"context-based" statement-related device

Abstract

[Problem] In assessing whether a statement is true or false, it is desirable to make an assessment for the entirety of a document rather than a portion thereof. It is also said that the truth or falsehood of a statement depends on context. There is a method referred to as "surrounding context". However, this method is not for entire documents. [Solution] Accordingly, a "global context" is proposed. Any given bundle of conjunctively joined statements is defined as a unit context. A global context is defined when a conjunction of unit contexts continues from the beginning to the end of a document. In the hypothetical case where a statement "p" and its negation "￢p" are latent within this global context, it can be said that the global context "p∧,, ∧￢p∧" is a contradiction. FIG. 6 is a schematic diagram of sentence context. In the diagram, (α)-(γ) are individual global contexts. Also, (a)-(e) are unit contexts. U1, U2, and so on are unit statements. A conjunction of unit statements is a unit context, and a conjunction of unit contexts is an individual global context. A connotation such as (b) also constitutes a conjunction. If "p" and "￢p" are in different unit contexts, there is no contradiction.

Description

"Context-based", sentence-related devices

[0001] The present invention relates to the context of texts such as documents and speech.

[0002] When examining the semantic content of documents, statements, etc., it is often said that "meanings vary depending on the context." In Patent Document 1, the vectors of multiple words scattered around an arbitrary word A are measured (referred to as "vector group A"). On the other hand, for a word B in another part of the text that has different characters from A, the surrounding vector group B is also measured. If vector group A and vector group B are similar, then phrases A and B are determined to have the same meaning despite their different characters, and are therefore considered synonyms. On the other hand, if a word C with the same characters is present in multiple parts of another text, vector groups C around phrase C are taken in multiple places (group C1, group C2, ..., group Cn). If group C is not similar, it is determined that the meanings of phrases C may be different even though the characters are the same. Such methods of determining ambiguous words are conceivable.

[0003] This method compares the "surrounding context" to determine the appropriateness of a word's usage, and is expected to be effective to a certain extent.

[0004] Patent Publication No. 2023-15565

[0005] However, with the surrounding context method, it is difficult to show whether a document requiring consistency is truly consistent throughout. This is because the scope of the surrounding context is generally limited to a few words before and after any given word. For example, if a book (say a traffic safety textbook) begins with "Cross the road if the light is green," and ends with "Cross the road if the light is red," we would be confused. We would think it is contradictory.

[0006] The main objective of the present invention is to determine whether the text is consistent throughout the entire document or utterance. The consistency of the text refers to whether the sentences and phrases that make up the text are consistent or not.

[0007] If we use the surrounding context comparison method mentioned above to address this issue, the surrounding words for both the green and red lights will be "cross" and have the same vector. Also, because both signals allow people to cross, we can imagine that similar words such as "wait for the light," "sidewalk," and "car" will be present in the surrounding area. Since the vectors of these surrounding words will be roughly the same, the green and red lights will likely be judged to be synonyms. However, the locations where the green and red sentences are written are far apart in the document, so one may wonder whether it is even worth comparing them. This is because comparing meanings is not valid when the contexts are different.

[0008] In response to our confusion, the traffic safety instructor said that the two stories, namely the stories about green and red traffic lights, are not contradictory. He said that the opening and closing stories can be understood by reading the textbook as one story. Indeed, the textbook does include a story about animals that cannot distinguish colors. It also talks about human color blindness and how traffic light colors are determined. The final story is a cautionary tale about realizing a safe society where people can cross the street even when the light is red.

[0009] The participants in the lesson expressed their dissatisfaction with the reversal of red and green, saying that if red means cross, does that mean green means stop? The instructor seemed keen to emphasize the need for rules, whether it's green or red. Apparently, all the participants in the lesson were former traffic violators who had been called in for re-education. It was also said that the textbook was partly handmade, and that this moralizing section was not included in good drivers.

[0010] What concerns me is that the audience for the lessons was a specific group, and that the textbook was a special, handmade one. If what the instructor was saying had been categorized like a document, with chapters, sections, and paragraphs, it might have been understood. This is because the areas were clearly separated. From the perspective of this invention, which I will explain later, different "areas" are explained as different "contexts." The textbook was actually a seven or eight-page booklet aimed at offenders. It appears to have been filled with the instructors' thoughts.

[0011] However, just because a document is divided into chapters, sections, and paragraphs does not mean that the items are independent of each other and that there is no need to examine their relationships. In fact, the opposite is true. Rather than simply examining a document in chapter, section, or paragraph units, it is necessary to examine the document as a whole, taking into account the relationships between these chapters, sections, and paragraphs, and to check whether the sentences are consistent and do not contradict each other. In today's world, with so many documents being published, the effort required to examine the meaning of documents is enormous.

[0012] "Context" is an effective means of examining the entire document, taking into account its chapters, sections, and paragraphs. "Tracing, tracing, and isolating the context" are well-known methods of document review. This invention realizes this well-known method by using "global context" that covers the entire document. It goes beyond the limitations of "peripheral context."

[0013] There are instances where the use of "context" is the only option. By changing the context, a meaningless sentence can be made understandable.

[0014] To understand how this context works, let us look at Example 1. Example 1 is a nonsensical sentence. It is taken from "Syntactic Structures" (p.15, Author: Noam Chomsky, Publisher: Mouton, 1973).

[0015] [Example 1] Colorless green ideas sleep furiously.

[0016] This sentence is difficult to understand. He says there is no color, yet the color is green. Furthermore, the green color is attached to ideas (plural), and the ideas are green. We are left wondering what it means for ideas to have color in the first place. It is a mysterious description.

[0017] However, if you think of this description as a symbol or metaphor, then that might be the case. However, even if you try to think of what it symbolizes or what it is analogizing, it is not easy to think of the object of the analogy. Perhaps the reason there are multiple thoughts is that there are many different thoughts. Perhaps that is the case. However, it is strange to say that thoughts are asleep. Thoughts do not sleep or wake up. However, if you think of this as an attempt to express something, then you half-nod and think that it might be so. However, the most impressive part is that it says that the sleep is raging. To be angry means to be awake, and therefore not asleep. Each of the words is incompatible in content with each other, especially with the word next to it.

[0018] However, looking back, we notice that in the above interpretations, we readers are trying to understand the meaning of meaningless sentences. Let's call this reader Interpreter A. Interpreter A tries to come up with a new interpretation by adding some kind of symbol or analogy, or by changing his or her perspective. In other words, when the original text is meaningless, that is, when the words are completely unrelated in meaning and are even contradictory, Interpreter A even imagines the original author's intention and wonders whether the original author intended each word to have such and such a meaning.

[0019] This shows that the act of creating context is strongly related to the act of human meaning comprehension. By supplementing the context appropriately, the original sentence, "Colorless green thoughts sleep furiously," appears to be richer in meaning. The context structure in this case has the same format as when explaining an object: "What is...?" That is, "colorless" is "this" and "this" (Context 1), "thoughts are green" is "this" (Context 2), "thoughts are furious" is "this" (Context 3), "thoughts are asleep" and "sleeping is furious" is "this" (Contexts 4 and 5). If we succeed in creating contexts 1 through 5, Example 1 will become meaningful. Thus, it seems possible to generalize that meaning can be supplemented by context.

[0020] A word about the sentence structure "What is...?" is used to define or explain any phrase or matter. A linguist who observed the language development of infants up to the age of two on a daily basis reported the following: The "relation" mentioned in this report seems to be the prototype of a definition.

[0021] "When a toddler says 'mamma,' they are using what is known as a one-word sentence, such as 'Mommy, give me something to eat,' or 'There's food over there.' When a child reaches about one and a half years old, one word alone becomes insufficient. In this way, two or more words are lined up, the relationship between them becomes clearer, and the word gradually develops into adult language." (Aizawa Yoshiko, Language Life, No. 223 (April 1970), Chikuma Shobo).

[0022] Furthermore, the inventor believes that once a child reaches the age of three, they begin to actively ask "why" and "how." In adult terms, they do not accept words that do not make sense. They begin to accept words that make sense. Their words, which are a means of expressing words, become richer.

[0023] The inventor believes that this structure of how young children receive words is the prototype of the "context" referred to in this invention. As children grow older, this knowledge acquisition structure that asks for reasons and causes transforms into a conditional sentence structure such as "in this case, this happens," which is the structural form referred to as "context" in this invention. Context is made up of sentences. Sentences have structural forms. The structural form of a sentence is responsible for the relationships between words and phrases and between sentences. In this invention, by systematizing this structural form of "context" and applying it to sentences, the meaning of the words and sentences in an entire document, which is a collection of sentences, can be consistently examined from the perspective of structural form, not in part but as a whole.

[0024] The main task above is achieving coherence in sentences. However, by focusing too much on coherence, do we become so tied to the accuracy of our knowledge that we forget the joy of inventing new ideas? This is a related issue that accompanies the main task above. This issue may also lead to the shame of making mistakes or the fear of saying something different from others. This attitude toward knowledge may stem from an educational approach that emphasizes the correctness of knowledge and not the process of acquiring it. However, can we not create a social culture in which adults also actively engage in the same questions of "why" and "how" that young children ask? We will not discuss this issue further. However, the "context" method of this invention supports the coherence of sentences and statements, even when new contexts are created. Therefore, with this support, we can ask diverse questions and come up with ideas without worrying about inconsistencies or errors. We can express a wide range of ideas with confidence. The inventors hope that the present invention will help people to broaden their ideas.

[0025] There are other challenges, too. Finding out what claims are made in a document is tedious, especially for large, long documents. It takes a lot of time. Searching for words does not determine whether a claim is true or false; that determination must be made on a sentence-by-sentence basis. This is a major problem in today's information-rich world.

[0026] The main problem is solved by inspecting sentences in the same context. However, in this case, questions arise as to what is a context and what is the same context. Furthermore, a definition is required for consistency, which is the original problem. Why consistency of meaning between sentences must be performed in the same context will be explained later in the section on "Forms for Carrying Out the Invention." Definitions and detailed explanations of terms will also be provided here. The following matters regarding the means will be described based on this premise, that is, that inspection of sentences, etc. must be performed in the same context.

[0027] The device for extracting and creating unit contexts of the present invention extracts unit sentences and conjunction phrases from any given sentence. This extraction is performed first. The sentence here is natural sentence. Natural sentence refers to sentences written in natural language. Natural language is the language we normally use, or what is commonly known as our native language. Next, the unit sentences are connected to conjunction phrases to create a unit sentence relation. Conjunction phrases are conjunctions and implications in logical inference, such as the "and" and "if" phrases in sentences like "sentence p and sentence q" or "if p then q." These conjunction phrases are used to connect unit sentences through logical inference. A unit sentence is a sentence obtained by extracting a subject, such as a subject and an object, and a predicate, such as a verb or adverb, from a simple sentence in standard Japanese grammar. However, this explanation is for ease of understanding. More details will be provided later. For the time being, the subject is considered an object phrase, and the predicate is considered an attribute phrase.

[0028] By doing this, sentence p consists of "<target phrase> {attribute phrase}". If the conjunction phrase "if" is written as the symbol "->", and "#p" means "in sentence p", then "if p, then q" can be expressed as "<target phrase #p> {attribute phrase #p} -> <target phrase #q> {attribute phrase #q}". The advantage of doing it this way is that it makes it easy to compare multiple target phrases and attribute phrases that are lined up in the same context.

[0029] "The same context" refers to a sequence of sentences (unit sentences) connected by a conjunction. Also, two sentences (antecedent and consequent) connected by implication are also in the same context. Furthermore, when a conjunction and an implication are connected by a conjunction, they are also in the same context. If we represent a sequence of conjunction sentences as "U1 ∧, , ∧ Un", and if we represent "antecedent ⇒ consequent" as "sentence A ⇒ sentence C", then a sequence connected by conjunction and implication will form the "same context" as follows: "(U1 ∧, , ∧ Un) ∧ (A ⇒ C)".

[0030] The conjunction phrase "if...then" always has the opposite condition, "otherwise." This is the mechanism (device) of the present invention. If an arbitrary sentence to be extracted (called the "original sentence") does not contain an expression equivalent to this opposite condition, this opposite condition will always be created, and a sentence that does not exist in the original sentence will exist in the sentence consisting of this unit sentence relational body.

[0031] Furthermore, sentences with opposite conditions also have the form of implication. This is because they have the form "if not, then XX." The antecedent sentence of an implication is the sentence "if not." The consequent sentence is the sentence "XX is." Then, when an implication based on this opposite condition is connected by conjunction with any sequence of conjunctions, it is expressed as "(U1 ∧, , ∧Un) ∧ (￢A ⇒ C2)." Here, "￢" is negation, and "￢A" is a negated sentence that says "not" in response to the previous antecedent A. Furthermore, "C2" is the consequent of "￢A." If there is no sentence equivalent to C2 in the original sentence, it is a new sentence that has been artificially added.

[0032] As can be seen in the addition of "C2" above, the whole sentence in which a new unit sentence is artificially added to a unit sentence relation is called a "unit sentence structure." Artificially adding a unit sentence means that a human being has considered the semantic content and added the sentence. However, we will call something that clearly takes meaning into consideration a unit sentence structure.

[0033] As will be explained later, meaning is deeply related to context. Because unit sentence structures are structures related to meaning, they can be said to be structures with various contexts. Therefore, it would be more appropriate to call unit sentence structures "unit sentence context structures." However, since the structures include context, we will omit "context" and call them "unit sentence structures" in a more informal way.

[0034] In addition, some unit sentence structures have no additional or modified words or sentences compared to the original text. Unit sentence structures differ from unit sentence relations in that their meanings have been manually examined and confirmed.

[0035] The matters related to the above "unit sentence context structure" will be mentioned later in the section [Example 1], where it is necessary to describe the means in detail. Here, in the section [Means for solving the problem], we will only describe the outline of the means.

[0036] A system (device) for sentences such as the one described above is created and used for text inspection, etc. When inspecting an original text, a sentence identical to the original text is created as a unit sentence structure, and the unit sentences and the words within the unit sentences are compared by tracing the same context. For example, if the target phrase is "O" and the attribute phrase is "A," and the unit sentence is in the same context as "<O>{A} ∧, ∧<O>{￢A} ∧," then no matter how far apart the two sentences are in the text, the attribute phrases are {A} and "￢A" (anti-A), and therefore there is a contradiction. In this way, inspection can be performed across the entire document. A system that can be processed by a computer is also provided.

[0037] In order to avoid wasting a lot of time judging the assertions and opinions of a large document, it is convenient for the document to be examined to have a unit sentence structure. An assertion expressed as a sentence, i.e., a sentence expressing the assertion "A is B," exists as a unit sentence in the unit sentence structure. Moreover, it exists as a "context" that forms a conjunction. Therefore, the context in which the assertion to be examined is asserted becomes clear.

[0038] The above is the basic concept of the means of the present invention.

[0039] 1. Contribution to the inspection and creation of large documents The present invention is expected to be systemized by computer. Here, systemization refers to the development of the present invention as an apparatus, method, or program. Anything developed through systemization will be called a system.

[0040] This system can then extract all the context from the beginning to the end of even large documents, semi-automatically, though not completely automatically. If applied to document inspection, this system can inspect the meaning of sentences and phrases for each context. This can significantly reduce inspection time and improve inspection quality. It can also be applied to document creation. The writer can inspect sentences as they are writing natural language. If the system points out any missing conditions, new sentences can be added and created. This method of looking at documents from the perspective of context is called "context-based."

[0041] 1.1. Sentence-based search for large documents When documents are structured as unit sentences, this system enables sentence-based searches for documents. Since it is a sentence-based search rather than a phrase-based search, it is a search for assertions. In other words, it shows the context in which the search statement is asserted within the document. The present invention visibly demonstrates the context-dependent nature of any assertion. Therefore, multiple search statements can be used as conjunctions or disjunctions. This type of search is performed on large documents. Moreover, if it is performed on a large number of documents, the effect is significant. When determining assertions, research time is saved and research errors are reduced. Mathematical processing such as vectors can be used to determine the similarity of assertions.

[0042] 2. Countermeasures against "knowledge risks" in generative AI In terms of industrial effects, the present invention provides a "context-based" countermeasure against so-called "knowledge risks" in generative AI. Generative AI provides grammatically correct sentences. Needless to say, grammatically correct sentences do not necessarily provide true knowledge.

[0043] The judgment of the truth of knowledge includes vocabulary errors, errors arising from the relationship between sentences, and even cases where it is impossible to judge something as true or false. Whether the knowledge is provided by generative AI or knowledge acquired casually, the judgment of truth or falsehood is ultimately made by humans.

[0044] It is not necessarily appropriate to include mathematical knowledge in the current discussion of truth judgment. However, in order to recognize the nature of truth judgment of knowledge, I would like to say that mathematical theorems that are considered true independently of human effort do not depend on the judgment of many people, but they are knowledge that depends on the "human" judgment of at least one person, often a genius.

[0045] The present invention judges the truth of knowledge "context-based," that is, it judges the truth of individual sentences in light of the context. This judgment is an empirical judgment, which is said to be the polar opposite of mathematical judgment. When making an empirical judgment, making it in accordance with the context is a good support for judging the truth of knowledge from confusing or convoluted sentences. This is because generative AI can easily create malicious, confusing, and convoluted sentences.

