JP3655501B2 - Information search method, information search apparatus, and recording medium - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention, for example, efficiently searches for data in an information device for managing and using document data related to monitoring, operation, operation, and maintenance of equipment such as a plant and real-time data directly obtained from the equipment. Related to the method.
[0002]
[Prior art]
Conventionally, techniques have been developed for interpreting natural language questions that people use on a daily basis, and searching for and presenting information that meets the user's intentions.
[0003]
In particular, in the field of document search, the user's query as the search condition and the document to be searched are both described in natural language. A practical natural language search interface is realized (for example, “document information search device (Japanese Patent Laid-Open No. 7-192020)”).
[0004]
An interface for searching a relational database in a natural language has also been developed (for example, “Database Natural Language Interface Device and Method (Japanese Patent Laid-Open No. 9-44508)”). In this method, using a knowledge describing the relationship between the structure of the database and the expression in the natural language, the user's question sentence is converted into a database search language such as SQL and the search is performed.
[0005]
On the other hand, not only an explicit question of a user but also a method of performing a search using environment information at the time of a user's question or data registration as a search condition has been developed. For example, the “search device (Japanese Patent Laid-Open Publication No. Hei 7-141389)” uses information such as the location and time at which the document was created / searched and the business schedule of the person who used the document at the time of document search. This is a way to enable searching even when the system cannot give explicit questions to the system.
[0006]
[Problems to be solved by the invention]
As described above, document search and a natural language search interface for a relational database whose structure and contents are relatively static are in a practical stage. In the office, most of the data to be used is document data or data stored in a relational database. Therefore, it can be said that the conventional natural language search interface is effective in office work.
[0007]
However, for example, in work such as plant monitoring and control, the information handled by the user is only part of the document data such as manuals and operation plans. In addition, sensor values, alarms, operation histories, etc. The work is carried out while referring to a large number of dynamic real-time data (in this case, particularly referred to as process data). Even in such business, there is an increasing need to search for and use necessary information by inquiring in speech or text in natural language. In addition, it is desired to search for a plurality of different types of data such as document data and real-time data (for example, process data) and to quickly cope with a process abnormality while using both.
[0008]
One of the problems of the present invention is to realize an integrated natural language search interface that enables both search and use of both document data and real-time data obtained from, for example, a plant to be monitored. It is intended to reduce the labor of accessing information in operations such as monitoring and control of facilities and to improve the efficiency of operations.
[0009]
The above-described “search device (Japanese Patent Laid-Open No. Hei 7-141389)” is a method of using environment information as a search condition for document search. In this case, the environment information is created mainly in an office. -It means the environment related to use, and its meaning is different from the real-time environmental information in the external real world, such as in the monitoring and control of facilities such as plants.
[0010]
In the present invention, not only document data but also an environment related to retrieval and use of both document data and real-time data is considered.
[0011]
It is an object to more actively support information access by users in the environment by using information of the outside world such as a plant as environmental information. Furthermore, the real-time data in the present invention is different from the environmental information in the “search device” described above in that it is data that is a target to be searched and used by the user and is also environmental information.
[0012]
As a specific example, for example, when an abnormality occurs in some equipment in a plant, the user can ask only "What is the recovery method?" The purpose is to realize the function of presenting. As in this example, in business such as monitoring and control, inquiries in the user's natural language tend to be concise and omitted expressions.
[0013]
Therefore, the present invention also has an object to provide appropriate information by interpreting a simple inquiry sentence according to environmental information (real-time data).
[0014]
[Means for Solving the Problems]
The information search method of the present invention generates a search command for searching document data and real-time data by interpreting an input search request sentence, and based on the generated search command, the document data and the real-time data. By searching and presenting at least one of them, document data and real-time data can be searched efficiently, for example, reducing the work of accessing information in operations such as monitoring and control of facilities such as plants, and making operations more efficient Can be achieved.
[0015]
In addition, the information search method of the present invention interprets an input search request sentence based on real-time data acquired at the time of the search request (more specifically, the search request sentence based on the real-time data at the time of the search request). Generating a search command for searching document data and real-time data, and searching for at least one of the document data and the real-time data based on the generated search command Thus, document data and real-time data can be searched efficiently, for example, the work of accessing information in operations such as monitoring and control of facilities such as plants can be reduced, and operations can be made more efficient. In addition, it is possible to efficiently search for appropriate information by simply inputting a simple search request sentence and interpreting it based on real-time data.
[0016]
Preferably, when generating the search command, it is determined whether to accept or reject the request based on the real-time data at the time of the search request.
[0017]
Preferably, when the search command is generated, the content of the interpretation of the search request sentence performed based on the real-time data at the time of the search request is presented.
[0018]
The information retrieval apparatus according to the present invention includes a document data storage unit that stores document data, a real-time data storage unit that acquires and stores real-time data, and the document data and the real-time data by interpreting an input search request statement. A search command generation unit that generates a search command for searching for the document data, and a search unit that searches and presents at least one of the document data and the real-time data based on the search command. For example, it is possible to reduce the labor of accessing information in operations such as monitoring and control of facilities such as plants, and to improve the efficiency of operations.
[0019]
The information search apparatus of the present invention stores document data storage means for storing document data, real-time data storage means for acquiring and storing real-time data, and an inputted search request sentence stored in the real-time data storage means. The document data and the real-time data are interpreted based on the real-time data at the time of the search request (more specifically, the interpretation result of the search request sentence is supplemented or narrowed down based on the real-time data at the time of the search request). By providing search command generating means for generating a search command for searching, and search means for searching and presenting at least one of document data and real-time data based on the search command, document data and real-time data are provided. Search efficiently, for example, for monitoring and control of facilities such as plants. That information to reduce the labor of access, business can be made more efficient. In addition, it is possible to efficiently search for appropriate information by simply inputting a simple search request sentence and interpreting it based on real-time data.
