Data acquisition method, computer device, and data acquisition system

Abstract

Provided is a data acquisition method in which a language model can generate an accurate answer. This data acquisition method executed by one or more computer devices connected to one or more language models comprises: generating, by a computer device, a prompt for identifying a database including information related to a question content of a user; inputting, by the computer device, the prompt to a language model; and receiving, by the computer device and from the language model, at least one of the location of the database, the name of the database, the format of the database, and information of a column of the database.

Description

Data acquisition method, computer device, and data acquisition system

[0001] The present invention generally relates to techniques for acquiring data using language models.

[0002] In recent years, with the development of language models, many dialogue systems have been provided that generate answers by referring to appropriate sentences when a user inputs a question.Language models are trained on large amounts of data through a process called pre-training, but they are not trained on company-specific data.Therefore, when a question about a company's proprietary data is asked, there is a risk that the model will induce plausible lies (hereinafter referred to as hallucination).

[0003] As a technique for suppressing hallucination, for example, Patent Document 1 discloses the following: "The method includes a step of generating, for a given document 301, at least one query data 302 that can be answered by the document using a language model, and a step of using data consisting of pairs (Document, Query) of each of the documents belonging to a specific domain and the query data generated for the document to train a search model 320 for a conversational bot." In other words, by inputting company-specific data into a language model in advance, generating anticipated questions and their answers, and storing them in a database, answers to questions during operation can be searched for from sets of anticipated questions and answers and obtained as reference information, thereby enabling accurate answers to questions to be generated.

[0004] JP 2023-76413 A

[0005] In the method of providing dialogue described in Patent Document 1, a large amount of data is input into a language model in advance, but creating anticipated questions and their answers and storing them in a database not only requires a considerable amount of time and financial cost, but also has the potential to induce hallucination if a question other than the anticipated question is input during operation. Furthermore, when the content of the question must be comprehensively judged by referring to multiple databases, the content of the anticipated question may be insufficient.

[0006] The present invention has been made in consideration of the above points, and aims to propose a data acquisition method etc. that enables a language model to generate an accurate answer.

[0007] In order to solve this problem, the present invention provides a data acquisition method executed by one or more computer devices connected to one or more language models, which includes the steps of: generating a prompt for identifying a database containing information related to the content of a user's question by the computer device; inputting the prompt into the language model by the computer device; and receiving at least one of the location of the database, the name of the database, the format of the database, and information on columns of the database from the language model as an answer to the prompt.

[0008] In the above configuration, the computer device receives, as an answer to the prompt, at least one of the location, name, format, and column of a database containing information related to the question, thereby narrowing down existing databases containing information related to the question. For example, with the above configuration, the computer device can obtain information related to the question from existing databases, thereby providing an accurate answer without having to build a dedicated database.

[0009] According to the present invention, it is possible to realize a data acquisition method that enables a language model to generate an accurate answer. Problems, configurations, and effects other than those described above will become apparent from the following description of the embodiment.

[0010] FIG. 1 is an explanatory diagram relating to a network according to a first embodiment. FIG. 2 is an explanatory diagram relating to a computer device according to a first embodiment. FIG. 3 is a diagram showing an example of a flowchart relating to preprocessing according to a first embodiment. FIG. 4 is a diagram showing an example of a database configuration list according to a first embodiment. FIG. 5 is a diagram showing an example of a prompt according to a first embodiment. FIG. 6 is a diagram showing an example of a flowchart relating to data acquisition processing according to a first embodiment. FIG. 7 is a diagram showing an example of a user interface according to a first embodiment. FIG. 8 is a diagram showing an example of a flowchart relating to data acquisition processing according to a second embodiment. FIG. 9 is a diagram showing an example of a prompt according to a second embodiment. FIG. 10 is a diagram showing an example of a user interface according to a second embodiment.

[0011] (I) First Embodiment One embodiment of the present invention will be described in detail below, although the present invention is not limited to this embodiment.

