on reasonable business interpretations. 3. If the risk assessment and warnings from the risk assessment system are largely unusable, the response will be left to a human, stating, "It is unclear what Person A is saying."

[0028] Use in multilingual investment analyst reports Technical translation, investment analyst report translation, news translation For example, it can be applied to surveys conducted in non-English languages ​​in emerging economies, or to compliance management of local company documents. The classification or allocation function of the risk assessment system is as follows: The following can be developed to enable the system to learn and respond with greater accuracy by adding artificial intelligence learning capabilities, as well as other data and processing routines. 1. If the risk assessment and warnings from the risk assessment system can be used as is, output the output results from the risk assessment system. "Person A says, '...(output of the risk assessment system)...'" 2. If the risk assessment and warning system can partially perform risk assessment, the following output will be produced: "Person A is probably saying something along the lines of, '...(output of the risk assessment system)...', but the part about 'the limitations of the risk assessment system that reflect the risk of misjudgment' is unclear. It's probably reasonable to understand it as 'the output reflecting the risk assessment and warnings made by the risk assessment system.'" Advice on reasonable business interpretations. 3. If the risk assessment and warnings provided by the risk assessment system are largely unusable, the following output will be displayed. "It's unclear what Person A is saying." In this case, forward it to a human translator. In summary, the present invention allows the output results to be varied based on the risk assessment and warnings issued by the risk assessment system, the circumstances under which they are output, and factors that cause communication gaps in the messages in the communication language, such as typographical errors, omissions, fragmented sentences, dangerous words, unclear quantities, numerical ranges, unknown units, easily confused units, and so on. The changes occur in stages, such as returning machine translation, returning machine summaries, returning instructions that outline the risks, requesting additional information, or forwarding the translation to a human.

[0029] Using translation software equipped with the above risk assessment system, (1) Technical translation, patent translation, etc. for developed countries: We provide translation between languages ​​such as Japanese, German, French, Chinese, Korean, Portuguese, Russian, and Italian. (2) Also, BRICs and Asia India (English) - Multilingual (including Japanese) Brazil (Portuguese) - Multilingual, including Japanese Russia (Russian) - Multilingual including Japanese Chinese - Multilingual languages ​​including Japanese Thailand, Vietnam, Indonesia, Myanmar, etc. - Multilingual, including Japanese, (3) Emerging economies South America (Spanish) - Multilingual (including Japanese) Middle East (Arabic) - Multilingual (including Japanese) Africa (French, English) - Multilingual including Japanese Applicable to languages ​​that can be translated by machine translation software, including Eastern European languages ​​(Polish, Czech, Ukrainian, etc.) and Japanese. Stage 1: Machine translation only + basic business risk check Stage 2: After machine translation into Japanese, double-check with machine translation into English. Stage 3: The results from Stage 2 are checked in Japanese or English by a person who understands that language. Stage 4: Apply the results of Stage 2 or 3 to business risks (including legal aspects and impact assessment in case of misunderstanding). Stage 5: In the event of a problem, arrangements will be made for mediation and resolution by Japanese and foreign legal experts.

[0030] Uses in medicine This addresses the aspects of depression and autism that are caused by miscommunication in language. This serves as a warning against people without medical knowledge using terminology that could mislead the recipient.

[0031] Use in security This system is used to review and evaluate emails, external documents, and internal documents written in foreign languages. It can become practical by incorporating a feature that utilizes the classification function of this invention, which outputs summaries rather than complete translations to respect privacy.

[0032] Use in safety management for accident prevention This system issues an alert when communication language in workspaces such as factories and businesses contains hazardous words, such as terms for hazardous substances or dangerous values, that are not commonly known. Algorithms used by internet content advisors for risk assessment could also be applied. Even if certain words are not normally recognized as dangerous words, it is conceivable to create a database of the likelihood of accidents occurring when combinations of words are used synchronously, and use this database for accident prediction and prevention. [Examples]

[0033] The following are possible embodiments of the system of the present invention. The potential applications of the present invention are diverse and, of course, not limited to those listed below. Example 1: Use in a multilingual investment analyst report Example 2: Technical Translation Example 3: Advice on topics of conversation during meals Example 4: Topics in entertainment, particularly song selection in karaoke. Example 5: Resolving the communication gap between the elderly and younger people in a nursing care facility. Example 6: Resolving the communication gap in conversations between people from regions with a local dialect and tourists or business travelers who are unfamiliar with the dialect. Example 7: Preventing dangerous words during online drinking parties Example 8: Online support at government service counters Example 9: Online support for corporate customer service. For example, a complaint reception desk. Example 10: Hotel Reservation System

[0034] The risk assessment system of the present invention can be applied to many uses. For example, by taking "industry" as the horizontal axis and arranging each industry such as medical, nursing care, chemical, commercial, machinery, electrical, leisure, etc., and "problems to be solved and the resulting effects / benefits" as the vertical axis, and examining the intersections on a one-to-one basis for each, a wider range of possibilities can be considered. For example, each item on the vertical axis could include preventing misunderstandings and problems, detecting danger, detecting and providing opportunities for emergency reporting, identifying and providing opportunities for fraudulent reporting, improving operational efficiency, and recording and post-incident processing. [Industrial applicability]

[0035] This invention can provide a communication risk management system, for example, a system that improves the efficiency of translation in multilingual machine translation, and can be used in industries in the fields of communications and software.

Claims

1. The system detects if a string contains one or more words that pose a risk of communication failure, including typos, omissions, fragmented sentences, dangerous words, unclear quantities or ranges in numerical values, unknown units, or easily confused units. A risk assessment system having a function to display and warn about the resulting communication risks, When two or more words that pose a risk of causing communication problems are detected, the system uses a database containing registered word combinations to perform the following actions: This displays the communication risks arising from combinations of two or more of those words. A risk assessment system that classifies communication risks into six categories: factual agreement, typographical error, misrepresentation, coincidence, circumstantial agreement, and circumstantial discrepancy.

2. The risk determination system according to claim 1, which has a function to detect, display, and warn about communication risks by combining the time element between receiving a message from an information sender and sending a message from the information receiver after the information receiver has received the message.

3. The risk determination system according to claim 1 or 2, which can determine the risk by adding the number of characters and / or words between words in at least one set of two detected words to the determination criteria.

4. Furthermore, if they are different, whether or not they are sentences from messages sent by the same information sender. The risk determination system according to claim 2 or 3, which can determine whether a word used earlier is a sentence in a message sent by an information recipient other than the information sender that sent it, after the recipient has received it, by adding this to the criteria for determination.

5. When determining whether a word poses a risk of causing communication problems, Words registered in the database, The types of communication risks related to words included within 100 characters before and after that phrase, A risk determination system according to claim 1 or 2, which uses a database in which words detected for determination are registered based on information about the type of communication risk of the words to be determined.

6. A risk determination system according to claim 1 or 2, which has a function to expand a single word into related words using a dictionary expansion function and to detect the expanded words.

Images