JP7095874B2 - Natural language analysis system, analysis method and program - Google Patents

Description

The present invention relates to a natural language analysis system, analysis method and program.

Conventionally, it has been known to generate a data structure by using morphological analysis by a mathematical language analysis method such as statistics and probability and parsing such as a syntax tree or an abstract syntax tree (see, for example, Patent Document 1). ..
Patent Document 1 Japanese Unexamined Patent Publication No. 2017-191457

However, in the analysis system using the conventional syntax tree or abstract syntax tree, there is a problem that it is difficult to acquire information that is not directly related to the search term from the obtained data structure.

In the first aspect of the present invention, a sentence acquisition unit for acquiring an analysis target sentence to be analyzed, a phrase decomposition unit for generating phrase data obtained by decomposing an analysis target sentence into a phrase or a compound phrase, and a syntax of the phrase data. For analysis, an analysis system including a DAG generation unit that generates a data structure of a directed acyclic graph (DAG) from a clause or a compound clause is provided.

In the second aspect of the present invention, for the stage of acquiring the analysis target sentence to be analyzed, the stage of generating the phrase data obtained by decomposing the analysis target sentence into clauses or compound clauses, and the syntactic analysis of the clause data. , Provides an analysis method comprising the stage of generating DAG data from a phrase or compound clause.

In the third aspect of the present invention, a program for causing a computer to execute the analysis method according to the second aspect of the present invention is provided.

The outline of the above invention does not list all the features of the present invention. A subcombination of these feature groups can also be an invention.

The outline of the configuration of the analysis system 100 is shown. This is an example of a flowchart for executing the information acquisition process. This is an example of a flowchart for executing the DAG generation process. It is a conceptual diagram which shows an example of the DAG generated by the DAG generation unit 40. An example of a more specific configuration of the analysis system 100 is shown. An example of a word decomposition table is shown. An example of a particle table is shown. An example of a clause assembly table is shown. An example of the connection pattern table is shown. An example of a duplicate word table is shown. An example of a node table is shown. An example of the link table is shown. An example of the GUI screen used in the analysis system 100 is shown. This is an example of a configuration when the analysis system 100 is realized as hardware. It is a conceptual diagram which shows an example of the DAG generated by the DAG generation unit 40. It is a conceptual diagram which shows an example of the DAG generated by the DAG generation unit 40. An example of the hardware configuration of the computer 1900 functioning as the analysis system 100 is shown.

Hereinafter, the present invention will be described through embodiments of the invention, but the following embodiments do not limit the invention to which the claims are made. Also, not all combinations of features described in the embodiments are essential to the means of solving the invention.

FIG. 1 shows an outline of the configuration of the analysis system 100. The analysis system 100 includes a sentence acquisition unit 10, a phrase decomposition unit 20, a particle table setting unit 30, a DAG generation unit 40, an output unit 50, a DAG structure information acquisition unit 60, and a duplicate word setting unit 70. Be prepared.

The sentence acquisition unit 10 acquires an analysis target sentence to be analyzed. The sentence acquisition unit 10 may acquire a plurality of analysis target sentences. The analysis target sentence may be input for each sentence or for each paragraph. The analysis target sentence may be input by the user from the keyboard, or may be input by another input method such as voice input. For example, the analysis target sentence includes information on the symptoms of each disease in the medical field.

The phrase decomposition unit 20 generates phrase data in which the sentence to be analyzed is decomposed into phrases or compound phrases. In one example, the phrase decomposition unit 20 decomposes the sentence to be analyzed into words and classifies each of the words into a predetermined category. Then, the phrase decomposition unit 20 generates a phrase or a compound phrase by concatenating the decomposed words. For example, the phrase decomposition unit 20 divides the sentence to be analyzed into words such as Chinese characters, hiragana, katakana, numbers, alphabets, and special characters. The phrase decomposition unit 20 concatenates the decomposed words based on the particle table described later to generate phrase data.

The particle table setting unit 30 creates a particle table in which a word is associated with a predetermined identification code for each category. The particle table may be registered in advance. The particle table is referred to by the phrase decomposition unit 20. The particle table will be described later.

The DAG generation unit 40 generates DAG data from decomposed clauses or compound clauses for parsing the clause data. The DAG generation unit 40 of this example creates DAG data based on a pattern of identification codes for clauses or compound clauses. The DAG of this example is associated with the clauses or compound clauses constituting the sentence as nodes.

The output unit 50 outputs the DAG data generated by the DAG generation unit 40. The output unit 50 may output the DAG data to an external device of the analysis system 100, or may output the DAG data to a display unit such as a display and display the DAG data.