[0012] In this embodiment, we will explain a technology for providing services using language models, which generates accurate answers while suppressing hallucination when generating answers to user questions based on a structured data group (hereinafter referred to as a database).

[0013] The following describes a data acquisition method for acquiring data from a database for a language model to generate an accurate answer. The data acquisition method is executed by at least one computer device. The computer device includes at least one memory and a processor, the memory storing computer-readable instructions, and the processor is configured to execute the instructions stored in the memory. In the data acquisition method, the at least one processor generates a prompt that serves as an input sentence for the language model, inputs it to the language model, and receives from the language model at least one of the location, format, and column of a database containing data related to the user's question.

[0014] Next, an embodiment of the present invention will be described with reference to the drawings. The following description and drawings are examples for explaining the present invention, and appropriate omissions and simplifications have been made for clarity of explanation. The present invention can be implemented in various other forms. Unless otherwise specified, each component may be singular or plural. In the following description, the same elements in the drawings will be denoted by the same numbers, and their description will be omitted as appropriate.

[0015] The designations "first," "second," "third," etc. in this specification are used to identify components and do not necessarily limit the number or order. Furthermore, numbers used to identify components are used in different contexts, and numbers used in one context do not necessarily indicate the same configuration in another context. Furthermore, this does not prevent a component identified by a certain number from also serving the function of a component identified by another number.

[0016] A computer system according to this embodiment will be described with reference to Figures 1 to 7. In Figure 1, reference numeral 100 denotes a network according to the first embodiment.

[0017] [Network 100] Figure 1 is an explanatory diagram of a network 100. In the network 100, all devices 101, 102, 103, 104, 105, and 106 (hereinafter referred to as devices) communicate bidirectionally via communication means. Examples of the network 100 include a local area network (LAN) and a personal area network (PAN). The communication means include wireless communication, wired communication, etc.

[0018] The device includes a storage device that stores the database. Examples of the device include a personal computer (PC), a notebook PC (e.g., a laptop computer), a cloud-based database, a data server, etc. Note that the device and communication means are not limited to those exemplified above.

[0019] 2 is an explanatory diagram of the computer device 200. In the data acquisition system of this embodiment, one or more computer devices 200 and one or more language models 210 are connected via a network 100. Note that devices are not shown in the figure because they may not be provided in some cases.

[0020] The computer device 200 receives a question from a user, generates a prompt based on the received question, provides the generated prompt to the language model 210, and obtains an answer from the language model 210. The language model 210 is, for example, an LLM (Large Language Model). The language model 210 generates an answer based on configuration information (hereinafter referred to as a database configuration list) that indicates the configuration of either or both of a database stored in a device connected to the network 100 and a database stored in the computer device 200, and returns the answer to the computer device 200.

[0021] The computer device 200 is operated by an input / output device 201. The input / output device 201 operates an input / output interface 202. The input / output interface 202 generates a screen on which, for example, a question to the language model 210 and the answer are displayed. When a question to the language model 210 is executed, the processor 203 operates to store, for example, the question content, processing content, etc. in the memory 204. It is also assumed that the processor 203 accesses the external network 100.

[0022] The functions (processing unit, etc.) of the computer device 200 may be realized, for example, by the processor 203 reading a program stored in an auxiliary storage device into the memory 204 and executing it (software), or by hardware such as a dedicated circuit, or by a combination of software and hardware. Note that one function of the computer device 200 may be divided into multiple functions, or multiple functions may be combined into one function. For example, the processing unit may be divided into a first generating unit, a first input unit, a first receiving unit, an acquiring unit, a second generating unit, a second input unit, a second receiving unit, etc. Furthermore, some of the functions of the computer device 200 may be provided as separate functions or may be included in other functions. Furthermore, some of the functions of the computer device 200 may be realized by another computer capable of communicating with the computer device 200. Furthermore, each hardware component of the computer device 200 may be one or more.