[0046] The "context-based" approach of the present invention provides a means of defense against this. The present invention makes it possible to clearly indicate the context and to closely examine unit sentences. These techniques help determine the authenticity of knowledge against "human intervention."

[0047] 2.1. Authentication of Identity of One's Own Theory One risk of generative AI is the risk that authentication of the identity of one's own theory may be compromised. This is because generative AI can easily create multiple theories. By constructing one's own theory as a "context-based" set of contexts consisting of conjunctions and the like using the present invention, the identity of one's own theory can be logically and therefore clearly authenticated. A theory constructed in this "context-based" manner can be compared with other theories via sentences, making the differences clear. By utilizing technologies such as blockchain, attributes of such a theory, such as the creation time and history of the theory, can also be trusted.

[0048] 3. Meeting International Standards When it comes to exporting products, it's already assumed that "if you make something good, it will sell," so meeting international standards is a given. There's also a strategic theory that raises the question of why Japanese companies lose out even when they make good products ("Rule-making Strategies to Win in the Global Market by Creating International Standards: Why Japanese Companies That Excel in Technology Lose" (author, Tama University Rule-making Strategy Institute, published by Asahi Shimbun Publications)). This view holds that creating international standards and market strategy should be considered equally. This seems reasonable. However, we should perhaps ask why it's difficult for Japanese people to conceive of the idea of ​​creating international standards.

[0049] Germany has a strong influence on the International Organization for Standardization (ISO) in areas such as international standards for automobile safety. It's an old story, but the German scholar Bälz, who was a "housekeeper" in the early Meiji period, criticized the Japanese academic approach for not questioning the basics (Bälz's Diary, published by Iwanami Bunko). Bälz said that Japanese people tend to seek immediate utility and end their academic inquiry there. Since this was education in the early Meiji period, when Japan was striving to enrich the nation, Japanese scholars and students had no choice but to seek immediate utility.

[0050] However, there are still unfortunate surveys today. Currently, 40% of graduate students in Japan are in engineering, while in Europe and the US it is around 10% (What is Science?, Author: Sakura Osamu, Publisher: Kodansha). These figures may be enough to suggest that Japan still places utility first in academia.

[0051] Not only automobiles, but also autonomous mobile devices require proof of safety based on international standards for safe operation. Increasing the number of autonomous mobile devices is an area where great expectations are high amid a declining population, aging society, and labor shortages. However, the more AI is installed and the more advanced the device becomes, the stronger the need for safety assurance. While demonstrating safety is essential, proving it is not required in the curriculum of Japanese engineering schools. This is where the "context-based" logic of this invention proves its effectiveness.

[0052] 4. Responding to a rational society What is the origin of "logic" in the first place? The economist Schumpeter is famous as the pioneer of innovation theory. He reflected on why humans think rationally. He argued that everyday economic needs are the source of humans' rational attitudes (see the section "The civilization of Capitalism" in Joseph A. Schumpeter's "Capitalism, Socialism, and Democracy"). He went on to state that all logic originates from patterns of economic decision-making.

[0053] Setting aside the pros and cons of the statement that logic originates in everyday economic behavior, it would not be a big mistake to consider rational spirit and logic to be the same thing. Logicism is inseparable from rational spirit. The "context-based" approach of this invention supports rational thinking.

[0054] International political conflict is a formidable obstacle to industrial development. If conflicting political discourse is written down and analyzed "context-based," the cracks in the context can be seen. New context can be added to the cracks, filling them. Might this contribute somewhat to political security? When political conflict is reflected in text, it is not as incomprehensible as Chomsky's example sentence 1 above. Later, Example 1 can be interpreted as an understandable sentence, Example 1-1. Might this be likened to eliminating the cracks?

[0055] Given the current conflicts in international politics, it is inevitable to say that peace is the foundation of industrial activity. The inventor hopes that the application of "context-based" technology will be widely deployed in the political world, thereby contributing to industry.

[0056] 5. Educational Applications The poet Ishikawa Takuboku is said to have disliked the grammar of Japanese. Perhaps this genius poet found grammar to be an obstacle to his creativity. The linguist Chomsky argued that humans have an innate structure for understanding language. Just as there is an innate structure for language, perhaps there is also an innate structure for understanding context. This is because, as mentioned above, it was "context" that made Chomsky's nonsensical sentences understandable. While this innate structure does not necessarily refer to the grammar that Takuboku disliked, it is not unrelated. At the very least, by helping students understand the "context base" of what I describe as unit sentences, their sequences, and divisions, it could serve as the basis for writing long, logical sentences. This level of understanding would not inhibit a poet's creativity.

[0057] Japanese language education must be a time for young children to expand on their questions of "why and how?" and to express them clearly. A "context-based" approach is effective. Logic transcends ethnic cultures and is a means of conveying meaning common to all humanity. Internationalization requires common means of communication. Business requires people with strong logical skills.

[0058] Context-based document inspection device (architecture diagram of the entire document inspection program device) Document map creation unit (subprogram. Architectural diagram of the document map creation unit) Document inspection unit (subprogram. Architectural diagram of the document inspection unit) Context-based document search device (architecture diagram of the document search program device) Diagram explaining dependency analysis of natural language Diagram of the path of unit contexts from unit sentence structures Diagram of the entire document structure (table of contents, headings, sentences) Example of the table of contents of a document Diagram of the entire document structure (symbolic representation of the sentence section) Example of document creation standards

[0059] 1. Starting Point and Central Concepts of the Invention Borrowing the proverb "necessity is the mother of invention," the starting point of this invention is a problem. The problem is that there are many unclear points when checking someone else's document. While unraveling the causes of this problem, the inventor realized that there is a general application to it. Below, we will take up document-related problems again and develop concepts that play a central role in solving these problems. This will be referred to as Example 1.

[0060] 1.1. Starting point of the present invention: Idea 1.1.1. The power of "context": even meaningless sentences depend on context The unclear sentence in Example 1 above can be understood as follows by adding context.

[0061] [Sentence example 1-1] It was a colorless world, but that world once had color. As a result, only words remain. For example, green means abundant harvest. Therefore, when we say "colorless green thoughts," these thoughts mean "abundant thoughts" in a colorless world. However, if those thoughts are also "asleep," this means that they have nowhere to live and are in a dead state. Moreover, if that sleep is described as "furious," this means that the person who has those thoughts is in a state that is the exact opposite of the quietness of sleep; they have their eyes closed and appear to be asleep, but in reality they are completely angry and awake.

[0062] This is the interpretation of "Colorless green thoughts sleep in a rage" (Example 1). Context plays a role in the interpretation. This is because a sentence that was completely incomprehensible becomes understandable by adding context.

[0063] 1.1.2. Words that leave you scratching your head can be resolved depending on the context. While not as unclear as Example 1, there are occasional sentences that leave you scratching your head. Here is an example from a tourist pamphlet. If you look at the picture in the pamphlet, you can understand the meaning, but not from the words alone.

[0064] [Example sentence 2] The sun appears in the morning and disappears at night. However, on festival days, the sun appears many times and people dance and sing all night long.

[0065] The reason we don't know is because the state of the sun changes in many ways. So, we surround the sun with <> and the state of the sun with {}. This gives us the following four states of the sun. We numbered the sentences 01, 02, etc.

[0066] [Sentence example 2-1] 01 The sun {appears in the morning}, 02 The sun {goes away at night}, 03 The sun {appears many times}, 04 The sun {dances and sings all night long}

[0067] All the suns are personified. However, anyone who sees the tourist pamphlet will not have any doubts. The overseas town facing the sea has beautiful sunrises and sunsets, and in a scene that appears to be a festival, many dancers are dancing passionately. That's how it is depicted in the pamphlet. Of the four suns, the last two are actually dancers. You can't tell from the text alone.

[0068] However, in the text-only example 2, there is a sentence that says, "But only on festival days." This sentence suggests that the first two suns may be different from the last two suns. This is one way of imagining it. So, if we add "But only on festival days" between the lines and rearrange the sentence, it becomes as follows:

[0069] [Sentence example 2-2] 01 The sun {appears in the morning}, 02 The sun {goes away at night}, 03 But on festival days, 04 The sun {appears in large numbers}, 05 The sun {dances and sings all night long}

[0070] The sentence "But on festival days" in line 3 separates the sun of nature from the sun of the dancer. Line 3 can be said to represent the boundary between contexts. This boundary means that the top two lines belong to the same context, while the bottom two lines belong to a different context. This way of looking at it is possible. The line that creates the boundary (line 3 in this example) will be called the "line that creates a new context," or simply the "new context."

[0071] It's a bit hesitant to generalize from this small example, but as the new context of this example shows, in some contexts two words can have different meanings even if they are made up of the same letters.

[0072] If we look closely at the differences in meaning in this example, comparing the multiple "suns" in Example 2-1, there is a sun that "goes away at night" and another that "dances and sings all night long." If both suns exist, it is a contradiction.

[0073] However, as we have already seen, this contradiction can be resolved by assigning the natural sun and the dancer to the two phrases. This is a "homonym" where one word has multiple meanings, and is usually called a "polysemous word." And, as shown in line 3 of Example 2-2, it can be said that the context creates polysemous words. Or, it can be said that it makes them valid.

[0074] On the other hand, when two words made up of the same characters belong to the same context, the two words have the same meaning. This is shown by the difference in the belonging of the words at the boundary of line 03 in Example 2-2.

[0075] 1.2 Central Concept of the Present Invention For the purposes of the following discussion, we will introduce one concept again. This concept plays a central role in the present invention. The claims of the present invention are based on this concept.

[0076] This concept, to put it another way, refers to formality. It also includes utilizing formality. For example, in the small example above, we checked the order of the sentences. This is considered to be arranging a "shape." "Arranging a shape" will be called "formalization." The product created by the act of arranging a shape will be called a "formal object," and the collection of these will be called a "formal body."

[0077] Furthermore, "formalization" refers to clearly expressing context within a sentence. In other words, it refers to making obscure or hidden context apparent within a natural language expression. Dividing a sentence into several phrases is also an act of formalization. Furthermore, adding context is also considered to be included in "formalization" in the sense that it advances formalization.

[0078] 1.2.1. Basic Concepts that Compose Sentences The concepts of the present invention will be further explained. To this end, some basic concepts that are considered to compose sentences will be described. An example sentence will be given, and the form that composes the sentence, that is, the basic elements, will be extracted. One of these basic elements is the following "unit sentence." This is a phrase unique to this invention. The previous example sentence 2-2 will be used to explain this phrase.

[0079] 1.2.1.1. Unit Sentence, Target Phrase, Attribute Phrase As in "<The sun> shows {its face in the morning}" in line 01 of the previous sentence example 2-2, a sentence consisting of < > and {} is named a unit sentence. The phrase enclosed by < > is named the "target phrase". The phrase enclosed by {} is named the "attribute phrase". Assume that auxiliary words can be excluded from the unit sentence.

[0080] In addition, a sentence written in the mother tongue as written in sentence example 2 is called a sentence written in a natural language. In the present invention, it will be simply called a "natural sentence". Therefore, it can be said that a unit sentence is artificially or manually constructed and extracted from a natural sentence.

[0081] The unit sentence uses a simple sentence as a hint. "The sun shows its face in the morning" is regarded as a simple sentence in normal mother tongue grammar. A simple sentence consists of a subject and a predicate. In the present invention, this subject is used as the target phrase, the predicate is used as the attribute phrase, and what is reconstructed with the target phrase and the attribute phrase is called a unit sentence. In the present invention, what consists of an object and its attribute is regarded as the basic unit of a sentence. Therefore, "The sun shows its face in the morning" is constructed as "<The sun> shows {its face in the morning}" as a unit sentence.

[0082] The target phrase does not refer to only one word. It may refer to a collection of multiple targets. Also, a state in which the target is modified by several phrases is also a target phrase. Since it includes such cases where there are many words, the expression "phrase" is used to refer to the target. And since a collection is regarded as a target, the target is regarded as one. For example, if two targets O1 and O2 are expressed as "O1 and O2", this is regarded as one target. Also, the expression "O1 or O2" is also regarded as one target.

[0083] The attribute phrase does not refer to only one predicate. A phrase in which multiple words are connected, which should also be called a predicate part, is also an attribute phrase. This is because it is regarded as one unit. In the previous {shows its face in the morning}, "in the morning" and "shows its face" are in a relationship where they are related to "shows". Although it consists of multiple words, it is regarded as one attribute. Also, when comparing attribute phrases, the continuous form of verbs and the like that make up the attribute phrase should be compared in the terminating form.

[0084] 1.2.1.2. Conditional phrases and conditional symbols In the third line of Example 2-2, "But, only on festival days," the condition "only" is used to indicate the difference or boundary between the previous and following sentences. Expressions like this are called "conditional phrases." "If...then" is a typical conditional phrase.

[0085] To formalize things further, let's express a condition with "if," as is often the case. Let's also introduce "else" to express the opposite condition of "if." Let's also introduce "endif" as the end of the set (pair) of "if" and "else." These are symbols. Hence, we name them "condition relation symbols."

[0086] In the present invention, the Japanese expressions "if," "otherwise," and "end" are also conditional phrases. These are phrases extracted from Japanese sentences, and therefore, when used in Japanese sentences, they are conjunction phrases. On the other hand, when they are used as part of expressions in unit sentences, as a substitute for "if" or "else," they are conditional symbols.

[0087] In addition, "if not" is sometimes expressed as "when not" in Japanese sentences. There are also slight differences in expression for other conditional phrases. When conditional phrases with slight differences are unified into one expression, the conditional "symbol" is used.

[0088] There is a question as to whether "if" always has an "else" pair. For example, the opposite condition of "But, only on festival days" in line 03 of Example Sentence 2-2 is not a festival day, so it would be an ordinary day. If we ask whether anything about ordinary days is written, we find that ordinary days are already written in lines 01 and 02. In this case, if we set a line before line 01 and make it line 00, we can say that the condition "if it's an ordinary day" is implicitly hidden in line 00. In that case, "But, only on festival days" in line 03 becomes "else." In this way, since it is possible to extract implicit conditions to the surface, it seems reasonable to create a structural form in which "if" and "else" always exist as a pair.

[0089] From the perspective of this invention, the system automatically generates "else" when there is a formal "if." After that, a human must reconsider the meaning. When reconsidering, the semantic content is examined, such as whether the "else" condition actually has no consequent.

[0090] 1.2.1.3. Conjunction phrases and conjunction symbols There are expressions that connect sentences. They are "Sentence A, and sentence B." However, there are no phrases between lines 01 and 02 in Example 2-2. However, according to logical inference, the sentences in the two lines are connected by a conjunction. Expressed as a phrase, it is "and." "And," "furthermore," and other words also function to connect sentences in the same way as "and." We will call such expressions that connect sentences "conjunction phrases." Expressed as a symbol, it is "∧." We will call this a "conjunction symbol." "And" can also be treated as a symbol. This treatment is the same as for words such as "if."

[0091] 1.2.1.4. Conjunction phrases and conjunction symbols Here, we introduce the concept of "conjunction phrases." In this invention, conditional phrases, conjunction phrases, and negation phrases, which are the negation of sentences, are considered to be elemental conjunctive phrases. They are defined as phrases that connect sentences. Furthermore, "if," "∧," and the negation symbol "￢" are "conjunction symbols."

[0092] 1.2.1.4.1. Handling of disjunction-related phrases Naturally, the expression "or" exists as a conjunction-related phrase. This has the function of selecting one of the sentences. In logical inference, this is called disjunction. There are two types of disjunction: inclusive disjunction and exclusive disjunction. These disjunctions can be expressed by implication, conjunction, and negation if the condition-related phrase "if...then" is considered to be an implication of logical inference. Therefore, disjunction is not used as a basic concept of this invention.

[0093] However, this does not mean that disjunctions are not used as conjunctions. Disjunctions are used when extracting unit sentences from natural language because natural language contains expressions such as "sentence A or sentence B," "sentence A or sentence B, or both," and "either sentence A or sentence B."

[0094] When such expressions appear in natural language, for example, a disjunction symbol "∨" is used to connect unit sentences according to the type of disjunction. However, in the sentence inspection process described later, the disjunction is reduced to a conjunction-related phrase consisting of three elements and inspected. Therefore, in this invention, the symbols for disjunction-related phrases according to the type of disjunction are not specifically defined.

[0095] 1.2.1.5. Unit Sentence Context Structure The entire sentence example 2-2 can be considered a structure consisting of unit sentences and conjunction phrases. Here, the conjunction phrases consist of conditional phrases and conjunction phrases. The structure can be likened to a building (architecture), consisting of pillars and walls. However, the conditional phrases do not extend vertically upward like pillars. Conditional phrases appear here and there throughout the sentence and control the flow of the sentence. Conditional phrases are like large rivers or canals with numerous tributaries, similar to water veins, mountain ranges, or blood vessels. Furthermore, unlike the branching of conditional phrases, conjunction phrases are linear connections between these various "veins." However, because these conjunction phrases do not form a geometric structure like the interior of a building, they will be broadly referred to as a "structure." Therefore, the entire sentence is a structure consisting of the "venus" of the sentence, that is, a structure consisting of context. Since a context consists of unit sentences, the entire sentence can be called a "unit sentence context structure."