[0020]
Preferably, the search command generation unit determines whether to accept or reject the request based on real-time data at the time of the search request stored in the real-time data storage unit.
[0021]
Preferably, the search command generation means presents the contents of the interpretation of the search request sentence performed based on the real time data at the time of the search request stored in the real time data storage means.
[0022]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
[0023]
FIG. 1 is a block diagram showing the configuration of the information search system according to this embodiment. The information search system shown in FIG. 1 is roughly composed of a document data management unit 1, a process data management unit 2, a search command generation unit 3, a data presentation unit 4, and a search command generation rule storage unit 31.
[0024]
The document data management unit 1 stores and manages document data (including registration and search). The document data storage unit 12 and a document data registration unit 11 that registers a document in the document data storage unit 12 The document data search unit 13 searches for a document from the document data storage unit 12.
[0025]
The information retrieval system shown in FIG. 1 can be configured on a computer. That is, among the above units, the document data storage unit 12, the process data storage unit 22, and the search command generation rule storage unit 31 can be accessed by a computer using a magnetic disk (floppy disk, hard disk, etc.), optical disk (CD-ROM, DVD, etc.). ), Configured in a recording medium such as a semiconductor memory, and other components executed on a computer stored in a recording medium such as a magnetic disk (floppy disk, hard disk, etc.), optical disk (CD-ROM, DVD, etc.), semiconductor memory, etc. It can also be configured as a program that can be executed.
[0026]
The document data management unit 1 may be realized using a conventional document database. As a document search function, a document search using a keyword logical expression as a search condition and a document search using a natural language query as a search condition. As long as both are included.
[0027]
The process data management unit 2 stores and manages (including registration and search) process data of a plant or the like that is real-time data. The sensor data from a control device such as a sensor attached to the facility. The process data acquisition unit 21 acquires process data such as an alarm and an operation history of the operator. Of the acquired process data, necessary data is stored in the process data storage unit 22. In addition, the process data search unit 23 responds to the inquiry with the process data (that is, the current value) that is always acquired by the process data acquisition unit 21 or the past process data that is stored in the process data storage unit 22 ( That is, the process data history) is searched.
[0028]
The process data management unit 2 may be realized by using means that has been conventionally realized by a control system for facilities such as plants.
[0029]
The search command generation unit 3 interprets a query sentence (search request sentence) in a natural language from the user, and adds the search request sentence input by the user from the document data management unit 1 or the process data management unit 2 described above. A search instruction for searching for the information is generated.
[0030]
As will be described later, the search command generation unit 3 obtains process data that is always acquired by the process data acquisition unit 21 as the state of equipment such as a plant, that is, environmental information, and a search input by the user based on this process data. Controls how the request sentence is interpreted.
[0031]
The search command generation rule storage unit 31 interprets a natural language request sentence and stores a plurality of knowledge / rules for generating a search command based on environment information. That is, the search command generation unit 3 generates a search command using the rules stored in the search command generation rule storage unit 31.
[0032]
The data presenting unit 4 is for presenting document data and process data to the user. The user's inquiry is made by the search command generation unit 3 by using the search command to the document data search unit 13 or the process data search unit 23. The document data search unit 13 or the process data search unit 23 searches for data. The search result data is presented to the user by the data presentation unit 4.
[0033]
The essential and novel point in the configuration of the present invention is that the user's request for both the document data management unit 1 and the process data management unit 2 is uniformly received, each data is searched and presented to the user. This is the processing content of the search command generation unit 3 that appropriately interprets the user's request based on the configured point and the state of the plant.
[0034]
Next, knowledge / rule data stored in the search command generation rule storage unit 31 will be described.
[0035]
2 and 3 are diagrams showing document classes, which are knowledge used for searching document data, and FIG. 2 shows an outline of the document classes in a tree shape. For example, FIG. 3 shows a detailed description of the document class 21 represented as “unsteady operation” in FIG.
[0036]
In FIG. 3, the document class “unsteady operation” is a concept belonging to “monitoring / control”, and the superclass is “monitoring / control”. Further, the related term “unsteady operation” means that the presence or absence of the word in the document data is a word that determines whether or not the document is classified into the document class. Words such as “start-up”, “stop” and “abnormal processing” are described as related terms of “unsteady operation”. In FIG. 3, the term “start-up”, “start-up”, “start-up”, “start-up” indicates that these four terms are synonymous in the context of “unsteady operation”. .
[0037]
FIG. 4 is a diagram showing an example of the configuration of the plant, and shows the configuration of a very simple plant equipment “reaction can”. In the search command generation rule storage unit 31, as knowledge corresponding to this, The knowledge described in the format as shown in FIGS. 5 to 7 is stored. FIG. 5 is a diagram showing the configuration of the plant equipment as shown in FIG. 4 and the class to which each element of the equipment belongs in a tree shape. In FIG. 5, nodes indicated by ellipses such as “sensor” and “water level” represent classes (abstract concepts), and nodes represented by rectangles such as “NO.1 reaction cans” are instances (specific concepts; exist in the target plant). To represent the actual equipment). In FIG. 5, the upper and lower relationships between classes and instances are indicated by vertical trees (connections indicated by solid lines). For example, “NO.1 reaction can” is an instance of “reaction can”, and “reaction can” is expressed as a subclass of “can”. Moreover, the connection shown with the broken line in FIG. 5 represents the relationship between concepts. For example, class 31 “sensor” measures “physical quantity”, “thermometer” measures “temperature”, “equipment” consists of “equipment”, “NO.1 reaction can” is “NO” .1 can thermometer "and" NO.1 stirrer "and other sensors and actuators (in other words," NO.1 can thermometer "and" NO.1 stirrer "sensors and actuators belong The relationship that the equipment is “NO.1 reactor” (in FIG. 7, this relationship is described in the item “equipment”) is described in FIG.