The DAG structure information acquisition unit 60 acquires DAG structure information. The DAG structure information is information related to the structure of past DAG data generated by analyzing the analysis target sentence. The DAG structure information acquisition unit 60 inputs the DAG structure information to the DAG generation unit 40. The DAG generation unit 40 generates DAG data based on the DAG structure information. In this way, the DAG generation unit 40 can generate DAG data by combining the existing DAG structure information and the clause data of the analysis target sentence.

The duplicate word setting unit 70 sets duplicate words that allow a shared node of DAG data. A shared node is a node in which overlapping clauses are shared by different parsing target sentences. The duplicate word setting unit 70 searches for duplicate words and generates a duplicate word table. For example, the duplicate word setting unit 70 examines a word that appears a plurality of times and displays it in a list. The duplicate word setting unit 70 may create a duplicate word table by adding, modifying, and deleting the target word to be shared among the words that appear multiple times. Further, the duplicate word setting unit 70 may have a duplicate word search function of examining words that appear multiple times in a paragraph of a sentence and displaying them in a list.

The DAG generation unit 40 shares a node when the DAG data matches the set duplicate word. By sharing duplicate words in this way, even different analysis target sentences can be associated in the DAG data.

The analysis system 100 of this example can output DAG data according to the analysis target sentence. For example, the analysis system 100 can generate a data structure for a symptom of a disease when the sentence to be analyzed is a sentence related to medical treatment. In this case, a symptom-related disease can be searched from the DAG data by searching with a keyword related to the symptom.

FIG. 2A is an example of a flowchart for executing the information acquisition process. The analysis system 100 acquires information necessary for analysis before analyzing the analysis target sentence. The analysis system 100 of this example acquires each table based on the past DAG structure information by step S200 and step S202. The flowchart of this example is an example of information acquisition processing, and is not limited to this.

In step S200, the analysis system 100 acquires the past DAG structure information. The past DAG structure information is DAG structure data acquired when the analysis system 100 analyzes the analysis target sentence in the past. However, the analysis system 100 may start the analysis of the analysis target sentence without acquiring the past DAG structure information.

In step S202, the analysis system 100 acquires each table such as a particle table and a duplicate word table. The table stored in the analysis system 100 will be described later. Through these steps, the analysis system 100 acquires information necessary for analyzing the text in advance.

FIG. 2B is an example of a flowchart for executing the DAG generation process. The analysis system 100 analyzes the input analysis target sentence in steps S204 to S210 to generate DAG data. The flowchart of this example is an example of the DAG generation process, and is not limited to this.

In step S204, the analysis target sentence is acquired. In one example, the analysis target sentence is acquired by the sentence acquisition unit 10. For example, the analysis target sentence may be directly input by the user or may be input from another device. The analysis target sentence may be input via a communication circuit. A plurality of related documents may be automatically acquired by a program by searching the WEB with a search keyword and used as an analysis target sentence.

In step S206, clause data is generated. For example, the phrase decomposition unit 20 decomposes the sentence to be analyzed into words to create a word decomposition table. Next, the phrase decomposition unit 20 refers to the particle table and assigns an identification code to the word. Next, a phrase assembly table is created by combining words without identification codes.

In step S208, DAG data is generated. For example, the DAG data is generated by the DAG generation unit 40 based on the clause data.

Steps S204 to S208 may be repeated according to the input analysis target sentence. For example, the analysis system 100 repeats steps S204 to S208 as many times as the number of input analysis target sentences. Steps S204 to S208 may be looped by the number of sentences or may be looped by the number of paragraphs. Steps S204 to S208 may be repeated until all the sentences to be analyzed are analyzed.

In step S210, DAG data is output. For example, the DAG data is output by the output unit 50. The DAG data may include a node table and a link table described later.

It should be noted that steps S200 to S210 may be executed by each hardware constituting the analysis system 100. Further, steps S200 to S210 may be executed on the computer by a program.

FIG. 3 is a conceptual diagram showing an example of the DAG generated by the DAG generation unit 40. The DAG generation unit 40 of this example creates a DAG from an analysis target sentence regarding the symptoms of a cold and influenza. (A-1) to (G-1) indicate the node number of each node.

The sentence acquisition unit 10 has acquired a plurality of analysis target sentences. For example, the sentence acquisition unit 10 acquires, as the first analysis target sentence, "general cold symptoms are nose stuffiness, cough, headache, and the like." In addition, the sentence acquisition unit 10 acquires "the symptoms of influenza are headache, myalgia, high fever, etc." as the second analysis target sentence.

The phrase decomposition unit 20 decomposes the sentence to be analyzed into a phrase or a compound phrase. The phrase decomposition unit 20 of this example changes "general cold symptoms are nasal blemishes, cough, headache, etc." to "general", "cold symptoms", "nose bleeding", and "nose bleeding". It is decomposed into "cough", "headache", "etc.", respectively. In addition, the phrase decomposition unit 20 indicates that "influenza symptoms are headache, muscle pain, high fever, etc.", "influenza symptoms are", "headache", "muscle pain", "high fever", "etc." There are. " That is, in this example, as described later, "na", "ha", "etc." are registered in the particle table.