[0096] 1.2.1.6. Unit sentence relational entities Although it may be confused with the above "unit sentence context structures," we will introduce the concept of "unit sentence relational entities," which are not "structures." Both have the same formal structure. However, in terms of the generation process, "unit sentence relational entities" belong to the process before "unit sentence context structures." "Unit sentence relational entities" are entities that faithfully reconstruct a given original sentence using unit sentences and connective phrases. On the other hand, "unit sentence context structures" are entities that have been arbitrarily modified by humans to "unit sentence relational entities." "Unit sentence relational entities" are equivalent to the original sentence. "Equivalent" here means that the simple sentences of the original sentence and the unit sentences of the unit sentence relational entities correspond one-to-one and have the same truth values.

[0097] In the following, for the sake of simplicity, the word "context" will be omitted from "unit sentence context structure" and it will be called "unit sentence structure." The term "structure" has already been used in "unit sentence relation structure." A structure contains both form and semantic content. On the other hand, a relation structure does not contain semantic content, but only contains form. This has been mentioned elsewhere. The "unit sentence structure" was also mentioned earlier in the section on "Means for Solving the Problem."

[0098] Using the above tools, we create unit sentence relations and unit sentence structures. Example 3 is used as the original sentence. (Quoted from "The Road to Dara-e Noor" (by Nakamura Satoshi))

[0099] [Example sentence 3] Everyone will inevitably die. Naturally, it is an inescapable law for living beings. No matter how much we boast about our possessions or how much fame we have, we cannot take it with us to the afterlife. When that time comes, it is more important that the traces of our lives leave some kind of warmth behind, and that we convey the warmth and truth of being human.

[0100] The following example sentence 3-1 is a unit sentence relation. The blank spaces <> and {} in example sentence 3-1 indicate that the phrase is not written in the original sentence.

[0101] [Sentence example 3-1 (unit sentence relation)] 01 <People> {will surely die}. 02 <> {naturally}, but 03 <> {it is an unavoidable law for living beings}. 04 If {how proud} of <many possessions>, 05 If {how much fame} you {have}, 06 <it> {cannot be taken away to the afterlife}. 07 else 08 <> {} 09 endif 10 <some warmth> {the traces of our lives left behind at that time}, 11 <it is more important} to convey the warmth and truth of being a human being. 12 else 13 <> {} 14 endif

[0102] However, in reality, Example Sentence 3-1 is artificially created here because the two sentences on lines 10 and 11 were placed on those lines by the writer of the present invention after he or she interpreted Example Sentence 3. However, the arbitrariness is limited to that location.

[0103] However, based on the definition of a unit sentence relation above, Example 3-1 cannot be called a unit sentence relation because it contains artificial arbitrariness. From that point of view, it should be called a unit sentence structure.

[0104] The distinction between unit sentence relations and unit sentence structures is not something that needs to be strongly emphasized, since ultimately unit sentence structures are what are formally recognized as sentences made up of unit sentences.

[0105] Nevertheless, the reason for distinguishing between the two is that in the case of unit sentence relational structures, automatic processing by computer can be introduced. In document inspection, which will be discussed later, it is unproductive to manually create unit sentences one by one. In order to improve productivity, it is necessary to create unit sentence relational structures as automatically as possible.

[0106] The two sentences in lines 10 and 11 of Example 3-1 were stated to be arbitrary by the writer. However, the inventor added the label "unit sentence relation body" to Example 3-1 in the hope that this level of automatic processing by a computer would be possible.

[0107] The automation will be described in [Example 2] (see Figure 5). However, this automatic processing is not unique to the present invention, as it uses existing language analysis tools. What is unique is the format of the unit sentences and unit sentence structures created using language analysis, or the device, so to speak. These are the claims of the patent.

[0108] The blanks in sentence example 3-1 are manually filled in. The inventors have arbitrarily named it sentence 3-2. As a result, the unit sentence relation body is considered to have become a unit sentence structure.

[0109] [Sentence example 3-2 (unit sentence structure)] 01 <People> {will surely die}. 02 <Death> is {natural}, but 03 <Death> is {an unavoidable law for living beings}. 04 if {how proud} of <many possessions>, 05 if {how much fame} you {try to gain}, 06 <you cannot take everything you own> {without taking it away to the afterlife}. 07 else (Note: This is a case where you {don't try to gain any fame}) 08 <give everything you own> {to society, concealing your name}. 09 endif 10 <some warmth> {the traces of our lives leave behind}, 11 <it is more important} to <convey the warmth and truth of being a human being>. 12 else (Note: This is the case when "one does not boast of having many possessions") 13 {It may be permissible to keep all of one's possessions quietly}. 14 endif

[0110] The unit sentence structure is further shaped (structurally formalized). The unit sentence structure of Example 3-2 can be expressed using only the shape (structural form) by substituting symbols (U1, A, B, etc.) as unit sentences. The result is Example 3-3. Note that the logical inference "implication" is applied to the conditional relation phrase "if" in Example 3-3. The symbol "⇒" is used. This is also added to the conditional relation symbols. The unit sentence structure of Example 3-3 below is written using symbols, so it can be called a "symbolic unit sentence structure."

[0111] [Sentence example 3-3 (symbolic unit sentence structure)] 01 U1 02 U2 03 U3 04 if A 05 if S 06 ⇒ T 07 else 08 ⇒ U 09 endif 10 B 11 B2 12 else 13 ⇒ C 14 endif

[0112] 1.2.2. Structural Form The above example sentence 3-3 is entirely made up of symbols. For example, the unit sentences in lines 01 to 03 (U1 to U3) are connected by a conjunction (the symbol "∧") in terms of inference rules. This structure is expressed as "U1 ∧ U2 ∧ U3". The phrase "symbolic unit sentence structure" will be called a "structural formalism" by placing emphasis on the "form" of "symbols."

[0113] The reason for placing emphasis on the "form" of the "structure" is as follows: The expression "symbolic unit sentence structure" is an expression that places emphasis on "symbols." In fact, Example Sentence 3-3 consists only of symbols. On the other hand, a "unit sentence" inherently has meaning, since the object phrase and attribute phrase of the unit sentence are actual words. However, since symbols do not contain meaning, "symbols" and "unit sentences" are contradictory. Therefore, we have placed emphasis on "symbols" in "symbolic unit sentences" and replaced them with "form," and as a structure, we have called it a "structural formal body," in the sense that emphasis is placed on form.

[0114] Furthermore, when emphasis is placed on the relationships and connections between sentences, the term "structural form" is used. This term is obtained by removing the "body" from "structural form body." As mentioned above, "body" refers to a collection of deliverables. When emphasis is placed on the relationships between the elements that make up a deliverable, the "body" is removed and only "structural form" is used. Here, "elements" in this invention refer to unit sentences, as well as target phrases and attribute phrases that are elements of unit sentences. Below, "context" will be explained using "structural form."

[0115] In terms of context, in Example 3-3, lines 04 (A) and 05 (S), which contain conditional relation symbols, become the conditions of the conjunction and become the beginning of a new context. The structural form of the context, including the consequent, is "A∧S⇒T". Lines 04 and 07 (negation of line 05) and line 12 (negation of line 04) also become conditions of the conjunction and become the beginning of a new context. Their structural forms are "A∧￢S⇒T" and "￢A⇒C", respectively. "￢" indicates the negation of a sentence.

[0116] On line 9 there is an "endif" which indicates the end of the statement. The next line, line 10, contains a new statement. This continues on line 11, which forms a conjunction just like lines 1 to 3. It is "B ∧ B2". However, unlike "U1 ∧ U2 ∧ U3" from line 1, the statements on lines 10 and 11 are under the context of the conditional relation symbol on line 4.

[0117] "U1 ∧ U2 ∧ U3" from line 01 is not under any conditional relation symbol and is independent. In that sense, it can be called a "single sentence" (single unit sentence).

[0118] On the other hand, lines 10 and 11 are "subordinate sentences" (subordinate unit sentences). The subordinate, or so to speak, master (conditional relation symbol) is line 4. The master ends with endif on line 9, so the subordinate sentences are connected to lines 6 and 8 above it, respectively. Expressed symbolically, traversing line 6 passes through lines 4 and 5, resulting in the structural form "A∧S⇒T∧(B∧B2)". Similarly, traversing line 8 passes through lines 4 and 7, resulting in the structural form "A∧￢S⇒T∧(B∧B2)".

[0119] The parentheses "( )" in (B∧B2) are not necessary because the symbol combination rules are the same as general logical inference rules. In other words, the combining symbols "∧" and "⇒" have stronger binding power in that order.

[0120] 1.2.2.1 Context Tracing and Unit Context The structural forms described above can be considered as paths through a sentence. Listed in the order of the above descriptions, they are as follows. Example sentence 3-4 below is annotated as "structural form."

[0121] [Sentence example 3-4 (structural format)] (a) U1∧U2∧U3 (b) A∧S⇒T (c) A∧￢S⇒U (d) ￢A⇒C (e) B∧B2 (f) A∧S⇒T∧B∧B2 (g) A∧￢S⇒U∧B∧B2

[0122] All of the above, i.e., each of the paths from (a) to (g), can be referred to as a "context." Furthermore, the top five of (a) to (e) can each be called an "element context" because they are the smallest units. In this case, an element context can be understood simply as a collection of unit sentences separated by an appropriate break. An appropriate break refers to the boundary between the final unit sentence of an independent sentence and the first unit sentence of an implication sentence. As can be seen in Example 3-4, it is the boundary between the final unit sentence "U3" in (a) and the first unit sentence "A" in (b) and (c). Furthermore, the "￢A" at the beginning of (d) also separates from the final "U3" in (a).

[0123] The consequent of an implication sentence is also a break. Both the consequent "T" (unit sentence) in (b) and the consequent "U" in (c) form a break. However, this break has a continuation. At the end of "T" and "U" (endif on line 9) there is another higher-level conditional relation phrase (if, else). Therefore, even if "T" and "U" leave the endif on line 9, they will not end until they reach the next higher-level endif. If there are any unit sentences before reaching the next higher-level endif, the sentence tracer must follow them and exit the exit (endif).

[0124] In fact, the structure of Example Sentence 3-3, which is the base of the structural form of Example Sentence 3-4, has unit sentences "B" and "B2" on lines 10 and 11. Therefore, both "T" and "U" trace "B" and "B2", and because line 12 is else (an instruction that "traversal ends here and does not go to the next line"), they skip over line 13 and reach the exit (endif on line 14).

[0125] In Example 3-4, this process is represented by (f) and (g). In other words, (b) and (c) follow "B∧B2" in (e) after leaving the endif on line 9, and the path they take is (f) and (g).

[0126] We will refer to (f) and (g) as a series of unit sentences, which we will call "unit contexts." As mentioned earlier, when a unit context is implied, there is a beginning. In the case of (f), the beginning of the context is the antecedent "A∧S," and in the case of (g), the beginning of the context is "A∧→S."

[0127] A series of unit contexts is also called a unit context. In this invention, "unit context" is a common basic term related to contexts. Later, the term "global context" will be introduced. This global context is defined in terms of unit contexts.

[0128] From the perspective of a reader, a sequence is a continuous tracing of unit sentences. The "sequence" and "tracing" are made possible by conjunction phrases. (Conjunction phrases are conjunction phrases, conditional phrases, or negation phrases.) Multiple continuous tracings (connections) of unit contexts can be identified by symbols or names, as in (a) through (g).

[0129] 1.2.2.2 Principles of Document Inspection Incidentally, inspection of a document must be carried out on the entire document, or in other words, on the entire "whole area." Inspection of only a portion of the document is insufficient.

[0130] 1.2.2.2.1. All paths Figure 6 shows all of the paths of a sentence. The diagram illustrates the symbolic unit sentence structure of Example 3-3 and the structural form of Example 3-4. By tracing all of the paths, the entire range of the sentence is covered.

[0131] Let us explain Figure 6. The lines in Figure 6 trace the structural path of Example Sentence 3-3 from the first line to the last line. The dots "・" in the lines represent unit sentences. The result of the traced path is Example Sentence 3-5.

[0132] [Example sentence 3-5 (path of structural form)] (α) U1 ∧ U2 ∧ U3 ∧ A ∧ S ⇒ T ∧ B ∧ B2 (β) U1 ∧ U2 ∧ U3 ∧ A ∧ ￢ S ⇒ U ∧ B ∧ B2 (γ) U1 ∧ U2 ∧ U3 ∧ ￢ A ⇒ ∧ C

[0133] The paths are represented by unit sentences and connecting symbols as (α), (β), and (γ), respectively. In each of (α), (β), and (γ), an arbitrary element context is connected to an element context that is bifurcated by a positive or negative conditional relation phrase. In summary from the perspective of logical inference, all paths are made up of conjunctions of element contexts consisting of conjunctions of unit sentences and element contexts consisting of implications that cause bifurcations in positive and negative directions. Many branches are created by combining binary branches. These become contexts.

[0134] A text consisting of many sentences, i.e., a large document, is composed of many contexts, combining conjunction contexts and implication contexts through the antonymy of implication. Therefore, when tracing from the beginning to the end of a sentence, the process involves starting from the root of the binary tree array and reaching the terminal leaf, then going back from the terminal to the previous branch point, and then tracing back to the terminal again. This is generally called a "depth-first search." Figure 6 uses this depth-first search method, although the number of paths is small. In Figure 6, this search method is used, and once a terminal leaf is reached, it is considered a single context. Then, when "traversing from the terminal to the previous point," the returned part (P) is treated as a separate traversal, creating a new independent path. In this case, the returned part P is returned to the beginning again (B), and a new path is traced from the beginning to the terminal, all the way to the final terminal leaf. This path includes path B, which returns to the beginning, and path P, which leads to the previous branch point. Incidentally, the order of the unit sentences that make up B and P is not the order of return, but the order from the beginning.

[0135] 1.2.2.2.2 Global Context The depth-first search traverses unit sentences from the beginning to the end of a sentence throughout the entire sentence. The ordered set of unit sentences traversed throughout the entire sentence is called the "global context."

[0136] Even if the beginning of a global context is a common unit sentence, it branches out along the way. Multiple global contexts are created. These contexts are created by tracing them individually. We will call such contexts "individual global contexts."

[0137] As mentioned above, the set of unit sentences in a global context can be either elementary contexts or unit contexts, which are conjunctions of elementary contexts. Since elementary contexts are also included in unit contexts, the global context can be said to consist of unit contexts from the beginning to the end of a sentence.

[0138] The example paths (α), (β), and (γ) are individual global contexts obtained in this way, even though they are very short sentences. The paths (α), (β), and (γ) are different paths, and each one represents the beginning of the sentence and the end of the sentence.

[0139] 1.2.2.2.3. Components of a global context Let's check the components of a global context. Figure 6 also shows element contexts (a), (b), (d), and (e). (However, (c), (f), and (g) were not shown properly, so they are omitted.) Then, (α) is composed of "(a), (b), (e)" or "(a), (f)". (β) is composed of "(a), (c), (e)" or "(a), (g)". (γ) is composed of "(a), (d)".

[0140] 1.2.2.2.4. Actual Inspection Inspection is performed on the global context. For example, (α), (β), and (γ) in sentence example 3-5 are one of the components that make up the global context. When inspection is performed on all of these individual global contexts, it can be said that inspection is performed on the entire sentence. This is because all of the individual global contexts, i.e., the global context, cover the entire path of the sentence. In the case of a large document, the inspection is cross-sectional, going beyond the chapters, sections, and paragraphs of the document. This is different from the small sentence in the example here. All individual global contexts are subject to inspection.

[0141] 1.2.3. Evaluation of meaning Evaluation of meaning is carried out after examining the structural form. The structural form examination process is a preliminary step to semantic evaluation, and is intended to organize the structural form in order to carry out semantic evaluation effectively. Evaluation involves cognition, thinking, and judgment, so these cognitive acts can be described as "artificial."

[0142] To illustrate the inspection process, Example Sentence 3-5 shows all the paths in Example Sentence 3 as a structural form. The meaning is evaluated by recognizing, thinking about, and judging sentences and words based on this structural form.

[0143] Because evaluation is artificial in this way, evaluations may differ depending on the evaluator. However, the approach is to adopt a certain method, namely, to make the evaluation formal. This is expected to make it possible to clarify large differences in evaluations and the basis for the evaluation.

[0144] 1.2.3.1. Creating sentences for semantic evaluation The semantic content of the three unit contexts in sentence example 3-5 was evaluated. To do this, sentences for evaluation were created according to sentence example 3-5 (α), (β), and (γ). The corresponding sentences are (A), (B), and (C) in sentence example 3-6 below.