[0038]
FIG. 6 is a detailed description of each class including the upper / lower relationship between the classes shown in FIG. 5 and the relationship between the concepts. For example, from the class “water level meter” in FIG. It can be read that the class is a super class and is also a super class of the class “upper limit” in FIG. Further, when looking at the class “water level meter”, it can be seen that the sensor measures the “water level” of the physical quantity. In addition, the related word, which is one of the description items of the class “water level meter”, describes “water level meter”, “liquid level meter”, and “level meter”. This means that there are variations (synonyms) such as “liquid level gauge” and “level gauge”.
[0039]
7 is a description of the detailed description of each instance including the relationship between the concepts of the instances shown in FIG. 5, for example, “NO.1 upper limit” shown fifth from the top in FIG. It is described as an instance of a class called “upper limit”. A device represented by each instance is provided with a tag that is a unique identifier for identifying the device. For example, as illustrated in FIG. 7, the tag “NO.1 upper limit” is “LS1001”.
[0040]
The document classification class described in FIGS. 2 and 3 and the device classification class and instance described in FIGS. 4 to 7 have the same format. For example, “equipment” “device” in FIG. “Operation”, “physical quantity” and the like may be considered as subclasses of class 22 “equipment / equipment” in FIG. As knowledge data stored in the storage unit 31, the classification of documents and devices may be expressed by one tree.
[0041]
FIG. 8 is a diagram showing a part of a search command generation rule for interpreting a natural language request sentence and searching and presenting data requested by a user. Each rule (here, four rules R1 to R4 are shown as an example) has eight items.
[0042]
The first item “pattern” is a check item for determining whether the user's request text matches each rule. Information that is important for interpreting the user's intention from the request text as well as judging whether to apply the rule from the similarity when the pattern described in this item is matched with the request text To extract. For example, the pattern of the rule R1 is a pattern that matches a request sentence such as “the temperature of the thermometer of the NO.1 reaction can”, “the water level of the can”, “the value of the thermometer and the water level meter”. The part designated by <> in the pattern means that the expression (for example, name) meaning the class or instance described in FIG. 5 appears in the request sentence. For example, “NO.1 reaction can” and “can” in the above example match the <equipment (X)> portion in the pattern of rule R1, and “thermometer” and “water level meter” are <sensor (Y )>.
[0043]
For the pattern notation method of FIG. 8, “*” mark matches zero or more repetitions, “+” mark matches one or more repetitions, and “|” This means that it matches the pattern (OR), and {} means grouping.
[0044]
In the pattern of the rule R3, the description <class (X)> means that the related words of the document class and the class instance of the plant facility described in FIGS. 2 to 7 are included in the request sentence. For example, when the request sentence “method of shutting down reaction can” is matched with the pattern of rule R3, the words “reaction can” and “shutdown” match <class (X)>. As a result, the request sentence is interpreted as a request for the “reaction can” of the equipment class from FIGS. 5 and 6 and the “unsteady operation” of the document class from FIGS. The pattern of rule R4 is “TRUE”, which matches any request statement.
[0045]
In FIG. 8, the second item “precondition” is an item describing a condition for applying the rule. For example, the precondition “controller_is_enable (CURRENT_STATE)” of the rules R1 and R2 means that the control equipment is available in the current plant state (CURRENT_STATE) in order to apply these rules. It is that. As will be described later, the rule R1 is a rule for acquiring the current value of the process data from the control device, and the rule R2 is a rule for issuing an operation command to the control device. Both rules are used by the control device. Applicable only when possible.
[0046]
FIG. 9 shows an example of the current plant status (CURRENT_STATE) for a part of the plant equipment. The process data acquisition unit 21 in FIG. 1 acquires data regarding the presence or absence and degree of abnormality of sensors and actuators in the plant equipment, including the state of the control device itself. For example, in FIG. 9, the tag is “CONTROLLER”. If the state of the equipment device is a serious failure, the above-mentioned precondition “controller_is_enable (CURRENT_STATE)” is false, and the rules R1 and R2 are not applied.
[0047]
Similarly, the precondition of the rule R3 in FIG. 8 is that “if the current state (CURRENT_STATE) is a serious failure (is_critical (CURRENT_STATE)), the request is related to monitoring / control (class X == monitoring / control). This rule only applies to The rule R3 is a rule for performing a document search. By controlling the applicability according to such a precondition, an inappropriate request is issued by the user regardless of the state of the plant, as will be described later. In case this can be rejected.
[0048]
On the other hand, the precondition of the rule R4 is true (TRUE), and this rule can be applied in any state.