The DAG generation unit 40 assigns a node number to each clause of the first analysis target sentence. For example, the node (A-1) corresponds to the phrase "general". Node (B-1) corresponds to the phrase "What are the symptoms of a cold?" The node (C-1) corresponds to the phrase "nose water". The node (C-2) corresponds to the phrase "cough". The node (C-3) corresponds to the phrase "headache". The node (D-1) corresponds to the phrase "is, etc.".

Further, the DAG generation unit 40 similarly assigns a node number to each clause of the second analysis target sentence. For example, the node (E-1) corresponds to the phrase "What are the symptoms of influenza?". The node (F-2) corresponds to the phrase "myalgia". The node (F-3) corresponds to the phrase "high fever". The node (G-1) corresponds to the phrase "is, etc.". In this way, the sub-particle "etc." aggregates the nodes connected in parallel with the next node.

The analysis system 100 of this example generates a DAG by setting the linking pattern to either series or parallel. Whether the connection of each node is serial or parallel may be determined according to particles and sub-particles. For example, since the nodes (C-1) to (C-3) are sandwiched between "ha" and "etc.", they are connected in parallel. Further, since the nodes (C-3), (F-2) and (F-3) are also sandwiched between "ha" and "etc.", they are connected in parallel.

The duplicate word setting unit 70 has registered "headache" in the duplicate word table. Therefore, the DAG generation unit 40 shares the "headache" of the node (C-3). As a result, the first analysis target sentence and the second analysis target sentence are associated with each other by the "headache" of the node (C-3). On the other hand, the duplicate word setting unit 70 does not register the phrase "etc." in the duplicate word table. Therefore, the DAG generation unit 40 does not share the nodes (D-1) and (G-1) indicating "etc.".

As described above, the analysis system 100 associates different analysis target sentences in the form of DAG by sharing the node. That is, although cold and flu are different illnesses, the two illnesses are associated by a DAG due to the common symptom "headache". The analysis system 100 can be represented by a DAG even if it is related to a large number of documents. Then, the analysis system 100 can automatically grow the DAG by analyzing the document and adding it to the DAG. As described above, the analysis system 100 can also be applied as an AI learning engine.

FIG. 4 shows an example of a more specific configuration of the analysis system 100. Each means of the analysis system 100 may be realized by an arbitrary hardware configuration or may be realized by a program.

The phrase decomposition unit 20 includes a word decomposition means 22, a word classification means 24, and a phrase assembly means 26. The word decomposition means 22 decomposes the sentence to be analyzed into words.

The word classification means 24 classifies the decomposed words into predetermined categories. For example, categories include Chinese characters, hiragana, conjunctive particles, parallel particles, commas, sub-particles, kuten, and the like. Then, the word classification means 24 may attach an identification code to the decomposed words corresponding to the categories. For example, the word classification means 24 identifies a word in hiragana that matches the particle "ha" used for adjunct, the parallel particle "ya", the particle "etc.", and the punctuation marks "," and ".". Is given.

The clause assembly means 26 assembles a clause based on the identification code. The phrase assembling means 26 concatenates words to which an identification code is not assigned until a word with a subsequent identification code appears. This assembles a phrase or a sequence of phrases.

The particle table setting unit 30 has a particle table acquisition unit 32 and a particle table storage unit 34. The particle table acquisition means 32 acquires a particle table, which will be described later. The particle table storage means 34 stores the particle table. The particle table is read from the particle table storage means 34 by the word classification means 24.

The DAG generation unit 40 includes a DAG creating means 42, a node connecting means 44, and a closed loop checking means 46. The DAG creating means 42 creates a DAG from the analysis target sentence. The node joining means 44 joins the generated DAG node with another DAG node. The closed loop checking means 46 checks the closed loop of the DAG generated by the node joining means 44.

Further, the DAG generation unit 40 may have a function of correcting a closed loop of a link due to topological sorting. By combining Kahn's method and Tarjan's method, the DAG generation unit 40 can investigate the node and the link that cause the closed loop. When there is a loop, the DAG generation unit 40 avoids the closed loop by newly creating a link destination node of the corresponding link and modifying the link destination.

The duplicate word setting unit 70 has a duplicate word table acquisition means 72 and a duplicate word table storage means 74. The duplicate word table acquisition means 72 acquires the duplicate word table described later. The duplicate word table storage means 74 stores the duplicate word table acquired by the duplicate word table acquisition means 72. The duplicate word table storage means 74 may output the stored duplicate word table to the node joining means 44.

FIG. 5A shows an example of a word decomposition table. The word decomposition table of this example may be used to generate the DAG of FIG. The word decomposition table of this example has columns for numbers, words, classifications, and identification codes. Each word is labeled with a classification and identification code.