[0145] [Example Sentence 3-6] (A) Everyone will surely die. Naturally, this is an inescapable law for living beings. No matter how much we boast about our possessions or how much fame we gain, we cannot take it with us to the afterlife. When that time comes, it will be more important that the traces of our lives leave some kind of warmth behind, that they convey the warmth and truth of being human. (B) Everyone will surely die. Naturally, this is an inescapable law for living beings. No matter how much we boast about our possessions, if we make no effort to gain fame, we will give it all back to society, anonymously. When that time comes, it will be more important that the traces of our lives leave some kind of warmth behind, that they convey the warmth and truth of being human. (C) Everyone will surely die. Naturally, this is an inescapable law for living beings. However, if we never boast about having many possessions, we may be allowed to keep all of our possessions quietly.

[0146] 1.2.3.2. Example of Semantic Evaluation (A) remains the original sentence (Sentence Example 3). The unit sentences in lines 01 to 06 and lines 10 and 11 of Sentence Example 3-2 are reflected in (A). The continuation of the sentence returns to the original sentence even after passing through the unit sentence structure, so there is no problem. Therefore, the semantic content is the same as the original sentence and there is no problem.

[0147] In (B), lines 01 to 03 of Example Sentence 3-2 occupy the beginning of (B), just like in (A), and regulate the entire sentence. Following this beginning are lines 04, 07, 08, 11, and 12. This shows (β) in Figure 3, and the continuation of these sentences is correct.

[0148] On the other hand, in terms of meaning, line 7, which is the negation of line 5's "How to gain fame," is a case of "not trying to gain any fame at all." "How to gain" is interpreted as "how much to gain," and its negation becomes "not trying to gain any at all." The next line, line 8, then reads, "I will give back all of my possessions to society, anonymously." (Note that phrases such as "all" follow the rules of quantification in logical inference.)

[0149] Continuing with the interpretation of Example 3-2, if line 7 says "not trying to obtain any fame," it can be interpreted as "not having anything to do with fame," and therefore line 8 could mean "spending his vast fortune ostentatiously, without hiding his name, for his own gain," without giving it back to society. However, if interpreted this way, it is semantically inconsistent with the following lines 10 and 11, which say that he values ​​"the warmth of something" and "human warmth and conveying the truth."

[0150] If line 8 of Example 3-2 is thus about self-interest ("For self-interest, without concealing one's name, and using it ostentatiously"), then lines 10 and 11 of Example 3-2 should be moved and inserted after line 6. This would result in a change in the unit sentence structure. This would result in a change in the rest of the sentence, that is, a change in context.

[0151] The result of the change is shown in Example 3-2-1. The line numbers remain the same as before. The digit offset also remains the same as before.

[0152] [Example sentence 3-2-1 (unit sentence structure)] 01 <People> {will surely die}. 02 <Death> is {natural}, but 03 <Death> is {an unavoidable law for living beings}. 04 if {how proud} of <many possessions>, 05 if {how much fame} we {gain}, 06 <we cannot take it away with us to the afterlife>. 10 <some warmth> {leaves a trace of our lives at that time}, 11 <it is more important} to <convey human warmth and truth>. 07 else (Note: This is a case where {no attempt is made} to gain <fame> at all). 08 <give it all back to society, anonymously}. 09 endif 12 else (Note: This is the case of "not boasting about having many possessions at all") 13 It may also be permitted to keep all of one's possessions secretly. 14 endif

[0153] 1.2.3.3. Objections to semantic evaluation and regularity of structural form Objections to semantic evaluation can arise from the arbitrary inclusion of words or sentences in the blank spaces of unit sentences or blank lines in the unit sentence structure. This was done previously in Example 3-2. Objections can be made to the meaning, but the way the sentence continues, i.e., the structural form, does not change.

[0154] The existence of this formality suggests that there may or may not be regularity in the process of extracting the continuation of a unit sentence from a unit sentence structure, and if there is, it could be processed by a computer. However, since human interpretation is required for meaning, it is not easy to apply computers to this. However, as we have just seen with context change, by taking meaning into consideration in accordance with the context, i.e., according to the form, it seems likely to support semantic testing, which tends to be dependent on personal arbitrariness and tends to be chaotic. For example, because a unit sentence consists of a target phrase and an attribute phrase, comparing target phrases in the same context avoids the waste of comparing meanings while leaving the context unorganized. Therefore, it is promising to utilize computers.

[0155] 1.2.3.3.1. The effect of context organization on wastefulness due to unorganized context Let's return to the issue of the role of context in examining texts. Is it really possible to avoid the wastefulness of comparing meanings without organizing the context through context? This is the next topic of discussion.

[0156] The first half of (C) in Example 3-6 is the same as (A) and (B). On the other hand, the second half of (C) is "However, if one never boasts about his many possessions, one may be permitted to keep all one's possessions quietly," which does not go through lines 10 and 11. This is because it does not go through line 4, "if one boasts about his many possessions," but goes through its negation, line 12.

[0157] However, given the original intent of author Nakamura Tetsu, is it permissible to take one's property to the grave after death? Perhaps, or rather, certainly, Nakamura's argument does not allow leaving property after death. Therefore, lines 10 and 11 should, in their original interpretation, come after the last line of Example 3-2. The arguments in lines 1-3 are so compulsory with respect to the other sentences. Furthermore, lines 4 and 5 use the emphatic expression "how" (hereinafter referred to as "emphasized phrase") to prevent other contexts from arising, i.e., other conditions from coming into effect. This suggests that, based on author Nakamura's original argument, contexts (B) and (C) have little basis for existence, or, to put it more bluntly, should not exist.

[0158] In other words, we have confirmed that unit sentence structures have structural forms, but do not necessarily have fixed semantic content. In Nakamura's case, after confirming the structural form, we did not adopt certain semantic interpretations. After confirming the structural form, we can record that we did not proceed with certain contexts (paths) for certain reasons. This will prevent the intrusion of unnecessary interpretations later on. In other words, by clarifying the context, waste is avoided.

[0159] On the other hand, this also shows that adding a new structural form can lead to another valid interpretation. Example 3-2-2 follows the case of "never boasting about having many possessions" in line 12 of Example 3-2-1, and assumes that in that case, "even after death, one will leave property to one's descendants, who will then selflessly support relief efforts."

[0160] [Example sentence 3-2-2 (unit sentence structure)] 01 <People> {will surely die}. 02 <Death> is {natural}, but 03 <Death> is {an unavoidable law for living beings}. 04 if {how proud} of <many possessions>, 05 if {how much fame} we {gain}, 06 <we cannot take it away with us to the afterlife>. 10 <some warmth> {leaves a trace of our lives at that time}, 11 <it is more important} to <convey human warmth and truth>. 07 else (Note: This is a case where {no attempt is made to gain} <fame>) 08 <give it all back to society, anonymizing}. 09 endif 12 else (Note: This is the case where "one does not boast of having many possessions") 12a if <person> {leaves property to descendants} 12b if <descendants> {supports selfless relief efforts} 12c <Mr. Nakamura> {partially approves of such use of property} 12d else (Note: This is the case where <descendants> {do not support selfless relief efforts}) 12e <Mr. Nakamura> {partially approves of such use of property} 12f endif 12g else (Note: This is the case where <person> {does not leave property to descendants}) 13 It may also {be permissible} to <secretly keep all of one's property>. 13a endif 14 endif

[0161] Lines 12a to 12c create a context in which the author, Mr. Nakamura, might allow it. However, even against this context, one could argue that there is no guarantee that descendants will do so. This is the "context" of the counterargument. In this way, we can advance the discussion by using the expression "context" rather than the expression "condition." In this way, we can express it as "The 'context' from lines 12g to 13 is a 'context' that should be reconsidered."

[0162] 1.2.3.3.2. Limitless Context Creation and Consistency of the Overall Meaning The following discussion continues without describing the structural form, as in Example 3-2-2. If an objection is raised regarding the guarantee of inheritance for descendants, then in line 12b, a structural context is created that calls for the social institutionalization of relief activities. If the use of private property after death were to be used in that direction, Nakamura would undoubtedly be in full agreement. The reason for the inference of agreement is that the inventor believes that Nakamura places the relief of people driven into poverty by various social circumstances at the top of his own code of conduct.

[0163] However, placing a claim at the top does not mean that counterarguments will not arise. Rather, the structural form makes it possible to relativize any claim in the opposite direction, or to support the claim in accordance with the top claim. In this example, the opposing relativization might be accepting a moderately ordinary life, or it might be a life attitude that minimizes interaction with others. On the other hand, the latter, supporting the claim, might be, as mentioned above, promoting some kind of social institution to honor Nakamura's achievements and pass on his legacy to future generations.

[0164] In this way, the fact that other contexts can be created for any top-level context comes from the fact that contexts can always create opposite structural forms. However, there is no guarantee that the contexts created from these structural forms will be persuasive when they are filled with meaning. This is because the meaning may be empty or ineffective. Or, even if a context has a valid meaning, if the context is confusing, the meaning may be difficult to understand. Furthermore, when tracing the entire statement in detail, what should have branched off as a context may contradict claims in other parts of the context farther away. Since contexts can formally be added without limit, it is necessary to check the consistency of the meaning of the entire statement.

[0165] 1.2.3.3.3. Organizing confusing contexts Ensuring consistency in overall meaning has a variety of beneficial effects. The effect, simply put, is to eliminate wasted thought and save time. In order to ensure consistency in meaning, it is necessary to clarify the structural form. To illustrate this, we will take up again an example of confusing context. That example is sentence example 1-1. Did clarifying the structural form, that is, organizing the context, actually make the semantic interpretation consistent? A structurally formalized version of sentence example 1-1 is shown below as sentence example 1-2.

[0166] [Example Sentence 1-2 (Unit Sentence Structure)] if <that world> is {a world without color}. However, if <that world> {was previously a world with color}. And, if <that world> {still has words related to color}. In that world, if <that word> is {green}, then: ・The <word green> {refers to abundance}, therefore, ・The expression <colorless green thoughts>> means {thoughts that are abundant in a colorless world}. Furthermore, if <that word> is {thoughts} and if <thoughts> are {used in contrast to "rage"}, then: ・The expression <thoughts are raging"> means {a state in which thoughts are completely undefined}. Also, else if <thoughts> are {used in contrast to "sleeping"}, then: ・The expression <the expression "thoughts are asleep"> means {a state in which thoughts are dead}. Therefore... else -- (Note: Unexamined. Meaning there is room for consideration.) endif endif ・The expression "thoughts sleeping in a rage" means {a dead state with no thoughts at all}. else ・-- endif ・<The expression "colorless green thoughts sleep raging"> means {In a colorless world, rich thoughts are in a completely unsettled state and are as good as dead} (Note: The target phrase is a combination of "colorless green thoughts" and "thoughts sleep raging", and the attribute phrase is a combination of "in a colorless world, rich thoughts" and "thoughts are in a completely unsettled state and are as good as dead". However, "in a dead state" is loosely translated as "as good as dead".) else ・-- endif else ・-- endif else ・-- endif else ・-- endif

[0167] As seen in Example 1-2, by creating a context using structural forms, the nonsensical sentence Example 1, "Colorless green ideas sleep furiously," can be interpreted as "In a colorless world, rich ideas are completely undefined and virtually dead." The new conditional statement ("if") is somewhat forceful. However, this is a kind of passageway created by an artificial context to make the interpreted sentence (given sentence: Example 1) somewhat more understandable. This is similar to the way that by creating a mechanical device, such as a passageway or waterway, a stagnant flow of people can be made to flow through a new passageway, or water can flow through a new channel, even for a meaningless sentence. In other words, the new context can be seen as clearing a complicated and stagnant meaning. It can be said that the unit sentence structure, consisting of the structural form of the context, leads to a certain semantic interpretation. Although this has not been measured over time, it is also an example of how many people agree with the claims of this invention.

[0168] This invention establishes and reinforces the common knowledge that meaning changes depending on the context by focusing on the structural form of a sentence. This allows us to implement, methodize, and program the semantic judgment of an entire document, which is an ordered collection of many sentences, thereby improving the efficiency of document inspection and supporting document creation.

[0169] 1.2.3.4. "Context Fluctuation" by Connective Phrases Incidentally, in interpreting the meaning of Nakamura's previous citation, we saw that even though the format of the else condition is the same as the structural format of if, the else condition does not actually hold. This can be interpreted as meaning that the emphasis phrase "how" does not actually hold other conditions, or other contexts, due to its strength of meaning. In other words, the expression "how" in lines 4 and 5 of Example 3-2-1 is strong, and while it does not completely negate the existence of the conditions in lines 7 and 12, i.e., the existence of the context, it weakens the basis for their existence.

[0170] This phenomenon is termed "context fluctuation" caused by conjunction-related phrases. This occurs because the structural form emphasizes one possible context over another, ignoring its existence. When converting natural sentences into unit sentence structures, it is necessary to pay attention to the conjunction-related phrases that correspond to this context fluctuation, and grasp and understand their function. The inventor feels the need to imagine a so-called "meta-entity" that controls any given context (structure). For example, the meta-entity may assign priorities between contexts. However, the details of the role of this meta-entity are a topic outside the scope of this invention.

[0171] 1.2.3.4.1. Similar Example: Emphasis Phrase "Priority" There is a similar example. Expressions such as "always take priority...", which can also be considered emphasizing phrases, can sometimes make other conditions less visible and prevent them from being established, resulting in a contradiction. An example of this is shown below. This can be said to be the "neutralization" of context that is actually necessary due to the emphasis phrase. The structural format method of the present invention unravels the context structure of the sentence and reveals the cause of this neutralization.

[0172] Example 4 is an example of software requirements. Software requirements are the items that a developer requests to be developed when developing software. The granularity of meaning can be general or detailed, and the method of communication depends on the situation, and can be written or oral. When documented, they are called "requirements specifications," and the requirements within those documents are sometimes called "requirements specifications." However, the distinction is not clear. This example is taken from "Integration of Two Kinds of Syntax for Requirements Description and Its Future Development" (Authors: Norihiro Urushibara; Chiharu Sasaki, Publisher: IEEE, 2018, https: / / ieeexplore.ieee.org / document / 8501291).

[0173] [Example 4 (required specifications)] Requirement 1: When the water level exceeds the alert level (40 cm), the discharge valve will be opened for 5 seconds, and at the same time, no new instructions to open the discharge valve will be accepted for 15 seconds (including the 5 seconds). Requirement 2: The same discharge valve opening and closing operations as above will be performed at the operator's instruction. Requirement 3: The operator's instructions will always take priority. (The instructions will be accepted even during the 15-second prohibited section, and will also take priority over the operator's own past instructions.)

[0174] Create a unit sentence structure from sentence example 4. To do this, create a unit sentence structure for each of the three requirements and combine them into one. This is sentence example 4-1. "-?" indicates that it has not yet been considered what to do in the case of the condition in question.

[0175] Let us explain example sentence 4-1. In example sentence 4-1, unit sentence structures are created in the order of requirements 1 to 3. Requirements 1 and 2 are replaced with unit sentence structures literally in the requirements sentences. Requirement 3 takes into consideration the "priority" case, but this also simply conditions the wording in parentheses ( ) of requirement 3. The last (4), "a simple combination of the above three requirements," is a simple combination of the conditions of the three requirements.

[0176] [Sentence Example 4-1 (Unit Sentence Structure)] (1) Requirement 1 (Unit Sentence Structure): if <water level> {exceeds the alert level (40 cm)} then keep the <discharge valve> {open for 5 seconds} and do not accept <new open instructions> {simultaneously for 15 seconds (inclusive)} else -? endif (2) Requirement 2 (Unit Sentence Structure): if <operator's instructions> {have been given} keep the <discharge valve> {open for 5 seconds} and do not accept <new open instructions> {simultaneously for 15 seconds (inclusive)} else -? endif (3) Requirement 3 (unit sentence structure): if <operator's instruction> {always takes priority} if <new open instruction> {￢will not be accepted simultaneously for 15 seconds (inclusive)} (Note) The requirement analyst interprets this conditional statement (if statement on the second line) as "accept" in accordance with the instruction "always takes priority" in Requirement 3. (Note: Attribute phrases with the negation symbol "￢" are double negated, meaning "accept.") else -? endif else -? endif (4) Combinations of the above three requirements (unit sentence structure): 01 if <water level> {exceeds the alert level (40 cm)} 02 keep the <discharge valve> {open for 5 seconds} 03 give <new open orders> {will not be accepted simultaneously for 15 seconds (inclusive)} 04 if <operator's instructions> {has been given} 05 keep the <discharge valve> {open for 5 seconds} 06 give <new open orders> {will not be accepted simultaneously for 15 seconds (inclusive)} 07 if <operator's instructions> {always take priority} 08 if <new open orders> {will not be accepted simultaneously for 15 seconds (inclusive)}, 09 (Note) The attribute phrase in this conditional statement (if statement) on line 8 contradicts the attribute phrase two lines above (line 6). 10. Therefore, the response in the case of "always takes priority" is inappropriate. 11 else 12 endif 13 else 14 endif 15 else 16 endif 17 else 18 endif

[0177] The unit sentence structure of (4) in Example 4-1 has been assigned line numbers. Also, the "-?" following the else line has been omitted. In (4), lines 6 and 8 are contradictory. Both have the same target phrase, but the attribute phrases are contradictory. The attribute phrase {Do not accept simultaneously for 15 seconds (5 seconds inclusive)}" and the negation symbol {￢Do not accept simultaneously for 15 seconds (5 seconds inclusive)}" are contradictory. Looking at the sentence sequence, or context, if line 6 is sentence P, then line 8 is ￢P, and in the same context, P and ￢P are arranged in a conjunction relationship, "... , , ∧P∧, ∧￢P⇒... , , ."