[0049]
In FIG. 8, a third item “parameter” and a fourth item “command” are descriptions relating to a search instruction generated by each rule and a method of generating the parameter. The rule R1 is a rule for acquiring and presenting the current value of the process requested by the user, and its command is “get_value (Y. tag)”. Here, “Y” is a sensor instance (for example, “NO.1 upper limit”), and its tag is used as a parameter of the command “get_value”. The rule parameter item describes a procedure for specifying the value of the parameter. For rule R1,
bind ((X, Y, Z), (Y. equipment == X and Y. physical quantity == Z), CURRENT_STATE)
Thus, X, that is, “equipment”, Y, that is, “sensor”, and Z, that is, “physical quantity” are specified. The constraint condition for that is (Y.equipment == X and Y. physical quantity == Z), but this is a restriction meaning that the equipment of Y is X and the physical quantity of Y is Z. is there. For example, in response to an inquiry such as “the water level of the reaction can”, an instance Y of the sensor in which the equipment X is an instance of the “reaction can” class and the physical quantity Z is the “water level” class. Desired. When there are a large number of sensors Y that satisfy this restriction, as will be described later, the sensor Y that most likely matches the user's intention is selected based on the current process state CURRENT_STATE.
[0050]
In FIG. 8, items 5 to 8 are templates for explaining to the user the result of interpretation performed by the system and information that can be provided in response to the user's request.
[0051]
The fifth item “menu” is for briefly presenting information that the system can provide to the user. For example, when the user issues a request to “open the valve of the reaction can”, the rule R2 is the rule that most closely matches the request by performing pattern matching. For the menu item “Z operation of Y of X”, the equipment X, the actuator Y, and the operation Z are specified from the user's request sentence by the above-described processing, and these are set as the variables X, Y, and Z of the menu. As a result, the user is presented with a menu “Opening reaction can valve”.
[0052]
Similarly to the sixth item “response when rejected”, the seventh item “response when successful”, and the eighth item “response when failed”, a message template for explaining the status of interpretation of the system to the user. It is. “Rejection response” is a response sentence template for informing the user that the request is rejected when the above-mentioned preconditions are not satisfied. “Response at success” and “response at failure” are templates of response sentences that are output in accordance with the success / failure of the result of executing the command. In FIG. 8, “REQUEST” is a variable meaning a user's request text.
[0053]
The search command generation rules according to the present embodiment have been described above with examples. However, in the present invention, the contents and description methods of these knowledge / rules are not limited to the above methods. The present embodiment is mainly intended for information retrieval in plant monitoring and control. However, when the target field is different, the contents of the knowledge / rule and the description method are different. The essential part of the search command generation rule of the present invention is that it has knowledge (concept hierarchy etc.) and rules (pattern etc.) for interpreting the intention from the user's request sentence, and environmental information such as plant. It has a framework for controlling how requests are accepted / rejected and interpreted.
[0054]
Next, the information search processing operation of the information search system of FIG. 1 will be described with reference to the flowchart shown in FIG. In this case, FIGS. 11 to 15 show screen display examples of request sentences from the user and system responses.
[0055]
The system accepts a question from the user, and if there is a request sentence Q from the user (step S1), first, for each element Ri of the set R of all search command rules, the pattern of the request sentence Q and Ri The similarity is calculated, and this is set as the score Pi of Ri (step S2). Here, the Pi calculation method depends on the pattern description method. For example, it can be simply obtained by the following calculation formula.
[0056]
Pi = (li × ni) / (Lq × Ni)
Lq = character string length of request sentence Q
li = character string length of the request sentence Q that matches the Ri pattern
Number of concepts in Ni = Ri pattern (description such as <equipment (X)> above)
ni = number of occurrences of the concept specified by the Ri pattern in the request statement Q
For example, when Q = “how much the temperature of the reaction can”, the score P1 of the rule R1 in FIG. 8 for this requirement Q is 10 characters for Lq, and the value of l1 is “reaction can”, “ Six characters of “temperature”, Ni value in this case are <sensor (Y)> and <physical quantity (Z)>, and ni value is “reaction can” and “temperature” Therefore, P1 = (6 × 2) / (10 × 2) = 0.6. Here, Ni is counted in consideration of pattern omission and repetition. That is, in the rule R1, since “{<sensor (Y)>} *” portion allows 0 or more repetitions, Ni is not included in the request sentence Q when there is no description thereof. On the other hand, the portion of “{<equipment (X)>}” also allows zero or more repetitions, but since the description “reaction can” appears once in the request sentence Q, the number of both Ni and ni Included.
[0057]
Similar to rule R1, the scores P2 and P3 of rule R2 and rule R3 are P2 = (3 × 1) / (10 × 2) = 0.15 and P3 = (5 × 2) / (10 × 2), respectively. ) = 0.5.
[0058]
Since the rule R4 is a special rule whose pattern is “TRUE (constant)”, it is assumed that this score is always the threshold value Ps. Ps is a threshold value indicating whether or not the rule should be applied. Here, when the value is “0.3”, for example, P4 = 0.3.
[0059]
After calculating the score Pi of each rule Ri, if there is no Ri having a score Pi exceeding the threshold value Ps (step S3), a response sentence indicating that the request sentence Q could not be interpreted is presented (step S4). . FIG. 11 shows an example of the screen. However, if there is a rule in which the score does not always fall below the threshold as in rule R4, this step is not reached, and the process proceeds to step S5 and subsequent steps.
[0060]
In step S5, the current process value state S (indicated as “CURRENT_STATE” in FIG. 8 is the same information, an example of which is shown in FIG. 9) is acquired. The state S is used in the subsequent processes of step S6 and step S8.
[0061]
The process from step S6 to S14 is a process for generating a search command and executing a search. In FIG. 10, the processing from step S6 to S14 is repeatedly executed in the descending order of Pi for the rule obtained in step S2, that is, Ri for which the score Pi for the request sentence Q exceeds the threshold value P. This method may be modified as described later.