The phrase decomposition unit 20 classifies each of the decomposed words into predetermined categories. For example, categories include Chinese characters, hiragana, conjunctive particles, parallel particles, commas, sub-particles, kuten, and the like. However, the word classification method is not limited to this example.

The words used to assemble the phrase are given identification codes corresponding to the categories of classification. In one example, the phrase decomposition unit 20 assigns an identification code to a predetermined hiragana in the table of the word decomposition table. For example, "2" is added as an identification code to the conjunctive particles "na" and "ha". "3" is added as an identification code to "ya", which is a parallel connection particle. "5" is added as an identification code to "etc.", which is a sub-particle that combines parallel connections into one. The reading point "," is given "4" as an identification code. "13" is added to the kuten "." As an identification code.

The phrase decomposition unit 20 does not have to assign an identification code to a word that is not used for assembling the phrase. The phrase decomposition unit 20 of this example assigns "0" as an identification code to a word that is not used for assembling the phrase.

FIG. 5B shows an example of a particle table. The particle table of this example may be used to generate the DAG of FIG.

The particle table stores the corresponding particles for each classification. For example, the particles include the particles "na" and "ha", and the identification code of "2" is associated with each other. In parallel, the particle "ya" is included, and the identification code of "3" is associated with it. The plurality of adjuncts include the particle "etc." and are associated with the identification code of "5".

FIG. 5C shows an example of a clause assembly table. The clause assembly table of this example may be used to generate the DAG of FIG. The clause assembly table has, for example, columns for numbers, clauses and compound clauses, classifications, divisions and identification codes. Each clause or compound clause is labeled with a classification, division and identification code.

The phrase decomposition unit 20 concatenates words without an identification code into a phrase or a continuous phrase. As a result, the character string that becomes the node of the DAG becomes a meaningful mass as much as possible, so that it is easy to handle. For example, the "general" clause is concatenated because the "general" identification code is "0" and the "na" identification code is "2". And, like the particle "na", the phrase of "general" is a connecting particle whose division is an adjunct, and "2" is added as an identification code.

Similarly, the phrase "cold symptom is" is concatenated because the identification code for "cold", "no" and "symptom" is "0" and the identification code for "ha" is "2". ing. And, the phrase of "the symptom of a cold" is a connecting particle whose division is an adjunct, like the particle "ha", and "2" is given as an identification code.

FIG. 5D shows an example of a connection pattern table. The connection pattern table of this example may be used to generate the DAG of FIG. In the connection pattern table, the pattern of the identification code attached to the series of clauses of the sentence to be analyzed is stored. The DAG generation unit 40 creates a series of DAGs using the pattern of the identification code stored in the connection pattern table.

In one example, the DAG generator 40 creates a chain of DAGs based on the pattern 2-2-3-4-5 in the clause assembly table. For example, pattern 2-2 shows a series connection. In patterns 2-3, 2 indicates the node at the beginning of the branch. Patterns 3-4-4 show parallel links. Pattern 5 shows a node where parallel connections converge. As a result, the DAG generation unit 40 has (A-1) → (B-1), (B-1) → (C-1), (B-1) → (C-2), and (B). -1) → (C-3), (C-1) → (D-1), (C-2) → (D-1), (C-3) → (D-1), Can be linked.

FIG. 5E shows an example of a duplicate word table. The duplicate word table of this example may be used to generate the DAG of FIG. Duplicate words for sharing at the time of creating a DAG are registered in the duplicate word table. "Headache" and "high fever" are registered in the duplicate word table of this example. Therefore, if a "headache" or "high fever" is present on a DAG node, the DAG is created as a shared node. As a result, the analysis system 100 creates a series of DAGs for each sentence to be analyzed, but if there are duplicate words between the DAGs, the paragraphs can be related by connecting them as duplicate words. ..

The closed loop checking means 46 may check the closed loop when the nodes are shared. When the closed loop exists, the DAG generation unit 40 creates a new node and releases the closed loop.

FIG. 5F shows an example of a node table. The node table is an example of DAG structure information. The node table of this example may be used to generate the DAG of FIG. The node number and the contents corresponding to the node number are registered in the node table.

Specifically, the content of "general" is registered for the node (A-1). The content "What are the symptoms of a cold" is registered for the node (B-1). The content of "nose water" is registered for the node (C-1). The content of "cough" is registered for the node (C-2). The content of "headache" is registered for the node (C-3). The content "is, etc." is registered for the node (D-1).

FIG. 5G shows an example of a link table. The link table is an example of DAG structure information. The link table of this example may be used to generate the DAG of FIG. In the link table, the link source of the node and the link destination are registered. The link source of the node and its link destination may be determined based on the connection pattern table.