[0178] To conclude, the reason for the contradiction is that Requirement 3 in Example 4 (Requirement Specifications) states "always take priority," but the scope of this statement was unclear. From the perspective of this invention, which focuses on the scope of context, it was not clear in the requirements specifications how far the priority extended to. The requirements analyst simply prioritized "always take priority." This is the cause.

[0179] The requirements analyst received Requirement 3 of the original requirements specification and expressed the results of the analysis in Requirement 3 of the unit sentence structure. Then, in the second line of Requirement 3 of the unit sentence structure, it was stated that any new development instructions would be accepted because the first line of Requirement 3 was a case of "always taking priority." The notation uses the negation symbol "￢" to indicate that it is not impossible to "accept." If Requirement 3 of the original requirements specification emphasized the expression of priority, saying "this instruction takes priority even over its own previous instructions," it would be inevitable that the requirements analyst would create a unit sentence structure literally without dividing the situation according to the conditions.

[0180] In this example, if one is misled by the meaning of a word and a condition creates a new context, one may fail to notice the new context, thereby violating a semantic domain, such as a contradiction. In this example, P and ￢P are placed in the same context: "... , , ∧P, ∧ ￢P⇒...". In this context, there is a "⇒", and it can be seen that "￢P" is the antecedent of implication. The antecedent indicates the start of a new context. There is a contradiction in the antecedent. Therefore, the antecedent of the contradiction indicates that a new, non-contradictory context should be created. By creating a new, non-contradictory antecedent, a non-contradictory semantic domain should be created.

[0181] In reality, the operator's instructions that should be given priority differ between earlier and later instructions. Therefore, it was necessary to distinguish between giving priority to the earlier or later instructions. However, it was broadly assumed that it always takes priority over any other instructions. There was still a detailed context. In this case, the antecedent "shape" goes beyond meaning, indicating a formal contradiction, and therefore indicating that the meaning should be reconsidered.

[0182] 1.2.3.4.2. Candidate expressions for connecting phrases The expressions "always take precedence" or simply "take precedence" are not expressions of conditional phrases. However, this example shows that expressions such as these that encourage a choice of priority should also be considered expressions that create new conditions. This prevents the risk of contradictions that may occur if the phrases are interpreted literally. A formal method that makes it relatively easy to set conditions contributes to creating contexts that are free of semantic contradictions.

[0183] In addition to emphasizing expressions, there are also expressions that function as conditional relative phrases despite being non-conditional. In the previous sentence example 2, there is "...but, only on festival days...," where the conditional relative phrase is not explicit, like an expression like "if...then." It is conceivable that there are other expressions that should be made into conditional relative phrases despite being non-conditional. Such expressions are named "candidate conditional relative phrase expressions." By collecting these expressions, it is possible to improve the accuracy of paraphrasing them into conditional relative phrases. As will be explained later in the explanation of computer processing, a database of candidate conditional relative phrase expressions can be created and applied to computer processing.

[0184] Conjunctions include not only conditions that express implication, but also conjunctions, negations, and disjunctions. Therefore, expressions that indicate these conjunctions must also be collected as candidate expressions.

[0185] 2. Example of a "Context-Based Document Inspection Device" As seen in the example sentences above, the present invention is characterized by capturing the characteristic that the meaning of words and sentences can change depending on the context. Based on this characteristic, a device, method, program, or system can be created and used, for example, to inspect an entire document. The feature of the present invention can be called a context-based, or "context-based" idea. For example, when this idea is applied to document inspection, it can be called a "context-based document inspection device." Furthermore, when it is applied to idea generation by combining various contexts rather than inspection, it can be called a "context-based idea generation system." Below, we will look at an example of document inspection.

[0186] As shown in Figure 1, the structure of the "context-based document inspection device" (000) consists of a document reading unit (100), a document map creation unit (200), and a document inspection unit (300). The document map creation unit (200) assigns addresses to the sentences (text data) read by the document reading unit (100). This is to identify the location of each individual sentence. The addresses are useful for distinguishing the location of a sentence when referencing it during inspection, etc. The document map creation unit is named "map" as an analogy to this addressing.

[0187] 2.1 Document Map Creation Unit (200) As shown in FIG. 2, the document map creation unit (200) consists of a unit sentence construction unit (210) and a unit sentence structure construction unit (220).

[0188] 2.1.1. Unit Sentence Construction Unit (210) In the unit sentence construction unit (210), the language analysis unit (211) performs language analysis on the sentence using, for example, morphological analysis and dependency analysis, the unit sentence element extraction unit (212) extracts target phrases and attribute phrases from the simple sentence to create unit sentences consisting of the target phrases and attribute phrases, the conjunction relation phrase extraction unit (213) extracts conjunction relation phrases that connect the sentences, and then the unit sentence relation construction unit (214) connects the unit sentences with conjunction relation symbols according to the types of conjunction relation phrases. The connected unit sentences are the "unit sentence relation bodies" named above.

[0189] 2.1.1.1. Language Analysis Unit (211) The language analysis unit (211) performs language analysis. For language analysis, a method of morpheme and dependency analysis of a sentence is used, as shown in FIG. 5. In FIG. 5, the following sentence example 5 is used as the input sentence (text data). In analysis example 5 (analysis of sentence example 5) in FIG. 5, the input sentence is subjected to morpheme analysis, and grammatical attributes (for example, part of speech names such as conjunctions and nouns) are identified for each analyzed morpheme. Then, the order of appearance of the morphemes (zero-origin) and dependency relationships such as the order in which this morpheme is related to each morpheme are indicated, such as "0 1D...".

[0190] Existing language analysis tools for these morphemes and dependencies can be used.

[0191] The following is a quote from the book, with some minor revisions.

[0192] [Example 5] If the error in this vertical position is large, when a car enters a curved road from a straight road, even if the car's speed is too high, if the speed control is appropriate, the car can successfully turn and automatically stay within the driving lane on the curved road.

[0193] 2.1.1.2. Unit Sentence Element Extraction Unit (212) The unit sentence element extraction unit (212) extracts simple sentences and unit sentences. In simple sentence extraction, a simple sentence is enclosed in " " as shown in sentence example 5-1. Here, a simple sentence is a sentence consisting of a subject and a predicate in normal grammar. A simple sentence is identified by referring to the data in analysis example 5 in Figure 5.

[0194] [Example sentence 5-1] Even if "the error in this vertical position is large" and "the vehicle enters a curved road from a straight road" and "the vehicle speed is too high," if "speed control is appropriate," "the vehicle can make the turn successfully" and "can automatically drive within the driving lane on the curved road."

[0195] A unit sentence is constructed by extracting target phrases and attribute phrases from a simple sentence. This is also done by referring to the data in Analysis Example 5 in Figure 5. Example sentence 5-2 is an example of this. (The target phrases and other information have already been discussed.)

[0196] [Example sentence 5-2] If <this vertical position error> is {large}, when {enters a curved road from a straight road}, even if <the vehicle speed> is {too high}, if <speed control> is {appropriate}, the {can successfully complete the turn} and <can automatically stay within the driving lane on the curved road>.

[0197] 2.1.1.3. Conjunction Related Word Extraction Unit (213) The conjunction related word extraction unit (213) extracts conjunction related words that link unit sentences. Example sentence 5-2 contains words that connect sentences, such as "if," "even when," and "if...then." These are conjunction related words. In this example, since it is a condition, it can be said to be a condition related word among conjunction related words. Conjunction related words such as "and," and "and" are also extracted from conjunction related words. Example sentence 5-3 below is the result.

[0198] [Example sentence 5-3] If <this vertical position error> is {large}, and if {enters a curved road from a straight road}, even if <the vehicle speed> is {too high}, if <speed control> is {appropriate}, then will {make the turn successfully} and <will be able to automatically} <drive within the driving lane on the curved road>.

[0199] Example 5-3 uses one line for each sentence, and shifts each conditional phrase one place to the right. As a result of this shift in places, the sentence ends as if there is still half left.

[0200] 2.1.1.4. Unit Sentence Relation Constructor (214) Therefore, the unit sentence relation constructor (214) fills in the half. To fill in the half, the opposite of "if" is used, "if not," and the phrase "end" is used to indicate the end of the set of "if" and "if not." Also, a line number is inserted on the left side. The result is Example Sentence 5-4.

[0201] [Sentence example 5-4] 01 If <this error in vertical position> is {large}, 02 if <the automobile> {enters a curved road from a straight road}, 03 even if <the vehicle speed> is {too large}, 04 if <speed control> is {appropriate}, 05 then the <automobile> {can successfully complete the turn}, 06 and <can automatically stay within the driving lane on the curved road> 07 otherwise 08 the end 09 otherwise 10 the end 11 otherwise 12 the end 13 otherwise 14 the end

[0202] Example sentence 5-4 satisfies the remaining half of the sentence. However, this is only in form, and there are no sentences with content from line 07 onwards. A sentence in this state, that is, a sentence in which there is no content-effective sentence (unit sentence) in the line but the sentence conjunction is satisfied, is the unit sentence relation body mentioned above. The unit sentences are connected by conjunction symbols.

[0203] 2.1.2. Unit Sentence Structure Construction Unit (220) Therefore, the unit sentence structure construction unit (220) inserts valid unit sentences between the lines of the unit sentence relation body. However, this is a manual process. The result of this process is shown first in Example 5-5. In every case, a sentence (unit sentence) is inserted on the line following the opposite conditional expression "if not."

[0204] [Example sentence 5-5] 01 If <this error in vertical position> is {large}, 02 if <the automobile> {enters a curved road from a straight road}, 03 even if <the vehicle speed> is {too high}, 04 if <speed control> is {appropriate}, 05 then <the automobile> {can complete the turn}, 06 and <drive in the driving lane on the curved road> {can automatically} 07 otherwise 08 then <the automobile> {can complete the turn successfully}, 09 and <drive in the driving lane on the curved road> {can automatically} 10 the end 11 otherwise 12 then <the automobile> {can complete the turn}, 13 and <drive in the driving lane on the curved road> {can automatically} 14 the end 15 otherwise 16 then <drive in the driving lane on the straight road> {can automatically} 17 End 18 Otherwise 19 If the <automobile> {enters a curved road from a straight road}, 20 even if the <vehicle speed> is {too high}, 21 if the <speed control> is {appropriate}, 22 then the <automobile> {can complete the turn}, 23 and {can automatically} <keep in the driving lane on the curved road} 24 Otherwise 25 then the <automobile> {can complete the turn successfully}, 26 and {can automatically} <keep in the driving lane on the curved road} 27 End 28 Otherwise 29 then the <automobile> {can complete the turn successfully}, 30 and {can automatically} <keep in the driving lane on the curved road} 31 End 32 Otherwise 33 then {can automatically} <keep in the driving lane on the straight road} 34 End 35 End

[0205] 2.1.2.1. Unit Sentence Addressing Unit (221) In the unit sentence structure constructing unit (220), the unit sentence addressing unit (221) assigns line numbers to unit sentences, as seen in Example 5-5 above. This is to uniquely identify unit sentences. However, other methods may be used as long as they can uniquely identify unit sentences.

[0206] The line number of a unit sentence is associated with the sentence number of the original sentence (the number indicating the order in which the sentence appears). The latter is called the "original sentence number," and the unit sentence number and the original sentence number are associated. This is because it allows you to identify where the unit sentence is located in the original sentence.

[0207] For example, if a computer reads the text of an original sentence, the first "original sentence number" of the original sentence is from the beginning of the sentence to the first period "." or ".". Of course, a sentence can also end with a space. Eventually, it will be possible to determine the symbol for the end of a sentence and count the number of that sentence. It is also possible to separately extract simple sentences from the original sentence and use the order in which those simple sentences appear as original sentence numbers.

[0208] 2.1.2.2. Unit Sentence Structure Editing Unit (222) The unit sentence structure editing unit (222) allows the user to modify unit sentence relations. Previously, in sentence examples 3-1 and 3-2, the same process was carried out as adding words to the target phrases of the unit sentences and adding words to the else clauses. The same process is carried out in the unit sentence structure editing unit (222).

[0209] The result of this process is the unit sentence structure of Example 5-5. The unit sentence structure has the same formal structure as the unit sentence relation structure. The difference is that it has been manually edited. In particular, the difference is significant in whether or not a sentence has been added to a blank line. Also, I wonder if there is a way to automate the process of adding words. By applying the "surrounding context" method mentioned above, it may be possible to automatically select words. In this case, the surrounding context plays a complementary role to the global context claimed by this invention. However, this complementarity is not the point claimed by this invention.

[0210] The present invention has the effect of clearly indicating the position (line) where a sentence should be inserted. Waiting for the insertion of a sentence, using Example Sentence 3-1, the insertion location is indicated as a blank line. Example Sentence 3-2 shows the result of the insertion. However, the insertion must be a sentence that the person reading Example Sentence 3 creates in accordance with the intent of the sentence. In other words, the blank line structural form is filled with a sentence (unit sentence) that combines words with appropriate meanings. Furthermore, Example Sentences 3-2-1 and 3-2-2 consider the meaning further, adding conditions to refine the meaning.

[0211] The mechanism that supports this refinement of meaning is called structural form, and is the mechanism provided by this invention. An example is the "unit sentence." A unit sentence is one of the structural forms that make up a context. A unit sentence has two elements: a target phrase and an attribute phrase. Our thoughts function differently depending on which of these two a given phrase belongs to. When the word "Mount Fuji" appears as the target phrase "<Mount Fuji> {}" or when it appears as "<> {Mount Fuji}," the direction of consciousness is different. When it appears as "<Mount Fuji> {}," attention is drawn to the attributes of Mount Fuji, wondering what has happened to it. It could be the first snowfall, or a sign of an impending eruption. On the other hand, when it appears as "<> {Mount Fuji}," consciousness is moved to search for a tall object, a person with outstanding abilities, or some other object, such as "<He> is {Mount Fuji}."

[0212] This is a mechanism for considering meaning. Furthermore, the present invention also devise "connective phrases" for the structural form that connects unit sentences, and uses these relationships as elements to connect the relationships between sentences and words in the entire natural language as contextual relationships.

[0213] These are mechanisms that can be thought of as refining the meaning of words and relating the meanings of sentences. A mechanism can be considered to be a device, method, or program that organizes the context of a sentence into a structural form.

[0214] 2.1.2.3. Unit Context Extraction Unit (223) The unit context extraction unit (223) extracts unit contexts from unit sentence structures. When a unit sentence structure is traced continuously from its beginning to its end, the traced unit sentences become individual global contexts.

[0215] When the unit sentence structure and the original document are equivalent, any individual unit context extracted here can be considered to be the individual global context of the original document. A note on "equivalent": When editing a unit sentence structure, the original text often does not contain a sentence equivalent to the else (if not) statement. For example, the original text does not contain a sentence equivalent to the "<>{}" line in Sentence Example 3-1 or the "-?" line in Sentence Example 4-1. In such cases, when a target phrase and attribute phrase are filled in the brackets (<>{}) to create a new unit sentence, adding a natural sentence with the same content to the original sentence makes the unit sentence structure equivalent to the original sentence.

[0216] An example of an equivalent sentence is Sentence Example 3-2. Sentence Example 3-2 is the original sentence, Sentence Example 3, to which a unit sentence has been added. However, this example was arbitrarily created by the inventor. It differs from the original sentence of the original author. In such cases, natural sentences are created based on Sentence Example 3-2, and a new "original sentence" is created.

[0217] Sentence example 3-5 is an example of a unit context extracted from a unit sentence structure. Sentence example 3-5 shows an individual global context that spans the entire region of the sentence from the beginning to the end. Sentence example 3, which is the source of sentence example 3-5, is a short sentence. However, the same idea can be applied to an entire document consisting of many sentences. The extraction of such global context is performed by the unit context extraction unit (223). It should be noted that the test is not performed on symbol sentences, but on unit sentences that contain target phrases and attribute phrases. This is because the test ultimately asks about meaning.

[0218] 2.2. Document Inspection Unit (300) The document inspection unit (300) inspects the document. The inspection is performed on the unit sentence structure. The inspection assumes that the original sentence is equivalent to the unit sentence structure. The inspection is performed for each unit context. The present invention is based on a so-called "rule" regarding words and sentences that there should be no contradictions in the same context. According to this rule, for example, words and phrases consisting of the same characters must be used with the same meaning in the same context. And if words and phrases consisting of the same characters are used with different meanings, these words and phrases should not exist in the same context. In other words, "polysemous words" that are the same but have different meanings are OK as long as they appear in different contexts.

[0219] This usage of "context" seems normal. However, the meaning of "context" is broad. For example, it is used in expressions such as "In this context, it is considered..." or "In another context, it is interpreted as...". In this case, "context" evokes a kind of "scene" in the listener's mind. It is similar to the development of a story in a play or novel. One event in a play may be described as joy in one scene, while the same event may be told as tragedy in another scene. Or it may be about sin and punishment, or love and hate. These scenes are further divided into smaller "scenes," and the story unfolds with various "folds."