[0062]
First, in step S6, whether or not the rule Ri is applicable in the state S is determined based on the precondition of the rule Ri. For example, as described above with reference to FIG. 8, when the user's request sentence is a question “How to use a recreational facility”, and the state S is a state where a serious failure has occurred For example, the precondition of rule R3 (is_critical (CURRENT_STATE)? (X == monitoring / control): TRUE) is not satisfied. This is because “is_critical (CURRENT_STATE)” is true, but “X”, that is, “retreat” class (see FIG. 2) does not belong to the “monitoring / control” class (see FIG. 2) (see FIG. 2). In 8 etc., the expression (A? B: C) means that if A is true, the truth value of B is taken, and if A is false, the truth value of C is taken). Therefore, in this case, in step S7, a response sentence notifying that the request is rejected, that is, not applicable to the rule (for example, rule 3 here) is presented. FIG. 12 is a screen display example when the request is rejected.
[0063]
On the other hand, if the precondition is satisfied, the value of the parameter passed to the command is determined in step S8. The parameter value is determined in consideration of the current state S. As an example, as shown in FIG. 9, the current state S is a state in which a device with the tag “LS1001” (that is, “NO.1 upper limit”) has detected a serious failure, Consider a case in which the system is inquired with a request statement “method”. In this case, rule R3 matches with the highest score, and “recovery”, that is, “unsteady operation” belongs to “monitoring / control” (see FIG. 3), so the precondition of rule R3 is satisfied.
[0064]
In step S8, the values of the parameters X, REQUEST, and K of the command “get_text (X, REQUEST, K)” executed by the rule R3 are obtained. Here, the command “get_text” is a document search command, the parameter “X” is a document class “unsteady operation”, and “REQUEST” is a request statement itself, that is, a character string “recovery method”. “K” is an auxiliary keyword, and the “get_text” command searches a document of class X using a natural language query sentence REQUEST and a logical expression of the auxiliary keyword as a search condition. . The auxiliary keyword in this case is obtained by the parameter item of the rule R3, that is, “K = (X == monitoring / control? Get_keyword (CURRENT_STATE): NULL)”. That is, here, since “unsteady operation” belongs to “monitoring / control”, the keyword K is generated by a command “get_keyword (CURRENT_STATE)”, that is, processing for obtaining a keyword from the current state.
[0065]
Here, processing for obtaining a keyword from the current state of the plant will be described. For example, in the current state, it is assumed that an abnormality has occurred in the device with the tag “LS1001”. It can be seen from FIG. 7 that the device with the tag “LS1001” means the sensor “NO.1 upper limit”.
[0066]
7, the “NO.1 upper limit” class is “upper limit”, and the superclasses of “upper limit” from FIGS. 6 and 5 are “water level gauge”, “sensor”, and “equipment”. By tracing the superordinate concept in this way, a word related to “LS1001” which is currently focused on is obtained together with the degree of the relation. That is, the concept that is one step higher than the “NO.1 upper limit” instance is the “upper limit” class from FIG. 6, and the related terms are “upper limit”, “upper limit water level”, and “water level exceeded” from FIG. 6. Similarly, the upper level concept of the “upper limit” class is “water level meter” from FIG. 6, and the related terms are “water level meter”, “level meter”, and “liquid level meter” from FIG. Further, “Sensor” is a higher-level concept of “water level meter” from FIG. 6, and its related terms are “sensor”, “measuring instrument”, “measuring instrument”, and “instrument” from FIG. Furthermore, the concept of “Sensor” that is one step higher is “apparatus” from FIG. 6 (If the flow from the above “NO.1 upper limit limit” toward the higher concept is schematically shown, the NO.1 upper limit limit → the upper limit is set. Limit → Water level gauge → Sensor → Equipment).
[0067]
Now, “LS1001”, that is, the concept that is one level higher than the “NO.1 upper limit” instance is the “upper limit” class from FIG. 6, and its related terms are “upper limit”, “upper water level”, “ The water level was exceeded. Therefore, each of these related words is separated by morphemes, and a search condition is generated as a logical expression of AND (* mark) and OR (+ mark).
[0068]
In addition, each time a trace is made higher than the lowest “NO.1 upper limit”, the relevance is multiplied by ½, so that the search condition generated as a result of tracing the superordinate concept as described above. The logical expressions and their relevance are as follows.
[0069]
Relevance formula
1/2 (upper limit * limit + upper limit * water level + water level * excess)
1/4 (water level * total + level * total + liquid level * total)
1/8 (Sensor + Measurement * Instrument + Measurement * Instrument + Instrument)
Similarly, in the relationship other than the superordinate concept, that is, the relationship “physical quantity” that is measured by “NO.1 upper limit”, the physical quantity described in the “NO.1 upper limit” instance from FIG. “Water level” and its related terms are “water level”, “liquid level” and “level” from FIG. Therefore, if the degree of relevance is multiplied by 1/2 each time it goes higher than the lowest “NO.1 upper limit”, it is generated as a result of following the relationship of “physical quantity” as described above. The logical expression of the search condition and its relevance are as follows.