Specifically, the node (A-1) of the "node table" is created from the information of "number 1" of the "clause assembly table". Since the identification code of "number 1" in the "phrase assembly table" is "2", it represents an adjunct. Further, since the identification code of "number 2" in the "phrase assembly table" is "2", it is a serial link. Therefore, the node (B-1) is registered in the link destination with respect to the link source node (A-1).

Next, since the identification code of "number 3" in the "phrase assembly table" is "3" and the identification code of "number 4" and "number 5" is "4", parallelism is shown. Since the identification code of "number 2" is "2", parallelism starts from the phrase following "number 2". Since the identification code of "number 6" is "5", the previous parallel clauses converge in the clause of "number 6". Therefore, the nodes (C-1) to (C-3) are registered in the link destination with respect to the link source node (B-1). Further, the nodes (C-1) to (C-3) are registered in the link source with respect to the link destination node (D-1).

The DAG generation unit 40 can create a DAG by connecting nodes using a node table and a link table. The DAG generation unit 40 may create a link table in parallel with the node table. The output unit 50 may output a node table and a link table as DAG data.

FIG. 6 shows an example of a GUI screen used in the analysis system 100. In the GUI screen of this example, the analysis target sentence is input by the user in the analysis target sentence input field. The analysis target sentence may be input from the keyboard, or may be input by another input method such as voice input. Further, the sentence acquisition unit 10 may acquire the analysis target sentence from the medium created by the digital data. The sentence acquisition unit 10 may be read by each HTML document program associated with a link on the Internet. The analysis system 100 executes the analysis of the analysis target sentence by clicking the analysis execution button after inputting the analysis target sentence. The GUI screen is not limited to this.

FIG. 7 is an example of a configuration in which the analysis system 100 is realized as hardware. The analysis system 100 includes a CPU 110, a main memory 120, an HDD 130, an input device 140, and a display 150. The hardware configuration of this example is an example and is not limited thereto.

The CPU 110 executes various arithmetic processes for realizing the sentence acquisition unit 10, the phrase decomposition unit 20, and the DAG generation unit 40. For example, the CPU 110 refers to the program read by the main memory 120 and executes various arithmetic processes according to the procedure indicated by the program.

The main memory 120 stores an acquisition program for an analysis target sentence, a phrase decomposition program, and a DAG generation program. The main memory 120 may appropriately store other programs. Further, a plurality of addresses may be assigned to the main memory 120. By specifying the address and accessing the stored data, the CPU 110 can execute arithmetic processing using the data.

The HDD 130 operates as a storage unit for storing each table. For example, the HDD 130 stores DAG structure information, a word decomposition table, a phrase assembly table, and a particle table. Further, a plurality of addresses may be assigned to the HDD 130.

The input device 140 is a user input device for inputting text. For example, the input device 140 is an input device such as a keyboard.

The display 150 displays a GUI screen for inputting a sentence to be analyzed. The display 150 may appropriately display information necessary for the operation of the analysis system 100. Further, the display 150 may display the analysis result of the analysis system 100.

The hardware configuration of the analysis system 100 may be connected to each other by a data communication path such as a “system bus”. As a result, information is transmitted / received and processed between the hardware.

Here, the analysis system 100 generates DAG structure information by automatically analyzing the analysis target sentence and records it in the HDD 130. For example, when the user inputs a sentence, the CPU 110 executes a phrase decomposition program, generates a table, and records it in the HDD 130.

For example, when the CPU 110 acquires a sentence such as "general cold symptoms are nose, cough, headache, etc." as an analysis target sentence, the CPU 110 executes word decomposition processing on the sentence. Create a "word decomposition table". The CPU 110 assigns an identification code of the "word decomposition table" by using the "particle table" read from the HDD 130 into the main memory 120. After that, the CPU 110 executes the clause assembly process, creates a "clause assembly table", and records it in the HDD 130.

After that, the CPU 110 executes a DAG generation program to interpret the pattern of the identification code of the "phrase assembly table". Then, the CPU 110 creates a "node table" and a "link table". Then, the CPU 110 interprets the DAG data output program and outputs the "node table" and the "link table".

FIG. 8 is a conceptual diagram showing an example of the DAG generated by the DAG generation unit 40. The analysis system 100 of this example adds data to the DAG structure information by adding sentences.

The sentence acquisition unit 10 has, as the analysis target sentence, the first analysis target sentence "the symptoms of influenza are headache, myalgia, high fever, etc." and the second analysis target sentence "in the case of high fever". Requires administration of antihypertensive agent. " In this example, "ha" and "etc." are registered in the particle table. In addition, "high fever" is registered in the duplicate word table.