[0220] At first glance, the "context" of this invention seems different from the broad meaning of these contexts. However, if the protagonist first says "A" in a certain scene and then says "Not-A," we may react to the contradiction by asking "why" or "that's strange," or we may understand and accept the protagonist's intentions.

[0221] The "context" in this invention focuses on the commonality of these structures and applies the same "structural form" to them. In this way, situations become contexts, and the meaning that flows there is formalized in unit sentences, etc., and then the meaning is examined.

[0222] From the above, it can be seen that in this invention, document inspection is characterized by being based on context. The document inspection unit (300) consists of an inspection selection unit (310) and a context inspection unit (320). The inspection selection unit (310) selects the area and inspection topic of the document inspection. The context inspection unit (320) inspects the document based on its context.

[0223] 2.2.1. Inspection Selection Unit (310) The inspection selection unit (310) consists of an inspection area selection unit (311) and an inspection subject selection unit (312). The inspection area selection unit (311) selects a chapter or section of the document to be inspected, or an arbitrary context, as the inspection area. The inspection subject selection unit (312) selects an inspection subject to check, such as whether words consisting of the same characters are used with the same meaning, or whether there are any contradictions between sentences.

[0224] 2.2.1.1. Inspection Area Selection Unit (311) The inspection area selection unit (311) selects which area of ​​the document to inspect. Figure 7 shows the general structure of a document. Following the title are the items of preface, chapters, sections, paragraphs, and afterword. The items are followed by text. Figure 8 shows an example table of contents. The table of contents in this example only lists chapters and sections, but the main text contains headings and text for items equivalent to paragraphs. The main text also contains items such as paragraphs, regardless of whether they are listed in the table of contents. Items below paragraphs are also possible. As can be seen above, items such as chapters and sections have a hierarchy, or it can be said that the table of contents has items arranged hierarchically. The numbers after the headings of the items are page numbers.

[0225] The table of contents in Figure 8 is taken from "Computing: Principles and Their Development" (by Masami Hagiya, The Open University of Japan Foundation, 2019).

[0226] In the case of inspection, you can select individual items to inspect from the table of contents, or you can select the entire document rather than selecting individual items.

[0227] The concept of the area when selecting the entire document is as follows. In Figure 9, the sentences in Figure 7 are represented by the "Δ" (delta) symbol, and each sentence (Δ) is assigned a unique number for each chapter, section, and paragraph, indicating that each sentence is unique. The Δ in "Preface" and "Afterword" are represented as "Δp" and "Δa," respectively, indicating that they are unique.

[0228] In this case, the sentence does not refer to natural sentences, but to natural sentences converted into unit sentence structures. Therefore, Δ constitutes a unit sentence structure. Also, Δ refers to sentences for each item. Here, an item refers to a chapter, section, paragraph, etc. It does not refer to only a paragraph. Therefore, a unit sentence structure for each item is called an "item-specific unit sentence structure."

[0229] Example sentence 6 below explains the relationship between the sentences in Figure 9.

[0230] [Example 6 of sentences (unit sentence structure of the entire document)] A. (Δp)∧(Δ1)∧(Δ11)∧Δ111 B. Same as above ∧Δ112 C. Same as above ∧Δ113 D. (Δp)∧(Δ1)∧(Δ12)∧Δ121 E. Same as above ∧Δ122 F. (Δp)∧(Δ1)∧(Δ13) G. (Δp)∧(Δ2)∧(Δ21)∧Δ211 H. Same as above ∧Δ212 I. Same as above ∧Δ213 J. (Δp)∧(Δ2)∧(Δ22) K. (Δp)∧(Δ3)∧(Δ31)∧Δ311 L. Same as above ∧Δ312 M. (Δp)∧(Δ3)∧(Δ32)∧Δ321 N. Same as above ∧Δ322 O. Same as above ∧Δ323 P. A∧B∧C∧、、、、∧N∧O∧Δa

[0231] The explanation is as follows: A to O in Example 6 are identification symbols for each line. P is a conjunction of A to O. P will be explained using Example 6-1 below.

[0232] Each row from A to O indicates that the sentence in the highest level item is related to the sentence in the lowest level item. The lowest level item refers to the last item in a chapter, section, or paragraph in Figure 9. For example, this indicates that the sentence in the lowest level item, paragraph, is related to the sentence in the section that is higher up in the paragraph, and the sentence in the paragraph is related to the sentence in the chapter that is higher up in the paragraph. Dependencies can be applied using conjunctions in logical inference. Therefore, they are represented by the conjunction relation symbol "∧".

[0233] In reality, it is rare for the "Preface" (Δp) to be written with a meaning that binds the lower-level sentences. However, there are cases where the "Preface" is the conclusion of the entire document. Example 6 assumes such a case. In each of lines A to P, after the "Preface," the higher-level item applies to all lower-level items.

[0234] Furthermore, there are cases where the "Afterword" is the conclusion of the entire document. In this case, as shown in the line with line symbol P, the conjunction of the higher-level line symbols A to O and the sentence (Δa) of the "Afterword" are connected by a conjunction. Sentence example 6 shows the above-mentioned relationships of the entire document.

[0235] When sentences overlap in conjunctions, all but one of the identical sentences can be removed. As with sentences, duplicate unit sentence structures can be removed. After removing the duplicate unit sentence structures, the final line symbol P becomes Example Sentence 6-1.

[0236] [Example 6-1] P. (Δp)∧(Δ1)∧(Δ11)∧Δ111∧Δ112∧Δ113∧(Δ12)∧Δ121∧Δ122(Δ13)∧(Δ2)∧(Δ21)∧Δ 211∧Δ212∧Δ213∧(Δ22)∧(Δ3)∧(Δ31)∧Δ311∧Δ312∧(Δ32)∧Δ321∧Δ322∧Δ323∧Δa

[0237] Sentence Example 6-1 shows that when inspecting an entire document, all sentences in the document should be inspected in the order in which they are written. For example, it shows that it is not necessary to inspect the same thing on a chapter-by-chapter basis. In Sentence Example 6, the same thing is the sentence written as "same as above" (unit sentence structure).

[0238] On the other hand, if you want to check any chapter, section, or paragraph, you can simply specify the item in question. The check will be performed by tracing the conjunction relationships of the specified item.

[0239] The object of inspection is the unit sentence structure. As mentioned earlier, there is not a one-to-one relationship between the original sentence and the unit sentence structure. There are sentences that do not exist in the original sentence but exist in the unit sentence structure. For example, in the previous sentence example 5-5, there is a unit sentence that does not exist in the original sentence, sentence example 5.

[0240] Example sentence 5 is a sentence consisting of only one period "." In other words, if everything up to the period is considered a sentence, then example sentence 5 has one sentence. Therefore, we will compare the original sentence (example sentence 5) with a sentence that has been restructured into simple sentences. Example sentence 5-1 is the sentence that has been restructured into simple sentences. Example sentence 5-1 has six simple sentences. In the unit sentence relation body of example sentence 5-4, the number of simple sentences and the number of unit sentences are the same. However, the unit sentence structure obtained by revising the unit sentence relation body will have a larger number of sentences. The sentences from the seventh line onwards in example sentence 5-5 do not exist in example sentence 5-1.

[0241] The test results are given to the original text. The results of the unit sentence structure test must be reported, including any mention of the unit sentence structures that do not exist in the original text. Based on this report, it is possible to then modify the original text to make it consistent with, or equivalent to, the unit sentence structures.

[0242] Returning to the selection of test areas, testing is primarily about semantics. Therefore, if you know that the chapters in a given document are semantically independent of each other, you don't need to test the entire document. In this case, you can test each chapter individually.

[0243] There is also the idea that items such as chapters, sections, and paragraphs are each considered to be independent contexts. Example 6 above does not recognize the independence of items such as chapters, sections, and paragraphs as contexts when treating the document as a whole. When a context, i.e., a conditional relation phrase is set there, the conditional relation phrase for the item must be organized to clearly show how it relates to other items. Each item in Example 6 is no longer a simple conjunction. However, if you know the semantic relationship between the items, you can create a path. However, even if you do not know the semantic relationship between the items, if the test results reveal the contradiction between the items as a structural form, you can see that the meanings of the items are also inconsistent with each other. From this point of view, the test in Example 6-1 above, which does not consider items to be independent contexts, is sufficient.

[0244] 2.2.1.2. Test Subject Selection Section (312) The test subject selection section (312) is where test subjects are selected. Test subjects are, so to speak, the test menu. There is no doubt that tests test meaning. However, when we talk about "meaning" in a test, we are asking whether the "meaning" of a word or phrase is correct or appropriate. Also, in this case, rather than asking whether the "meaning" is correct or appropriate, it may be better to ask whether the "expression" is correct or appropriate. "Meaning" is diverse.

[0245] Also, when we talk about the "meaning" of a sentence, for example, if we take a contradiction as meaning, there is a formal contradiction and an empirical contradiction. A "formal contradiction" refers to a contradiction in form, without even considering the meaning. In other words, when comparing two unit sentences, if the target phrases are the same between the two sentences and the attribute phrase A is in an inverse relationship, i.e., {A} and {￢A}, then there is an immediate contradiction formally, that is, in form, without even considering the meaning of the target phrase or attribute phrase.

[0246] On the other hand, an empirical contradiction occurs when two things have different attribute words that appear to be compatible in meaning, but are incompatible in fact. If heavy rain falls and that heavy rain causes the river to recede, it does not make sense. An explanation is needed. The two things are contradictory in the world of everyday experience. Such an empirical contradiction is called an experiential contradiction.

[0247] However, it is possible to explain the relationship between heavy rain and a decrease in water level without contradiction. It is also possible to say that while there was heavy rain in the mountains, downstream, river construction and, for example, automatic control of the river's floodgates were successful, resulting in a sense of a decrease in water level. This explanation is based on a difference in the situation or scene. From the perspective of linguistic expression, it can be said to be an explanation based on a different context. This was mentioned earlier in the context of various scenes in a play, etc. It is also similar to Example 1-1 (Chomsky's example).

[0248] 2.2.2 Context Checking Unit (320) The context checking unit (320) checks the context. Before reaching the context checking unit, processing is performed to align the context.

[0249] The reason why context is the subject of inspection is at the heart of this invention. The background to this invention is that even words consisting of the same characters can have different meanings depending on the context, and in such cases, there are limitations to comparing a limited range of words as vectors (see "surrounding context"). Therefore, in order to inspect the meaning of words or sentences, inspection must cover the entire document. This is the main thrust of this invention. Based on this thrust, we have developed an apparatus, method, and program for inspecting documents.

[0250] The context checker (320) has a check execution unit (321) and a check support database (DB) set (322).

[0251] 2.2.2.1 Inspection execution unit (321) The inspection execution unit (321) executes an inspection according to the selected inspection item. An example of an inspection item is shown in the table in Figure 10. In addition, when executing an inspection, the inspection support database set (322) is referenced.

[0252] 2.2.2.2. Inspection Support Database (DB) Set (322) The inspection support database (DB) set (322) is a DB related to words, phrases, and sentences. The DB plays a role in supporting the actual inspection. Details of the DB will be described in the next section, "Example of Document Inspection."

[0253] 3. Example of Document Inspection An example of document creation rules will be taken and compared with the idea of ​​the present invention. This will be used as an example of the present invention.

[0254] 3.1. Example of Document Creation Rules The table in Figure 10 shows an example of compliance requirements when writing software development instruction documents. Software development instruction documents are generally called "requirements specifications." Although they are documents for a specific field, they share many commonalities with documents that require strictness. Therefore, these compliance requirements are considered as a test menu (subjects), and the proposal of this invention is verified to see if it can withstand this test menu. This is also used to verify the feasibility of the context inspection unit (320).

[0255] The table has a category called "Sentence Attributes" in the first row, and the attribute is natural language. The present invention also targets natural language. However, the natural language in this case is Japanese.

[0256] The second line of the table, "Relationship between sentences," shows that ambiguity can be either "between two sentences" or "within one sentence." The third line shows the category of "phrase or sentence." The second line above, "Ambiguous between two sentences," categorizes it as "depending on phrase" or "depending on sentence." The second line above also shows that "Ambiguous in one sentence" is "depending on phrase."

[0257] These ambiguities can be categorized as follows. The tasks are given a task symbol "Q" and are indicated by the Q number and the task name. [Classification of ambiguity] (1) Between two sentences (1.1) Ambiguity due to words and phrases Q1: "Unique word" Q2: "Formal contradiction" Q3: "Experiential contradiction" (1.2) Ambiguity due to sentences Q4: "Conditional coverage" (1.2.1) Ambiguity due to the relationship between the antecedent and consequent of a sentence Q5: "Antecedent contradiction" Q6: "Consequent contradiction" Q7: "Pay attention to both" Q8: "Pay attention to the order of events" (2) Within one sentence (Ambiguous) (2.1) Ambiguity due to words and phrases Q9: "Domain-specific word" Q10: "Boundary word"

[0258] 3.1.1. Ambiguity Between Two Sentences Ambiguity between two sentences can be either lexical or sentence-specific.

[0259] 3.1.1.1. Phrasal Ambiguity When ambiguity occurs between two sentences, it can be caused by phrasing.

[0260] Q1: "Unique Word" The title of the problem with problem number Q1, "Unique Word," is that the "cause" of the ambiguity written in the table in Figure 10 is "homonyms" or "heteronyms." This refers to the situation where the same word or phrase is used with different meanings in two sentences, or where different words or phrases are used with the same meaning in two sentences. As a result, the expressions in the two sentences become ambiguous. The "solution" solution provided by the table is to adhere to the rule that the same word or phrase must be used with the same meaning within the same document. In other words, if homonyms or heteronyms are used in two sentences, it will be confusing and lead to a misunderstanding of the meaning.

[0261] However, simply comparing two sentences does not address the fact that words and phrases change meaning depending on the context, so in Q1 of the table, we must accept our proposal to compare sentences depending on the context.

[0262] There are some words that must not change even when the context changes. If domain-specific words such as technical terms are used with different meanings in different contexts, it can cause confusion. The points that must be observed regarding domain-specific words are listed in Q9 of the table.

[0263] The point raised in Q1 is about general words and phrases. It is restrictive for writers to have to use the same expression even when the context is different. The proposal of this invention flexibly responds to such writer's reality. It is superior to the solution in Q1.

[0264] Q2: "Formal contradiction" A "formal contradiction" occurs between two sentences. However, this also occurs in the same context. The solution to Q2 in the table must also be revised to say that it is not simply two sentences, but two sentences in the same context. This invention is clearer.

[0265] Q3: "Experiential Contradiction" An "experiential contradiction" indicates that both events do not exist as empirical facts. The cause of this is that the same object has multiple attributes. The example of the sun in Example 3 above is an example of this. In the example, the sun {appeared in the morning} and {danced and sang all night long}.

[0266] The solution to the table in Figure 10 is to confirm that the attributes differ depending on the perspective. This seems to indicate that changing the perspective will resolve the contradiction between the attributes "appearing in the morning" and "dancing and singing all night long." "If you change your perspective" is a common phrase we use when pointing out ambiguity and reconciling meanings. However, here too, the solution is not simply to change the perspective, but to point out that the two sentences should be in different contexts or that the attribute phrases should be changed. This is because a contradiction would arise if phrases referring to incompatible attributes existed in the same context.

[0267] 3.1.1.2. Sentence Ambiguity Sometimes the ambiguity between two sentences is due to the sentence rather than the words.

[0268] Q4: "Condition coverage" "Condition coverage" is a concern when writing conditions, as there is a risk of missing a condition. Therefore, the solution in the table of Figure 10 is to ask for the opposite condition to ensure that no condition is missed.

[0269] However, there are expressions that make it easy to notice the conditionality and expressions that do not. For example, if the instruction is to group the data into one bundle if there are more than 100 items, and the instruction is to group the data into bundles of 100 items, the former uses the expression "if...", making it easy to notice the conditionality. Therefore, it is also easy to write the countercondition. On the other hand, in the latter, the condition is not directly written in the words, so you have to infer the content and extract the conditionality from it.

[0270] These differences in expression styles seem to depend on what might be called "business culture." If there is a norm that requires explicit expression of conditional expressions such as "if" or "then," it would be in the case of software development work, which requires a table like the one in Figure 10. In software development, various conditions are intertwined, so there may be a norm that requires explicit expression of conditionality. On the other hand, in documents other than this type of business, the expression "Collect the data into bundles of 100 items" is more concise and would be highly evaluated in administrative departments, for example.

[0271] The above has advanced the concern about the comprehensiveness of conditions into a discussion of whether the expressions make it easy to generate counterconditions. Returning to the original point, the point is to examine not only the stated conditions but also the counterconditions. It would be ideal if a computer or other device had the functionality to automatically generate counterconditions. In this invention, as we have already seen, countercondition symbols are utilized to first generate the countercondition as a structural form. The actual meaning of whether the countercondition is necessary or not is examined in accordance with the structural form. In the previous sentence examples 3-1 and 3-2, the meaning is examined in accordance with the structural form.