[0070]
Relevance formula
1/2 (water level + liquid level + level)
Further, in the relation of “equipment” having “NO.1 upper limit” as a component, “NO.1 reaction can” is shown in FIG. From FIG. 7, it is the “reaction can” class, and the related terms are “reaction tank” and “reaction tank” although not shown in FIG. Further, although the upper-level component of the “reaction can” is not shown in FIG. 6, as can be clearly seen from FIG. 5, it is “can” and the related terms are “tank” and “tank”. . Furthermore, although the upper-level component of “can” is not shown in FIG. 6, as is clear from FIG. 5, it is “equipment” (the component moving upward from the above “NO.1 upper limit”). The flow of NO.1 is schematically shown as NO.1 upper limit → NO.1 reaction can → reaction can → can → equipment). Therefore, if the degree of relevance is multiplied by ½ each time the lowermost “NO.1 upper limit” is traced, it is generated as a result of following the relationship of “equipment” as described above. The logical expression of the search condition and its relevance are as follows.
[0071]
Relevance formula
1/4 (reaction * can + reaction * tank + reaction * tank)
1/8 (can + tank + tank)
The sum of these logical expressions is used as an auxiliary search condition K for the command “get_text”.
[0072]
Next, in step S9, it is determined whether or not the command is to be executed, and if so, the process proceeds to step S10. Whether to execute the command includes a method of automatically executing based on the score Pi of the rule Ri, and a method of explicitly selecting whether the user executes the command. FIG. 13 is an example of the latter, which is an example of a screen in which the menu of each rule (see the explanation of FIG. 8) is presented to the user in order from the highest score, and the user selects a command to be executed. Of these, a case is shown in which a search result is displayed when “temperature of reaction vessel thermometer” is selected by the user. In this case, the processing flow of FIG. That is, a process for creating and presenting a menu after step S8 is added, and after presenting the menu for all the rules of interest, the user is asked to determine whether to execute the menu. The command is executed by sending the command and parameter values obtained in step S8 to the document data management unit 1 or the process data management unit 2 and executing a search (step S10). For example, in the example of the rule R3 in FIG. 8, a command “get_text” and its parameters “REQUEST”, “X”, and “K” are sent to the document data management unit 1 to search for a document.
[0073]
The document data search unit 13 out of the document data belonging to the document class “X” (“unsteady operation” in the above example) matches the natural language search condition “REQUEST” (for example, “recovery method”). Search for. A document search method using a natural language as a search condition is performed in the same manner as in the past. At this time, the word “recovery” in “REQUEST” is a document class “unsteady operation” as shown in FIG. In the context of “return” and “abnormal processing”, these words are also used as search conditions.
[0074]
The auxiliary keyword search conditions described above are adapted to the search result of the natural language search conditions, and are used to narrow down and search for documents that match the plant state.
[0075]
As described above, if the result of command execution such as document search is successful (step S11), a response message indicating success and the execution result are presented to the user (step S12). FIG. 14 is an example of a screen presenting a list of retrieved documents as a command execution result. As described above, as a result of adding a document search condition based on the state of the plant to the user's request sentence “recovery method”, an operation plan regarding a recovery method such as “exceeding the water level” is searched and the user is searched. Presented.
[0076]
As described above, according to the present invention, it is possible to interpret a user's simple and ambiguous request sentence based on the state of the process data, and to present information suitable for the user's intention. In addition, as shown in FIG. 14, the processing performed by the system according to the response sentence “Recovery method (upper limit, limit, water level, excess, ...)” is displayed. (In this example, the document was searched) and how the request was interpreted based on the state of the process data (in this example, it was interpreted based on the situation such as “water level” “excess”) Made. With this function, the user can easily know whether the system has interpreted as intended, and can obtain only the information he wants without executing a command due to incorrect interpretation of the system. .
[0077]
As another example, the flow of processing when the user issues a request to “open the valve” will be described. For this request statement, in step S2, rule R2 is chosen as the rule with the highest score, and if this is applicable in step S6 (ie if the control device is available), in step S8, “set_value” The values of the parameters Y and Z of the command (Y.tag, Z.signal ") are generated. This command is a command for setting a Z signal to the Y actuator. Here, when there are a large number of valves in the plant, which valve should be opened cannot be specified from the request statement. At this time, if the current state of the plant is a state where the device with the tag “TI1401” is abnormal (that is, “NO.1 in-can thermometer” indicates an abnormal value), the system It is determined that the valve related to the abnormal state is a valve intended by the user.
[0078]
The procedure will be specifically described below. First, it can be seen from FIG. 7 that the device with the tag “TI1401” is a tag indicating “NO.1 can thermometer”. From FIG. 7, the equipment to which “NO.1 can thermometer” belongs is “NO.1 reaction can”, and “NO.1 reaction can” is “NO.1 injection valve A” “NO. .1 injection valve B, “NO.1 discharge valve”, and “NO.1 cooling water valve” are found to have four valves (the above first NO.1 can thermometer is schematically shown in the first relation) As shown, NO.1 can thermometer → NO.1 reaction can → NO.1 injection valve A, NO.1 injection valve B, NO.1 discharge valve, NO.1 cooling water valve).
[0079]
On the other hand, FIG. 7 also shows that the “NO. 1 can temperature controller” adjusts the “NO. 1 cooling water valve” by using the measured value of the “NO. The second relationship with the NO.1 can thermometer is schematically shown as follows: NO.1 can thermometer → NO.1 can temperature controller → NO.1 cooling water valve).
[0080]
The “NO.1 in-can thermometer” and the “NO.1 cooling water valve” have a two-way relationship as in the first and second relationships, and the number of stages following the relationship is both Since both are two stages, the degree of association between “NO.1 in-can thermometer” and “NO.1 cooling water valve” is 1/4 + 1/4 = 1/2. On the other hand, the other three valves “NO.1 injection valve A”, “NO.1 injection valve B”, and “NO.1 discharge valve” all have only the first relationship. The degree of relevance between these three valves is ¼.