The analysis system 100 creates DAG data (D-1, E-1, E-2, E-3, F-1) from the first analysis target sentence. Here, when the second analysis target sentence is added, "high fever" matches and "high fever" is registered in the duplicate word table, so that the node (E-3) is shared and the node (F-) is shared. 2) is linked to the node (E-3).

In addition, the analysis system 100 of this example certifies that "high fever" and "in the case of high fever" match by using the method of calculating the coincidence rate described later. Then, the analysis system 100 determines that "high fever" and "in the case of high fever" match, and adds data to the DAG structure information.

FIG. 9 is a conceptual diagram showing an example of the DAG generated by the DAG generation unit 40. In this example, "ha" and "ga" are registered in the particle table.

The sentence acquisition unit 10 stated that the first analysis target sentence, "Influenza prescription is administered by drug A in Japan," and the second analysis target sentence, "Influenza prescription is the United States." Then, drug B is administered. "

The analysis system 100 creates DAG data (A-1, B-1, C-1) from the first analysis target sentence. The analysis system 100 adds the DAG data of the second analysis target sentence to the DAG data of the first analysis target sentence.

The DAG generation unit 40 of this example calculates the matching rate between the word included in the phrase or compound phrase of the analysis target sentence and the word included in the phrase or compound phrase of another analysis target sentence. The DAG generation unit 40 shares a node when the match rate is equal to or higher than a predetermined threshold value.

The node joining means 44 calculates the matching rate between the word included in the "influenza prescription" and the word contained in the "influenza prescription". For example, the node connecting means 44 has the words "influenza" and "prescription" of the node (A-1) of the first analysis target sentence and the "influenza prescription" of the second analysis target sentence. Compare the words "flu" and "prescription".

For example, since "influenza" matches, 100%, and "prescription" and "prescription" match 2/3 characters, so 66%, and the total match rate is 166/200 = 83%. When the matching rate for sharing is set to 80% or more, the analysis system 100 determines that the clauses of "influenza prescription" and "influenza prescription" match, and shares the node. On the other hand, since the match rate between the node (B-1) and the node (B-2) is 0%, a new node is added. Moreover, since the match rate of the node (C-1) is 100%, it is shared. Therefore, the analysis system 100 can link the second analysis target sentence as (A-1)-(B-2)-(C-1).

As described above, the analysis system 100 can analyze the meaning of big data documents and analyze the context by analyzing the analysis target sentence using natural language processing and DAG. That is, the analysis system 100 can realize data analysis in consideration of the causal relationship of symptoms and diseases that are not directly related to the search keyword.

Further, the analysis system 100 does not merely extract the features of translation and sentences in natural language processing, but also analyzes using the data structure of DAG. Therefore, the analysis system 100 can correctly store important causal relationships as data when analyzing a large-scale document such as a machine failure or a diagnosis of a disease. Therefore, the analysis system 100 can obtain a DAG data structure obtained by decomposing a document into clauses by paying attention to a causal relationship.

The analysis system 100 is not limited to Japanese, and can be similarly applied to other languages. In this case, by storing various tables peculiar to each language, it can be applied to languages other than Japanese in a manner corresponding to the grammar and words of each language.

FIG. 10 shows an example of the hardware configuration of the computer 1900 that functions as the analysis system 100. Further, a plurality of computers may cooperate to function as the analysis system 100.

The computer 1900 according to the embodiment is connected to a CPU peripheral portion having a CPU 2000, a RAM 2020, a graphic controller 2075, and a display device 2080 connected to each other by a host controller 2082, and to a host controller 2082 by an input / output controller 2084. An input / output unit having a communication interface 2030, a hard disk drive 2040, and a DVD drive 2060, and a legacy input / output unit having a ROM 2010, a flexible disk drive 2050, and an input / output chip 2070 connected to the input / output controller 2084. Be prepared.

The host controller 2082 connects the RAM 2020 to the CPU 2000 and the graphic controller 2075 that access the RAM 2020 at a high transfer rate. The CPU 2000 operates based on the programs stored in the ROM 2010 and the RAM 2020, and controls each part. The graphic controller 2075 acquires image data generated on a frame buffer provided in the RAM 2020 by the CPU 2000 or the like, and displays the image data on the display device 2080. Alternatively, the graphic controller 2075 may include a frame buffer internally for storing image data generated by the CPU 2000 or the like.

The input / output controller 2084 connects the host controller 2082 to a communication interface 2030, a hard disk drive 2040, and a DVD drive 2060, which are relatively high-speed input / output devices. Communication interface 2030 communicates with other devices via the network. The hard disk drive 2040 stores programs and data used by the CPU 2000 in the computer 1900. The DVD drive 2060 reads a program or data from the DVD-ROM 2095 and provides it to the hard disk drive 2040 via the RAM 2020.