[0272] Returning to the original discussion, the question of whether an expression is likely to produce an opposite condition is a question of whether the expression is conditional. In the expression "Group the data in groups of 100," where the conditionality is hard to notice, the conditionality lies in the "group each...group..." part.

[0273] It was mentioned earlier that there are cases where emphatic expressions require a change to a condition or reconsideration. In this case, the expression has a hidden condition. I will now touch on something related to this discussion: the examination support database set (322). This examination support database set (DB) will be mentioned later, but within this DB there is a "D1: Conjunction relation phrase candidate expression DB." D1 refers to emphatic expressions related to conditions, the hidden conditional expression in this case, and others as conjunctive relation phrase candidate expressions, and is used to support the structural formalization of contextual structures such as conditions.

[0274] 3.1.1.2.1. Ambiguity due to the relationship between the antecedent and consequent of a sentence Ambiguity between two sentences can sometimes be caused by the relationship between the antecedent and consequent of the sentence itself.

[0275] Q5: "Antecedent Contradiction" According to the "Cause" section of the table in Figure 10, an "Antecedent Contradiction" refers to a case where there is a contradiction in the antecedent of a conditional sentence. And the "Solution" section says that a contradiction in the antecedent should be corrected, whether the antecedent is a formal contradiction or an experiential contradiction.

[0276] The semantic rules for logical inference state that a sentence whose antecedent is false is true regardless of whether the consequent is true or false. However, the table states that contradictions in the antecedent are to be corrected, so it is not permitted that there are contradictions in the antecedent in the document. The idea behind this invention is the same.

[0277] Since the antecedent of an implication is always true, the implication "A⇒B" between antecedent A and consequent B is equivalent to "A∧B". This is called the "conditional equivalence of implication and conjunction" method. It is also called the "CEIC" method for short. CEIC stands for "Conditional Equivalence of Implicature and Conjunction".

[0278] The CEIC method provides simplicity when tracing the path of a sentence, such as when two implications are connected by a conjunction. For example, "(A ⇒ C) ∧ (A1 ⇒ C1)" is equivalent to "A ∧ C ∧ A1 ∧ C1." Of course, this is assuming the CEIC method. Alternatively, it can be likened to a simplification "device" that sifts through complexity. Solving the logical formula "(A ⇒ C) ∧ (A1 ⇒ C1)" results in complex combinations of antecedents and consequents. However, since humans typically write sentences under the assumption that both the antecedent and consequent are true, the path of the sentence, or the logic, should be "straightforward" and "simple." The CEIC method, which assumes the antecedent is true, is consistent with the assumptions used when writing documents. This is why we adopt this method.

[0279] When conditions overlap in natural language, that is, when implication sentences are complicated in natural language, it is somewhat difficult to find contradictions in the natural language. However, it is easy if the natural language is a unit sentence structure. Example 5, which has overlapping conditions, is a natural language. Example 5-4 is a unit sentence relational body. Here, the sequence of conditions is clear. Example 5-5 is a unit sentence structure, and the sentence can be edited line by line. The contradiction in the antecedent can be corrected here.

[0280] Example sentence 5-5 will be used to illustrate a form in which the antecedent becomes a formal contradiction. If the negation of the sentence in line 02 of Example sentence 5-5 is inserted into line 03, the entire antecedent becomes a contradiction. This will be Example sentence 5-6. Let the inserted line be 02α. Since "￢" negates an attribute phrase, lines 02 and 02α become a formal contradiction. This is a simple form of formal contradiction. The entire antecedent from lines 01 to 04, which contain a formal contradiction, will also become a formal contradiction.

[0281] [Example sentences 5-6] 01 If <this vertical position error> is {large}, 02 if <the car> {enters a curved road from a straight road}, 02α if <the car> {enters a curved road from a straight road}, 03 even if <the vehicle speed> is {too large}, 04 if <the speed control> is {appropriate}, 05 then the <car> {can turn well},

[0282] Let us assume that this 02α line is also found in Example 5, which is written in natural language. This is Example 5-7.

[0283] [Example sentence 5-7] If the error in this vertical position is large, when the car enters a curved road from a straight road, or if the car does not enter a curved road from a straight road, if the car's speed is too high, the car will not be able to make the turn and will not be able to automatically drive within the driving lane on the curved road.

[0284] In Example Sentence 5-7, the expression "Also, if" is used as a prelude to the next expression, and although "if the car does not enter the curved road from the straight road" contradicts the previous line (line 02 of Example Sentence 5-6), it is, so to speak, casually (?) slipped in. By making Example Sentence 5-7 into a unit sentence structure, it becomes like Example Sentence 5-6, and the contradiction becomes clear. Therefore, the method of creating unit sentence structures is effective in preventing formal contradictions in the antecedent. It can be a solution to prevent contradictions.

[0285] When the antecedent is an experiential contradiction, the test must not only be a formal test, but also a semantic test. It is difficult to determine what the experiential contradiction is for "When a car {enters a curved road from a straight road}" in line 02. Let's suppose that there is a person standing there (on the curved road). In this case, the unit sentence structure is Example Sentence 5-8.

[0286] [Example sentences 5-8] 01 If <this vertical position error> is {large}, 02 if {enters a curved road from a straight road}, 02β if {is standing on the curved road}, 03 even if is {too large}, 04 if is {appropriate}, 05 then {can make the turn successfully},

[0287] If a car enters a certain place (a curved road) and there is a person there, an accident will most likely occur. The meaning of experiential contradiction is diverse, but it may refer to phenomena that are not possible in a physical sense, or something that should not happen, as in the case of this accident, or it may refer to something that violates the law.

[0288] This situation is different from a formal contradiction. In the case of a formal contradiction, an event either occurs or does not occur. On the other hand, an experiential contradiction that occurs in the experiential world means that both of the described events occur. In example sentence 5-8, both line 02 and line 02β occur. The solution to Q5 in the table is simply to "correct it." In the case of an experiential contradiction, it is necessary to assume a situation in which line 02 and line 02β occur simultaneously, and take some kind of measure to deal with this.

[0289] Writing the results of taking measures involves editing the structural form of the unit sentence structure. The four sets of conditions before 02β occurred created one context. Now that line 02β has been inserted, measures must be taken to prevent an accident from occurring. From the perspective of editing the unit sentence structure, it is necessary to think of measures that will prevent an accident, in other words, measures that will not result in an experiential contradiction, and to insert a statement in the line following line 02β that will prevent the experiential contradiction. In other words, the contradictory context must be edited to make it consistent.

[0290] However, it is difficult to edit a natural sentence like Example 5 to make it consistent by tracing and organizing the context. Therefore, a method using unit sentence structures should be introduced as a solution to Q5 in the table. In addition, in unit sentence structures, implications can be simplified using the CEIC method.

[0291] Q6: "Consequent Contradiction" According to the "Cause" section of the table in Figure 10, "Consequent Contradiction" refers to a case where there is a contradiction in the consequent. And the "Solution" section says that the contradiction in the consequent should be corrected, whether the consequent is a formal contradiction or an experiential contradiction.

[0292] The solution is the same as in the case of the antecedent contradiction in Q5. Utilize the unit sentence structure. First, let's assume that the natural sentence is Example 5-9.

[0293] [Example sentence 5-9] If the error in this vertical position is large, when a car enters a curved road from a straight road, even if the speed of the car is too high, if the speed control is appropriate, the car can turn smoothly, but if it does not turn smoothly, the car can automatically drive within the driving lane of the curved road.

[0294] Using the unit sentence structure, we get Example 5-10. After line 05, there is a sentence that is the opposite of line 05, in line 05α. This is a formal contradiction in the consequent.

[0295] [Example sentences 5-10] 01 If <this error in vertical position> is {large}, 02 if <the car> {enters a curved road from a straight road}, 03 even if <the speed of the vehicle> is {too large}, 04 if <speed control> is {appropriate}, 05 then the <car> {can complete the turn successfully}, 05α However, the <car> {can complete the turn successfully}, 06 and <can automatically stay within the driving lane on the curved road} 07 When not 08 The end 09 When not 10 The end 11 When not 12 The end 13 When not 14 The end

[0296] Regarding the formal contradiction in this latter sentence, it is necessary to express it in a way that clearly indicates a formal contradiction, and so, unlike the previous sentence example 5-7, the expression lacks naturalness. However, if the conjunction "when" is used instead of "however," and an expression that directly contradicts the previous sentence, "Cars can turn well," such as "sometimes, they can't turn well," is avoided, the contradiction in the natural sentence may not be noticed. Of course, if the following sentence, "Cars can automatically stay within their lanes on curved roads," is changed to "generally, they can do so automatically," the reader will likely perceive it more smoothly.

[0297] However, it is precisely these expressions that clarify the context through the unit sentence structure. What are the conditions under which "sometimes" a sentence can bend smoothly or not? What is the frequency and conditions under which "generally" a sentence can be bent smoothly?

[0298] The fact that such questions arise is evidence that expressions such as "sometimes" and "generally" function to avoid situations that would otherwise lead to experiential contradictions. In other words, by carefully conditioning the situation of "sometimes" and the content of "generally," many non-contradictory contexts emerge.

[0299] In this case, the "however" in line 05α of Example 5-10 is replaced by a condition. This condition corresponds to line 07, which is the opposite line of line 04, and after line 07 comes the sentence <car> {￢turns smoothly}. This is the process of editing the structural form of a unit sentence structure. In fact, such a sentence has been inserted in line 08 of Example 5-5.

[0300] The unit sentence structure method shows that even if an experiential contradiction appears to be a contradiction on the surface, by carefully analyzing the conditions, it is possible that what appears to be a contradiction actually exists in reality. The method for generating and editing the structural form of a unit sentence structure is clearly shown. The unit sentence structure method should also be introduced into the solution to Q6 in the table.

[0301] Q7: "Pay attention to both cases" According to the "Cause" section of the table in Figure 10, "Pay attention to both cases" means that you should be careful when the antecedents of two sentences are the same but the consequents are different, or when the antecedents are different but the consequents are the same.

[0302] An example of a case where two sentences have the same antecedent but different consequent is "P ⇒ Q" and "P ⇒ R." "If it rains heavily, the river will flood" (P ⇒ Q) and "If it rains heavily, work from home" (P ⇒ R) are events that normally occur in the same situation. In other words, in accordance with the sentence structure, they can be established in the same context. However, "If it rains heavily, go to work" (P ⇒ S) is a different situation. However, as we saw earlier, the issues of heavy rain and reduced water levels were resolved by inserting a context, that is, by separating the contexts. Cases where it is not possible to insert a different context are where caution is required.

[0303] Next, two sentences with different antecedents but the same consequent are "P ⇒ Q" and "R ⇒ Q." For example, if there is a sentence such as "Since Japan is a country with a large area (P), agricultural products are produced in abundance (Q)" (Sentence A), and if there is also a passage in that sentence that says, "However, Japan's land area is small (R), so 'if we make various efforts', agricultural products can be produced in abundance (Q)" (Sentence B), then some caution is required. If sentence B did not contain the expression "if we make various efforts," sentences A and B, in the same context, would state the same consequent with different antecedents (P and R). Or, the antecedents are not simply different, but are contradictory antecedents (P (large area) and R (small area) are contradictory in content), and state the same consequent.

[0304] However, sentence B includes the phrase "if we try various methods," which may prevent sentences A and B from being in the same context. This is because B can create a different path from A, "if we try various methods." Therefore, sentence B is not a contradiction.

[0305] We have seen before that contradictions can be avoided by separating the contexts. Here too, it is clear that sentence comparisons should be made while keeping in mind the differences in context. Therefore, the method of unit sentence structure can be utilized.

[0306] Q8: "Pay attention to the order of events" According to the "Cause" section of the table in Figure 10, there are sentences that are not written in the order in which the events occurred, and changing the description to the order in which the events occurred will "solve" the problem.

[0307] Q8 recommends that, for example, in a sentence such as "In order to achieve purpose A, means B must be taken," it should be written in chronological order, such as "If means B is carried out, purpose A will be realized." This is said to eliminate ambiguity.

[0308] However, this recommendation is unfamiliar to those who believe that the end comes first, then the means, or that the normal way of thinking is to first come up with a conclusion, then the means. However, if, for example, a computer could automatically rearrange events in chronological order, this would be somewhat more acceptable.

[0309] However, "for" is not the only conjunction that allows the reordering of events. The "if" in "If you want to achieve your goal, the means are these" is also the subject of this chronological conjunction. There are also "ifs" that relate means and goals without the need to change the chronological order, such as the "if" in "If these means are used, the goal will be achieved." For these reasons, it is necessary to clarify the differences in the functions of these conjunctions in the order of events. This process is also manual.

[0310] We believe that the idea of ​​Q8 in the table is valid. The present invention complements this idea. In the present invention, events are assumed to be unit sentence structures in chronological order, and the inspection is performed.

[0311] 3.1.1.2. Ambiguity in a single sentence This is when the ambiguity occurs in only one sentence.

[0312] 3.1.1.2.1. Ambiguity due to words and phrases Ambiguity that occurs in only one sentence is due to words and phrases. Both Q9's "domain-specific words" and Q10's "scope word limitation" are ambiguities caused by the words and phrases in a single sentence. They are not issues that arise from the relationship between two or more sentences. Below, we will look at Q9's "domain-specific words" and Q10's "scope word limitation."

[0313] Q9: "Domain-specific terms" According to the "Cause" section of the table in Figure 10, "Domain-specific terms" are undefined for "Cause" and require that domain-specific terms be well defined for "Solution." This document creation rule appears to be aimed at improving understanding among non-specialists. This is because a lack of understanding is sometimes described as "ambiguity."

[0314] This appears to be a question of language, not context, since technical terms must remain constant in any context.

[0315] However, for example, in legal documents, it is often the case that the wording remains unchanged, but the interpretation differs from when the law was first enacted. From the standpoint of this invention, it is possible to point out that the underlying context has changed.

[0316] Furthermore, the same words often have different meanings not due to the passage of time but due to differences in fields of expertise. For example, taking the table of contents in Figure 8 as an example, Chapter 14 is titled "Bioinspired Computational Models." This chapter is likely to be relevant to neuroscientists, mathematicians, biologists, medical scientists, and computer scientists. If they were unaware of each other's world of knowledge, words from their own field would sound different in other fields. If "genetics" were to be handled by computers, it would not be the same as biology. When defining terms, it is necessary to keep each field of expertise in mind and make definitions that are understandable. In Chapter 15, regarding natural phenomena, words consisting of the same letters have different meanings in different fields.

[0317] It is possible to call each specialized field its own "context." We do not have an abundance of different words for each field of expertise. Therefore, academic terms from one field are adapted to new academic fields. When adapted, the original words take on the specific meanings of the new field and become integrated into it. We sometimes liken this new field to a "context." For example, we might say that the word has "adapted to its new context." This is similar to the architectural term "architecture" that is used in the computer world today. If we trace the origins of the word, one might think that "architecture" first originated in architecture, but it is said that its original meaning was "foundation," and that this word was adapted to architecture.

[0318] The word "context" is very useful when talking about the applicability of such words to different fields. It can be said to be a descriptive term related to words. However, the word "context" is not limited to this and is used in a variety of different ways.

[0319] However, the scope of this invention is outside the scope of discussions of the origins of words and their various meanings.The present invention aims to clarify the core of "context" by understanding it from a formal perspective, such as the conjunction of sentences, and to utilize it as a real device for clarifying meaning.

[0320] Q10: "Range boundary words" Regarding "range boundary words," according to the "Cause" section of the table in Figure 10, the "cause" is that the words expressing the range do not explicitly describe the boundary, and the "solution" to this is to explicitly express the boundary that indicates the range.

[0321] For example, determining the boundaries of a phrase that indicates a range can be surprisingly problematic. "Above" and "below" are considered unconventional when used incorrectly by both the writer and the reader, but what about the following? "Above" and "Above" are examples. The former includes "above" and, if "above" is a numerical value, then that numerical value is included. However, "above" can be surprisingly difficult. When it comes to "above 34 degrees," it becomes unclear whether 34 degrees is included or not.

[0322] So, for example, the reader might ask the writer (if that is possible), or if not, look for similar expressions elsewhere in the document, assuming that the same writer would have used similar scope words.

[0323] Asking others and searching for similar examples can be interpreted as setting up a new context for oneself, since it is the same as asking "if...then." This is similar to how, in Example 1 (Chomsky's example), various situations were set up to match the individual words in an attempt to understand the incomprehensible sentence.

[0324] In general, phrases that indicate a range should clearly indicate the boundaries of the range to prevent misunderstandings. In this respect, the standard in Q10 of the table is effective for documents that require strict understanding.

[0325] On the other hand, in documents where strictness is not required, or where strictness is even considered tasteless, terms such as "above" and "below" have different functions. A certain novel may say, "The woman looked both over and under 30," but it does not say, "The woman looked both over and under 30."