[0023] [Preprocessing] FIG. 3 shows functional blocks related to processing (preprocessing) before the data acquisition processing of this embodiment is performed, and is a diagram showing an example of a flowchart related to the preprocessing.

[0024] In step S300, the processing unit of computer device 200 lists the paths of devices connected to network 100. The device paths include the IP addresses (Internet Protocol Addresses) of devices present on network 100, URLs (Uniform Resource Locators) of SQL (Structured Query Language) databases, and FTP (File Transfer Protocol) paths. After the processing unit has listed the paths of all devices on network 100, the process proceeds to step S301.

[0025] In step S301, the processing unit determines whether the number of paths of devices on the network 100 acquired in step S300 is greater than 0. If the processing unit determines that the number is greater than 0, it proceeds to step S302, and if the processing unit determines that the number is 0, it proceeds to step S305.

[0026] In step S302, the processing unit searches for databases present in the devices on the network 100 that were listed in step S300. If the location of the database has already been manually specified, steps S300 to S303 may be skipped and processing may begin with step S304.

[0027] In step S303, the processing unit lists the database paths of devices on the network 100. For example, the database path is the table name in the case of an SQL database, the relative or absolute path of the CSV (Comma Separated Values) file in the case of CSV data, and the relative or absolute path of the database within the FTP file in the case of FTP data. The database path may be obtained by automated search using a programming language, or the path may be specified manually. By changing the path format depending on the type of database, any type of database can be targeted for data acquisition. When all database paths have been listed, the processing unit proceeds to step S304.

[0028] In step S304, the processing unit lists the columns of the database of the devices on the network 100. A column of the database is a string of characters used to describe each piece of data in the database. By obtaining the columns, it is possible to confirm the structure and meaning of the database.

[0029] For example, if the database is SQL, the table name is customer_Complint (customer complaint), and the column names are "customer_id," "complaint_content," and "Product_name," it is easy to infer from the columns that the database contains the complaint content collected from customers, and that the complaint content and the target product are recorded along with the customer ID.

[0030] From the above viewpoint, in step S304, the processing unit acquires a list of columns by linking them to the path of the database. In step S304, the columns may be manually specified in advance by linking them to the database, or may be automatically acquired by a program.

[0031] In step S305, the processing unit creates a list of paths to databases in the storage device (hereinafter, on the local device) of the computer device 200. In this embodiment, it is assumed that databases on the local device may be accessed, not just those on the network 100.

[0032] In step S306, the processing unit determines whether the number of device paths on the local device is greater than 0. If the processing unit determines that the number is greater than 0, the processing proceeds to step S307, and if the processing unit determines that the number is 0, the processing proceeds to step S308.

[0033] In step S307, the processing unit lists the columns of the database on the local device, which may be any database, similar to the database of the device on the network 100.

[0034] In step S308, the processing unit stores the configuration list of all databases on the devices on the network 100 and the local device in the memory 204. At the same time as storing the list in the memory 204, the processing unit may store the list in the database as an intermediate file.

[0035] According to the preprocessing of this embodiment, it is possible to clarify the location of a database in either a database on a local device or a database on a device on network 100, and by obtaining the location of the database as well as the columns as a simple structure of the database, it is possible to obtain the meaning of the database.

[0036] [Database Configuration List 400] Figure 4 shows an example of a database configuration list (database configuration list 400). The database configuration list 400 shows a list in JSON format, but other formats can also be included in the present invention as long as they describe the database configuration. The database configuration list 400 includes a "path" that indicates the database path, a "table_name" that indicates the table name, a "format" that indicates the data format, and a "column" that indicates the column name.

[0037] By listing the database path, an retrieval method is provided for accessing the database later. The table name indicates the name of the database. Examples include customer_complaint (customer complaint) and purchase_history (purchase history). The data format indicates the type of database. More specifically, examples of the database format include the database extension and the type of database structure (CSV, SQL, PDF, etc.). The column name indicates the name of the column in the database. With regard to the column name, all column names of the database may be listed, or certain column names may not be listed for confidentiality reasons.