[0081]
Based on the above calculation, it is determined in step S8 that “Y” of the command “set_value (Y. tag, Z. signal)” is highly likely “NO.1 cooling water valve”, and the tag “ MV1101 "is used as a parameter (see FIG. 7). Since Z is an “open” operation in response to a user request, “Z. signal” is “1” from FIG. As a result, the command “set_value (MV1101, 1)” is executed in step S10, and the result is presented to the user in step S12 or step S13.
[0082]
In addition, since the process which determines the valve of operation object from the state of the plant demonstrated in the above example cannot always determine the object which a user intends correctly, as shown in FIG. May be presented and selected / confirmed. In this case, a process to be selected by the user is added as a part of the parameter generation process (step S8). The probability of which candidate matches the user's intention among the candidates to be selected by the user can be determined based on the state of the plant as described above. In the above example, of the four valves, the “NO.1 cooling water valve” is most likely. Therefore, as shown in FIG. 15, the selection effort by the user can be reduced by presenting the most likely candidate as an upper candidate at the head or by presenting it with emphasis.
[0083]
As described above, according to the above-described embodiment, retrieval and use of document data and process data in operations such as plant monitoring control can be performed without effort by simple natural language input. In addition, it is possible to control whether or not to accept a user's request depending on the state of the process data of equipment such as a plant, and to increase the accuracy of the interpretation of the request. Appropriate data can be retrieved and presented even for various request sentences. In addition, if the request is interpreted based on the state of the process data, the explanation of why it was interpreted and prioritization of the presentation of candidates when there are multiple interpretations are performed, so the user can Can be easily understood, and the request-response interaction between the user and the system can be performed smoothly.
[0084]
As a result of the effects described above, the user is freed from the workload of searching for his / her necessary data from a large amount of data, and can concentrate on work such as monitoring and control. Safety also increases.
[0085]
In the above embodiment, an example of application to monitoring control of a plant or the like has been described. However, the application target of the present invention is not limited to a plant. It is clear that it can be applied to various tasks using document data and real-time data for transportation systems, intelligent buildings, automated warehouses, etc.
[0086]
In addition to the industrial field, for example, it can be applied in the service industry, the financial industry, the medical field, and the agriculture, forestry and fishery industry. In each of these cases, the real-time data called process data in the above embodiment should be replaced with various types of real-time data such as product inventory, reservation status data, securities price data, patient data, and weather data. Basically, it is possible to realize efficient information retrieval and use in the same manner as in the above embodiment.
[0087]
【The invention's effect】
As described above, according to the present invention, document data and real-time data can be efficiently searched. For example, information access in operations such as monitoring and control of facilities such as plants is reduced, and operations are made more efficient. Can be achieved.
[0088]
In addition, it is possible to efficiently search for appropriate information by simply inputting a simple search request sentence and interpreting it based on real-time data.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an information search system according to an embodiment of the present invention.
FIG. 2 is a diagram showing an example of a class of document classification, which is knowledge for search command generation.
FIG. 3 is a diagram showing a detailed description example of a document class.
FIG. 4 is a diagram showing a configuration example of a part of a plant.
FIG. 5 is a diagram showing an example of plant classification / configuration, which is knowledge for generating a search command.
6 is a diagram showing an example of a detailed description of each class including the upper / lower relationship between classes and the relationship between concepts shown in FIG. 5. FIG.
FIG. 7 is a diagram showing an example of detailed description of each instance including the relationship between the concepts of the instances shown in FIG. 5;
FIG. 8 is a diagram showing an example of a search command generation rule that is knowledge for search command generation.
FIG. 9 is a diagram showing an example of a current state (CURRENT_STATE) of a plant for a part of plant equipment.
10 is a flowchart showing the flow of processing of the information search system of FIG.
FIG. 11 is a diagram showing a screen example of a system response to a user request, showing a screen display example when a user request sentence cannot be interpreted.
FIG. 12 is a diagram showing a screen example of a system response to a user request, and shows a screen display example when a request sentence is rejected because it cannot be applied to a rule.
FIG. 13 is a diagram showing a screen example of a system response to a user request, and presents a menu of each rule in order from the highest score to the user, and causes the user to select a command to be executed. Is shown.
FIG. 14 is a diagram showing a screen example of a system response to a user request, showing a screen display example presenting a list of documents retrieved as a command execution result.
FIG. 15 is a diagram showing a screen example of a system response to a user request, and shows a screen display example for presenting to the user the contents obtained by interpreting the user request based on the process data, and for selecting and confirming the content. Is shown.