Further, the input / output controller 2084 is connected to the ROM 2010, the flexible disk drive 2050, and the relatively low-speed input / output device of the input / output chip 2070. The ROM 2010 stores a boot program executed by the computer 1900 at startup, and / or a program depending on the hardware of the computer 1900. The flexible disk drive 2050 reads a program or data from the flexible disk 2090 and provides it to the hard disk drive 2040 via RAM 2020. The input / output chip 2070 connects the flexible disk drive 2050 to the input / output controller 2084, and inputs / outputs various input / output devices via, for example, a parallel port, a serial port, a keyboard port, a mouse port, and the like. Connect to controller 2084.

The program provided to the hard disk drive 2040 via the RAM 2020 is stored in a recording medium such as a flexible disk 2090, a DVD-ROM 2095, or an IC card and provided by the user. The program is read from the recording medium, installed on the hard disk drive 2040 in the computer 1900 via the RAM 2020, and executed in the CPU 2000. The program is installed on the computer 1900 and causes the computer 1900 to function as each configuration of the analysis system 100.

When the information processing described in the program is read into the computer 1900, the sentence acquisition unit 10, the phrase decomposition unit 20, and the particle table setting, which are specific means in which the software and the various hardware resources described above cooperate with each other, are set. It functions as at least a part of a unit 30, a DAG generation unit 40, an output unit 50, a DAG structure information acquisition unit 60, and a duplicate word setting unit 70. Then, by realizing the calculation or processing of information according to the purpose of use of the computer 1900 in the present embodiment by this specific means, a unique analysis system 100 according to the purpose of use is constructed.

As an example, when communicating between the computer 1900 and an external device or the like, the CPU 2000 executes a communication program loaded on the RAM 2020, and a communication interface is based on the processing content described in the communication program. Instruct 2030 to process communication. Under the control of the CPU 2000, the communication interface 2030 reads out the transmission data stored in the transmission buffer area or the like provided on the storage device such as the RAM 2020, the hard disk drive 2040, the flexible disk 2090, or the DVD-ROM 2095, and transfers the transmission data to the network. The received data transmitted or received from the network is written to a receive buffer area or the like provided on the storage device. As described above, the communication interface 2030 may transfer the transmission / reception data to / from the storage device by the DMA (direct memory access) method, and instead, the CPU 2000 may transfer the transfer source storage device or the communication interface 2030. The transmitted / received data may be transferred by reading data from the data and writing the data to the communication interface 2030 or the storage device of the transfer destination.

Further, the CPU 2000 is a whole or necessary part from files or databases stored in an external storage device such as a hard disk drive 2040, a DVD drive 2060 (DVD-ROM 2095), and a flexible disk drive 2050 (flexible disk 2090). Is read into the RAM 2020 by DMA transfer or the like, and various processes are performed on the data on the RAM 2020. Then, the CPU 2000 writes the processed data back to the external storage device by DMA transfer or the like. In such processing, the RAM 2020 can be regarded as temporarily holding the contents of the external storage device. Therefore, in the present embodiment, the RAM 2020 and the external storage device are collectively referred to as a memory, a storage unit, a storage device, or the like. Various information such as various programs, data, tables, and databases in the present embodiment are stored in such a storage device and are subject to information processing. The CPU 2000 can also hold a part of the RAM 2020 in the cache memory and read / write on the cache memory. Even in such a form, the cache memory plays a part of the function of the RAM 2020. Therefore, in the present embodiment, the cache memory is also included in the RAM 2020, the memory, and / or the storage device, unless otherwise indicated. do.

Further, the CPU 2000 includes various operations described in the present embodiment, information processing, condition determination, information retrieval / replacement, and the like, which are specified by the instruction sequence of the program for the data read from the RAM 2020. Is processed and written back to RAM 2020. For example, when the CPU 2000 determines a condition, whether or not the various variables shown in the present embodiment satisfy conditions such as large, small, above, below, and equal to other variables or constants. If the condition is satisfied (or not satisfied), it branches to a different instruction sequence or calls a subroutine.

Further, the CPU 2000 can search for information stored in a file in the storage device, a database, or the like. For example, when a plurality of entries in which the attribute value of the second attribute is associated with the attribute value of the first attribute are stored in the storage device, the CPU 2000 describes the plurality of entries stored in the storage device. By searching for an entry in which the attribute value of the first attribute matches the specified condition and reading the attribute value of the second attribute stored in that entry, it is associated with the first attribute that satisfies the predetermined condition. The attribute value of the second attribute obtained can be obtained.

The program or module shown above may be stored in an external recording medium. As the recording medium, in addition to the flexible disk 2090 and DVD-ROM2095, an optical recording medium such as DVD, Blu-ray (registered trademark) or CD, a magneto-optical recording medium such as MO, a tape medium, and a semiconductor such as an IC card are used. A memory or the like can be used. Further, a storage device such as a hard disk or RAM provided in a dedicated communication network or a server system connected to the Internet may be used as a recording medium, and a program may be provided to the computer 1900 via the network.

Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various changes or improvements can be made to the above embodiments. It is clear from the description of the claims that the form with such changes or improvements may be included in the technical scope of the present invention.

The order of execution of each process such as operation, procedure, step, and step in the apparatus, system, program, and method shown in the claims, specification, and drawings is particularly "before" and "prior to". It should be noted that it can be realized in any order unless the output of the previous process is used in the subsequent process. Even if the scope of claims, the specification, and the operation flow in the drawings are explained using "first", "next", etc. for convenience, it means that it is essential to carry out in this order. It's not a thing.

10 ... sentence acquisition unit, 20 ... phrase decomposition unit, 22 ... word decomposition means, 24 ... word classification means, 26 ... phrase assembly means, 30 ... auxiliary word table setting unit, 32 ... Auxiliary table acquisition means, 34 ... Auxiliary table storage means, 40 ... DAG generation unit, 42 ... DAG creation means, 44 ... Node connection means, 46 ... Closed loop check means, 50. ... Output unit, 60 ... DAG structure information acquisition unit, 70 ... Duplicate word setting unit, 72 ... Duplicate word table acquisition means, 74 ... Duplicate word table storage means, 100 ... Analysis System, 110 ... CPU, 120 ... main memory, 130 ... HDD, 140 ... input device, 150 ... display, 1900 ... computer, 2000 ... CPU, 2010 ... ROM, 2020 ... RAM, 2030 ... communication interface, 2040 ... hard disk drive, 2050 ... flexible disk drive, 2060 ... DVD drive, 2070 ... input / output chip, 2075 ... Graphic controller, 2080 ... Display device, 2082 ... Host controller, 2084 ... Input / output controller, 2090 ... Flexible disk, 2095 ... DVD-ROM

Claims (9)

The sentence acquisition unit that acquires the analysis target sentence to be analyzed,
A phrase decomposition unit that generates phrase data obtained by decomposing the analysis target sentence into phrases or compound phrases, and
A DAG generator that generates DAG data from the phrase or compound clause is provided for parsing the phrase data.
The phrase decomposition unit decomposes the analysis target sentence into words, classifies each of the words into predetermined categories, and classifies them into predetermined categories.
Further, a particle table setting unit for creating a particle table in which a predetermined identification code for each category and the word are associated with the word is provided.
The phrase decomposition unit generates a phrase or a compound phrase by concatenating the decomposed words based on the particle table.
Analysis system. The DAG generation unit generates the DAG data from the phrase or compound clause based on the connection pattern table in which the pattern of the identification code attached to the series of clauses or compound clauses of the analysis target sentence is stored.
The analysis system according to claim 1. It also has a DAG structure information acquisition unit that acquires DAG structure information including the structure data of past DAG data.
The analysis system according to claim 1 or 2 , wherein the DAG generation unit generates the DAG data based on the DAG structure information. Further provided with a duplicate word setting unit having a duplicate word table in which duplicate words that allow the shared node of the DAG data are registered.
Any one of claims 1 to 3 in which the DAG generation unit sets a node corresponding to the content as a shared node when the content of the DAG data matches the duplicate word registered in the duplicate word table. The analysis system described in item 1. The DAG generation unit calculates the matching rate between the word included in the phrase or compound phrase of the analysis target sentence and the word included in the phrase or compound phrase of another analysis target sentence, and the matching rate is predetermined. The analysis system according to any one of claims 1 to 4 , which shares a node when the threshold value is equal to or higher than the specified threshold value. At the stage of acquiring the analysis target sentence to be analyzed,
The stage of generating phrase data obtained by decomposing the analysis target sentence into clauses or compound clauses, and
A step of generating DAG data from the phrase or compound clause for parsing the phrase data is provided .
The stage of generating the phrase data is
The stage of decomposing the analysis target sentence into words and classifying each of the words into predetermined categories, and
The stage of creating a particle table in which a predetermined identification code for each category and the word are associated with each other, and
Based on the particle table, the stage of generating a phrase or compound phrase by concatenating the decomposed words, and
Have
analysis method. Further provided with a step of setting a duplicate word table in which duplicate words are registered, which allows the shared node of the DAG data.
A claim that the step of generating the DAG data includes a step of setting a node corresponding to the content as a shared node when the content of the DAG data matches the duplicate word registered in the duplicate word table. The analysis method according to 6. At the stage of generating the DAG data, the match rate between the word contained in the phrase or compound phrase of the analysis target sentence and the word included in the phrase or compound phrase of another analysis target sentence is calculated, and the match rate is calculated. Has a stage to share a node when is greater than or equal to a predetermined threshold
The analysis method according to claim 6 or 7. A program for causing a computer to execute the analysis method according to any one of claims 6 to 8 .

Images