[0326] In this way, words that indicate a range are also required to function differently depending on the context. Here too, there is a context in the "broad sense." Writers do not use expressions such as "above" and "below" as words that clearly indicate boundaries. We seem to unconsciously select contexts in the "broad sense," and use and understand words accordingly. However, this is not the subject of this invention.

[0327] 3.2. Evaluation of Examples of Compliance Requirements for Documentation Creation Based on the above discussion, we will evaluate the compliance requirements in the table in Figure 10. This table points out requirements that must be met in a specific field (software development instruction documents). However, as mentioned in each of the items in Q1 to Q10, it became clear that the ambiguity and accuracy of sentences and phrases in any item should be judged based on context. This proves the context-based approach of the present invention. Naturally, a context-based perspective is also required for document inspection.

[0328] 3.3. Database for supporting document inspection The evaluation of document creation compliance items in Figure 10 emphasizes the context-based perspective. Therefore, the database used in the inspection also needs to have a function to support the context-based approach. The context-based approach is a key feature of the present invention. Below, the function of the database will be explained, including the context-based approach.

[0329] D1: "DB of candidate expressions for conjunction phrases" The "DB of candidate expressions for conjunction phrases" typically contains expressions that indicate conditional conjunctions. "If" is an obvious conditional expression in writing, but there are other expressions that can also be used as conditional phrases. These are expressions that are candidates for conditional phrases. In addition to conditions, there are also expressions that indicate conjunction phrases of conjunction, negation, and disjunction. These expressions are stored as candidates for conjunctions, and when a similar expression is found in natural language, it is manually determined whether it corresponds to an arbitrary conjunction phrase.

[0330] For example, the target is the emphatic expressions and hidden conditional expressions mentioned earlier. Such expressions are not stored from the beginning, but are discovered during inspections and other processes, and the number is increased empirically. Machine learning can also be incorporated. This makes it possible to gradually improve the accuracy of context in document inspection. A "DB of candidate conjunction phrase expressions" refers to such a DB.

[0331] D2: "Word / Sentence Combination Matching DB" The "Word / Sentence Combination Matching DB" relates to word combinations and sentence combinations. Word combinations refer to the combination of target phrases and attribute phrases, which are elements of unit sentences. Sentence combinations refer to the combination of unit sentences. Matching refers to whether the combination of words or sentences is compatible or incompatible.

[0332] For example, in Example 2-2, all of the target phrases (<sun>) are compatible with their attribute phrases. Furthermore, in terms of sentences, lines 1 and 2, "The sun {appears in the morning}" and "The sun {goes away at night}" are compatible without contradiction. However, as we saw in the first posting of Example 2-2, lines 2 and 5, "The sun {goes away at night}" and "The sun {dances and sings all night}" are incompatible. As mentioned earlier, one is absent at night, while the other is present at night. Therefore, these two sentences are incompatible with each other. They are empirically contradictory. However, the incompatibility between the two sentences is due to the fact that they are in different contexts, but the context is taken into account when comparing them. A context exists in line 3 of Example 2-2. The sentences (unit sentences) belonging to the same area, such as lines 01 and 02, and lines 04 and 05, exist without contradiction.

[0333] The suitability of the above combinations depends on the context. Therefore, data on whether the suitability depends on the context must also be included in the combination data in the DB.

[0334] The attribute phrases in the unit sentences are managed in the infinitive form grammatically. This is for the purpose of character comparison. The management of attribute phrases in the infinitive form is similar to other DBs.

[0335] D3: "Sentence Contradiction DB" The "Sentence Contradiction DB" is a DB of contradictory sentences. A sentence is a unit sentence. However, the main sentences in this DB are contradictions between events with different target phrases. When the previous sentence "Heavy rain causes water to decline" is registered as a contradiction, the data in the DB is structured in such a way that the target phrase is generalized as x, and the two sentences "<x> is {heavy rain}" and "<x> {causes water to decline}" are contradictory when they are in some kind of relationship R (for example, a causal relationship).

[0336] Here too, a context is necessary, because when there is a relationship R, it is nothing other than the setting of a condition, that is, a context.

[0337] D4: "Word Category DB" The "Word Category DB" is a database that records the incompatibility of any attribute phrase with other attributes due to the characteristics of the category to which that attribute belongs. For example, "heavy" and "light" are incompatible. However, "heavy" and "fast" or "slow" are compatible. The category of "heavy light" is "weight," and the category of "fast slow" is "speed." In this way, attribute phrases that belong to the same category are incompatible.

[0338] Therefore, if there are two unit sentences that have the same target phrase but different attribute phrases, and the two attribute phrases belong to the same category and are located in the same context, the two unit sentences may not be valid. It is a good idea to register the categories of attribute phrases so that this check does not have to be done visually one by one.

[0339] It also supports visual inspection of context. By displaying a list of only target phrases in the same context and visually inspecting the category to which the target phrase belongs, the accuracy of the sentence can be improved. The same applies to attribute phrases. This is the effect of the present invention, which breaks down a sentence into target phrases and attribute phrases.

[0340] To create this "word category database," one can refer to the "Classification Vocabulary List" of the National Institute for Japanese Language and Linguistics. For example, "light and heavy" is included in the classification "1. Body type, 1.1 Abstract relations, 1.193 Angle, light and heavy, temperature, etc." However, since any object is "light," it can also be expressed as rising at an "angle of XX degrees," so this category cannot be applied as is. However, it can be used by further subdividing "angle, light and heavy, temperature."

[0341] D5: "Domain-Specific Word DB" The "Domain-Specific Word DB" is a dictionary-like DB that explains specialized words in a specialized field in a way that can be understood by people in any field. These words are not context-variable. As a rule, they are invariant throughout the entire document. Inspecting the similarity of domain-specific words (words with slight differences in characters) is also effective.

[0342] As mentioned earlier, any technical term can have different meanings depending on the field of expertise. Any technical term can be used in medicine, physics, or other fields, in other words, in a "broad" context. In reality, the field of expertise may need to be more specific to be practical.

[0343] D6: "Range Boundary Word DB" The "Range Word DB" is a database that stores words that express ranges, such as "greater than," "more than," and "less than." During inspection, the computer notifies the user of range words that exist in the document. In response to this notification, the user visually checks whether the document clearly expresses a range. Like domain-specific terms, these are words that apply to the entire document.

[0344] Regarding scope terms, we have already mentioned that there are scope terms with or without boundaries, and that scope terms can have a broader meaning without regard to boundaries. With this understanding in mind, it is important to recognize that any scope term, in this case, applies to the entire document.

[0345] From a practical standpoint, I will discuss how to handle range terms as a database. There are about 300 range terms in the Japanese language. Therefore, it is easy to register them in the database in advance.

[0346] In this case, registering related words according to the range word contributes to determining whether it is correct or not in relation to the context. For example, if the range word is "greater than" or "less than," specific words such as "meters" or "square kilometers" are entered according to categories such as length or area (as mentioned above). Then, when inspecting a sentence, if the range word and its specific words are found in the same context, only the target words that appear in that context, for example, are displayed in a list. This makes it possible to inspect, for example, if a context related to length or area contains a target word related to weight. This saves time during visual inspection. This is related to the "word category DB" section mentioned earlier.

[0347] It is entirely possible to learn the phrase data obtained from such work and reflect it in the database. The use of context as a structural form in the database is a distinctive feature of the present invention. Furthermore, when creating a new embodiment using the present invention, it is possible to use AI to learn empirical data on phrases. However, such matters related to AI are outside the scope of the present invention.

[0348] 4. Example of a "Context-Based Document Search Device" In this invention, a sentence-based search is performed on a document to be searched, assuming that the document has the unit sentence structure described above. As already seen, a unit sentence structure is a set of contexts consisting of conjunctions of unit sentences. Therefore, the document to be searched has a structure that allows for search by conjunction. The document search system will be described below.

[0349] In the system of Figure 4, the document to be searched is a unit sentence structure document (432). The entire system is named the "context-based document search device." Its distinctive feature is that it is based on "context." The input data for searching is a sentence (410). The unit sentence construction section (421) of the search preparation section (420) constructs the input sentence into a conjunction of unit sentences. Implication sentences and disjunction sentences are also reconstructed into conjunctions. If the sentence is inappropriate for searching, the user is prompted to re-enter it. The search subject selection section (422) selects the search target and search area, etc. If there are multiple documents in the unit sentence structure document (432), multiple documents can be selected as the search target, and the search order can also be specified. Furthermore, the search area can be freely selected, such as chapter, section, or paragraph. In the document search unit (430), the search execution unit (431) executes a search for the unit sentence structure document (432) in accordance with the search menu designated by the search subject selection unit (422), and outputs the search results in accordance with the search menu designated by the search subject selection unit (422).

[0350] When a search is deemed a good fit, the unit sentences that make up the conjunction may be separated in the context that it fits. The order of the unit sentences may also be different from the order of the searched unit sentences in the context. However, if the unit sentences are separated or in a different order, manual confirmation of the meaning is required.

Claims

1. (1.1) A sentence consisting of natural language shall be called a natural sentence, and (1.1.1) a collection of natural sentences shall be called a text. (1.2) From all the sentences in the text, unit sentences and connection relation phrases are extracted in order. (1.3) Then, following the order generally, a unit sentence relation body is created by reconstructing the extracted unit sentences and connection relation phrases. (1.3.1) For the unit sentence relation body, a person can add or modify it using unit sentences and connection relation phrases. (1.3.1.1) What is added or modified shall be called a unit sentence structure. (1.3.1.2) In the unit sentence structure, a set in which unit sentences are combined by conjunction in logical inference is called a unit context. (1.4) Extracting a unit context from the unit sentence structure is called context extraction. (1.5) Creating a new unit context according to the rules for creating the unit context is called context creation. (1.6) Checking the correctness of phrases and sentences for each unit context is called document checking. (1.7) Searching for a conjunction of unit sentences with respect to the unit sentence context structure is called context search. (1.8) A context-based, so-called "context-based" sentence-related device capable of performing the above context extraction, context creation, context checking, and context search.

2. (2.1) Regarding the unit sentence: (2.1.1) The unit sentence is a sentence consisting of an object phrase expressing an object and an attribute phrase expressing an attribute of the object. (2.1.2) The object phrase and the attribute phrase are phrases extracted from the natural sentences of the text. (2.2) Next, regarding the connection relation phrase: (2.2.1) The connection relation phrase functions to connect the unit sentences. (2.2.1α) This is because the connection relation phrase corresponds to an expression that relates sentences such as conjunctions in the natural text. (Note: Symbols such as α and β are appended to the concepts of the same hierarchical lines. The same applies hereinafter.) (2.2.2) Furthermore, the connection function of the connection relation phrase has the functions of conjunction, implication, and negation in logical inference. (2.2.3) The unit sentence relation body refers to what is formed by connecting the unit sentences by the respective functions of the connection relation phrase with conjunction, implication, and negation. (2.3) Next, regarding the "conjunction connection" in "logical inference, connected by conjunction": (2.3α) The antecedent and consequent described below are both unit sentences. Assuming that the unit sentence is represented by a symbol such as U. (2.3β) Also, assuming that the conjunction in logical inference is represented by the symbol "∧" and the implication is represented by the symbol "⇒". (2.3.1) When the connection relation phrase connecting the unit sentences is a conjunction: (2.3.1.1) The unit sentences constitute the unit context. (2.3.1.1.1) Therefore, the unit sentences can be metaphorically said to be on the same path. (2.3.2) On the other hand, in the antecedent (referred to as A) and the consequent (referred to as C) of the implication: (2.3.2.1) When the antecedent A of the implication is true: (2.3.2.1.1) The antecedent A of the implication can be connected to the consequent C in a conjunctive relationship. (2.3.2.1.2) And in the present invention, assuming that a false sentence should not exist in the document: (2.3.2.1.2.1) The implicative relationship "A⇒C" between the antecedent A of the implication and the consequent C of the implication becomes a conjunctive relationship "A∧C". (2.3.2.1.2.1α) This method of regarding the implication as conditionally equivalent to the conjunction. (2.3.2.1.2.1β) Will be called the "conditional equivalence of implication and conjunction" method.(2.3.2.1.2.1.1) Therefore, it can be said that the antecedent A of the implication and the consequent C of the implication constitute a unit context. (2.3.2.1.2.1.2) Therefore, it can be metaphorically said that the antecedent A of the implication and the consequent C of the implication are on the same path. (2.4) Next, regarding the unit sentence structure: (2.4.1α) All the antecedents described below are unit sentences. (2.4.1β) Also, assuming that the symbol "¬" represents the negation of a unit sentence, (2.4.1) When the connection relation clause in the unit sentence structure is an implication: (2.4.1.1) If there exists an antecedent "¬A10" that negates A corresponding to the antecedent A10 of the implication I1: (2.4.1.1.1) There will exist an implication I2 different from the implication I1 with ¬A10 as the antecedent. (2.4.1.2) On the other hand, if the antecedent ¬A10 does not exist: (2.4.1.2.1) A person can arbitrarily create the antecedent ¬A10. (2.5) Next, regarding the article: (2.5.1) When a unit sentence structure is separately created based on the article: (2.5.1.1) Let the article be called the "original article". (2.5.1.2) Also, in the unit sentence structure: (2.5.1.2.1) Arrange the unit sentences row by row. (2.5.1.2.2) Have an addressing function that assigns unique line numbers etc. to the unit sentences. (2.5.2) Also, in the original article from which the unit sentences are extracted: (2.5.2.1) Number the article in units of simple sentences in the ordinary grammar sense. (2.5.3) The numbers of the unit sentences are associated with the numbers of the simple sentences etc. of the original article. (2.5.4) When the original article refers to the entire document: (2.5.4.1) When the entire document is composed of items such as chapters and sections: (2.5.4.1.1) Assume that the articles of each item also individually form a unit sentence structure. (2.5.4.1.1α) Let this item - specific unit sentence structure be called the "item - specific unit sentence structure".(2.5.4.1.2) In all of the item-by-item unit document structures corresponding to the entire document, if each unit document that transversely extends across and is within each of the item-by-item unit document structures is conjunctively connected to each other, (2.5.4.1.2.1) A context that continuously traverses the whole from the beginning to the end of the entire unit document structure can be extracted. (2.5.4.1.2.1α) Such a context shall be referred to as the "global context". (2.5.4.1.2.1.1) Moreover, the global context can be extracted individually. (2.5.4.1.2.1.1α) And such individually extracted context shall be referred to as the "individual global context". (2.6) The apparatus according to claim 1, including all of the matters described in claim 2 above.

3. (3.1) Regarding the unit text structure: (3.1.1) When the unit text structure has a corresponding original text: (3.1.1α) In the unit text structure: (3.1.1.1) When the connection relation clause expresses an implication: (3.1.1.1.1) The antecedent (¬A) obtained by negating the antecedent (A) of the implication and the consequent of the antecedent ¬A are automatically generated as a structural form. (3.1.1.1.1.1) Then, if a person arbitrarily adds meaning to the structural form to create a unit text, (3.1.1.1.1.1.1) It is possible to create a unit text structure with a content different from the original text. (3.1.2) On the other hand, when the unit text structure does not have the corresponding original text: (3.1.2α) In the unit text structure: (3.1.2.1) A person can create a structural form. (3.1.2.2) A person can arbitrarily add meaning to the structural form to create a unit text. (3.1.2.3) Therefore, a person can create a unit text structure independently even without an original text. (3.2) The device according to claim 1, including all the matters described in claim 3 above.

4. (4.1) Regarding the inspection of documents, (4.1α) the inspection is related to meaning. (4.1.1) Regarding sentences, (4.1.1α) for two sentences to be semantically contradictory, it is in the same context. So, (4.1.1.1) in the contradiction inspection, (4.1.1.1α) the two unit sentences to be inspected are inspected within the same unit context. (4.1.1.1β) And the inspection area is either (4.1.1.1.1α) limited to a specific area or (4.1.1.1.1β) carried out for the entire document. (4.1.1.1.1) In the former case, the inspection can be carried out for the context of any area. (4.1.1.1.2) In the latter case, the inspection can be carried out for the entire context. (4.1.2) Regarding words and phrases, (4.1.2α) if two words or phrases are considered "strange" or "abnormal", etc., which are broadly called "ambiguous", (4.1.2.1) for two words or phrases to be ambiguous, (4.1.2.1.1) in terms of unit sentences, (4.1.2.1.1.1) it is in the target phrases or the attributive phrases. (4.1.2.1.1.1.1α) And since it is effective to compare them in the same context, (4.1.2.1.1.1.1) the two target phrases or the two attributive phrases are inspected within the same unit context. (4.1.2.1.1.1.1.1α) The inspection area is either (4.1.2.1.1.1.1.1β) limited to a specific area or (4.1.2.1.1.1.1.1γ) carried out for the entire document. (4.1.2.1.1.1.1.1) In the former case, the inspection can be carried out for the context of any area. (4.1.2.1.1.1.1.2) In the latter case, the inspection can be carried out for the entire context. (4.2) Also, when two sentences are contradictory or ambiguous, (4.2.1) by making additions or modifications to the corresponding unit context, (4.2.1.1) the contradiction and ambiguity will be resolved.The apparatus according to claim 1, comprising all of the matters recited in claim 4 above (4.3).

Images