[Explanation of symbols]
1. Document data management department
11: Document data storage unit
12 ... Document data registration section
13 ... Document data search section
2. Process data management department
21 ... Process data acquisition unit
22 ... Process data storage unit
23 ... Process data search section
3 ... Search command generator
31 ... Search command generation rule storage unit
4. Data presentation part

Claims (9)

An acquisition means for acquiring facility / equipment data output from each facility / equipment;
First storage means for storing document data and the equipment / equipment data;
Second data storing knowledge data indicating document data and a plurality of classes for classifying each facility / equipment, upper / lower relations of the plurality of classes, words belonging to each class, and names of facilities / equipment Storage means;
A search instruction for searching for a pattern included in a natural language search request sentence including at least one of the plurality of classes, a word belonging to each class, and a facility / device name, and document data corresponding to the pattern And a plurality of generation rules each including a description regarding a method for obtaining a parameter necessary for executing the search instruction, and any one of the search instructions for searching for facility / equipment data A third storage means;
An input means for inputting a search request sentence in natural language;
Generating means for generating a search command and a parameter based on the search request sentence input by the input means;
Search means for searching document data and facility / equipment data stored in the first storage means using the search command and the parameter;
Presentation means;
In an information search method in an information search device comprising:
An input step in which the input means inputs the search request sentence;
The generation means refers to the knowledge data, and acquires a class to which a word and a facility / equipment name included in the input search request sentence belong, and a higher class of the class; and
The generation means uses the words and equipment / device names included in the input search request sentence, and the class acquired in the first acquisition step, the patterns of the plurality of generation rules, and the input A calculation step of calculating a similarity between the search request sentence and
A selection step in which the generation unit selects a generation rule having a similarity pattern equal to or higher than a predetermined value applicable to the input search request sentence from the plurality of generation rules;
Using the description included in the selected generation rule, the generation unit includes at least one of the input search request sentence and the knowledge data, and the facility / equipment data acquired by the acquisition unit. A second obtaining step for obtaining the parameters;
A search step in which the search means performs a search using the search command included in the generation rule obtained by the generation means and the parameter obtained by the generation means;
The presenting means presenting a search result in the search means;
Information retrieval method including The generation rule includes a condition for applying the generation rule to the search request statement,
The selecting step includes selecting a generation rule including a pattern having a similarity equal to or higher than a predetermined value applicable to the input search request sentence and satisfying the condition from the plurality of generation rules. The information search method according to claim 1.   The information search method according to claim 2, wherein the condition is a condition relating to at least one of the words in the facility / equipment data and the search request sentence.   2. The information retrieval method according to claim 1, wherein the presenting means presents information regarding a generation rule and information regarding a parameter obtained by the generating means for the input search request sentence. First acquisition means for acquiring facility / equipment data output from each facility / equipment;
First storage means for storing document data and the equipment / equipment data;
Second data storing knowledge data indicating document data and a plurality of classes for classifying each facility / equipment, upper / lower relations of the plurality of classes, words belonging to each class, and names of facilities / equipment Storage means;
A search for searching for a pattern included in a natural language search request sentence including at least one of the plurality of classes, a word belonging to each class, and a facility / device name, and document data corresponding to the pattern. A plurality of generation rules each including one of a command and a search command for searching facility / equipment data and a description relating to a method for acquiring a parameter necessary for executing the search command are stored. Third storage means for
An input means for inputting a search request sentence in natural language;
A second acquisition unit that refers to the knowledge data, acquires a class to which a word and a facility / device name included in the search request sentence input by the input unit, and a higher class of the class;
Using the words and equipment / device names included in the search request text input by the input means and the class acquired by the acquisition means, each pattern of the plurality of generation rules, and input by the input means A calculation means for calculating the similarity between the search request text and
A selection unit that selects a generation rule having a similarity pattern equal to or higher than a predetermined value applicable to the search request sentence input by the input unit from the plurality of generation rules;
Using the description included in the generation rule selected by the selection means, from at least one of the search request sentence and the knowledge data, and the equipment / equipment data acquired by the first acquisition means, Third acquisition means for acquiring the parameter;
Document data and facility / equipment data stored in the first storage unit using the search command included in the generation rule selected by the selection unit and the parameter acquired by the third acquisition unit A search means for searching for,
Means for presenting search results in the search means;
An information retrieval apparatus comprising: The generation rule includes a condition for applying the generation rule to the search request statement,
The selection unit includes a generation rule including a pattern having a similarity equal to or higher than a predetermined value applicable to the search request sentence input by the input unit, and satisfying the condition, from the plurality of generation rules. 6. The information search apparatus according to claim 5, wherein the information search apparatus is selected.   The information search apparatus according to claim 6, wherein the condition is a condition related to at least one of the words in the facility / equipment data and the search request sentence.   6. The information according to claim 5, wherein information relating to the generation rule selected by the selecting means and information relating to the parameters obtained by the third obtaining means are presented for the search request sentence input by the input means. Search device. Computer
First acquisition means for acquiring facility / equipment data output from each facility / equipment;
First storage means for storing document data and the equipment / equipment data;
Second data storing knowledge data indicating document data and a plurality of classes for classifying each facility / equipment, upper / lower relations of the plurality of classes, words belonging to each class, and names of facilities / equipment Storage means;
A search for searching for a pattern included in a natural language search request sentence including at least one of the plurality of classes, a word belonging to each class, and a facility / device name, and document data corresponding to the pattern. A plurality of generation rules each including one of a command and a search command for searching facility / equipment data and a description relating to a method for acquiring a parameter necessary for executing the search command are stored. Third storage means for
An input means for inputting a search request sentence in natural language;
A second acquisition unit that refers to the knowledge data, acquires a class to which a word and a facility / device name included in the search request sentence input by the input unit, and a higher class of the class;
Using the words and equipment / device names included in the search request text input by the input means and the class acquired by the acquisition means, each pattern of the plurality of generation rules, and input by the input means A calculation means for calculating the similarity between the search request text and
A selection unit that selects a generation rule having a similarity pattern equal to or higher than a predetermined value applicable to the search request sentence input by the input unit from the plurality of generation rules;
Using the description included in the generation rule selected by the selection means, from at least one of the search request sentence and the knowledge data, and the equipment / equipment data acquired by the first acquisition means, Third acquisition means for acquiring the parameter;
Document data and facility / equipment data stored in the first storage unit using the search command included in the generation rule selected by the selection unit and the parameter acquired by the third acquisition unit A search means for searching for,
Means for presenting search results in the search means;
A machine-readable recording medium on which a program for causing the machine to function is recorded.

Images