[0038] [Prompt 500] Figure 5 shows an example of a prompt (prompt 500). A prompt is an instruction statement (input statement) given to the language model 210. The language model 210 receives the prompt 500 and performs processing based on the instructions specified in the prompt 500. The prompt 500 is broadly divided into a command statement 501, a database configuration 502, and a question 503.

[0039] In this embodiment, command statement 501 describes, for example, an instruction such as "Please tell me the database configuration that should be referenced to answer the question from the database configuration below." This command statement 501 makes it possible to output the database configuration that should be referenced in response to the question in question content 503, which will be described later.

[0040] The database configuration 502 describes a database configuration list such as that shown in Fig. 4. The computer device 200 may provide the entire database configuration list to the language model 210, or may reconstruct the database configuration list based on the contents of the database carefully selected by an administrator and provide it to the language model 210.

[0041] A question related to the present embodiment is entered in the question content 503. For example, a question such as "Which database should I refer to in order to refer to the complaint content of customer ID 1234?" is conceivable.

[0042] [Data Acquisition Processing] Fig. 6 shows functional blocks related to the data acquisition processing of this embodiment and is a diagram showing an example of a flowchart related to the data acquisition processing. Note that the functional blocks show an example and are not limited to the functions described in this embodiment.

[0043] In step S600, the processing unit of the computer device 200 reads a database configuration list (e.g., database configuration list 400) from memory 204. At this time, the processing unit may read an intermediate file stored on a local device instead of from memory 204, or may read a database configuration list designated by an administrator instead of a configuration list of all databases.

[0044] In step S601, the processing unit inputs the content of a user's question. A possible question is, for example, "Which database should I refer to in order to refer to the content of the complaint for customer ID 1234?" as mentioned above, but the question is not limited to the above.

[0045] In step S602, the processing unit generates a prompt (for example, prompt 500) from the database configuration list and the question content.

[0046] In step S603, the processing unit inputs a prompt to the language model 210.

[0047] In step S604, the processing unit receives the configuration of the database to be referenced as a response from the language model 210 and outputs it via the input / output interface 202.

[0048] [Screen 700] Figure 7 shows an example of a user interface (screen 700). Screen 700 includes an input section 701 and an output section 702, and the output content changes depending on the input content. In response to the input content "Which database should I refer to to view the complaint content for customer ID 1234?", the output content displayed is {"path":"sql: / / XXXX.YYY.ZZZ", "table_name":"customer_complaint", "format":"sql", "column":["customer_id", "complaint_content", "product_name"]}. When viewing the complaint content, this output content suggests to the user which database from all databases is likely to contain the complaint content, based on the table name and column name.

[0049] In this embodiment, the data structure is output, but by asking, "Which database should I refer to to view the complaint details for customer ID 1234? Please output in SQL text," the output content of { SELECT complaint_content FROM customer_complaint WHERE customer_id=1234;} is displayed. This output content makes it possible to specify the desired database and customer ID to obtain pinpoint data. In this embodiment, the SQL database structure and SQL text have been used as examples, but CSV and FTP also work in a similar way, making it possible to accurately obtain desired data from a database.

[0050] [Behavior during actual work] As explained in detail above, according to this embodiment, it is possible to present a means of acquiring a database that can be inferred from the content of the question asked by the questioner, and not only can it acquire the location of a database that the questioner is not aware of, but it can also clearly show how to acquire and check the data, thereby enabling accurate information gathering and data understanding.

[0051] (II) Second Embodiment A computer device 200 according to a second embodiment will be described with reference to Figures 1 to 5 and 8 to 10. In this embodiment, the same components as those in the first embodiment (Figures 1 to 5) are designated by the same reference numerals and will not be described again, and the following mainly describes the components that differ from those in the first embodiment (Figures 8 to 10).

[0052] [Data Acquisition Processing] Fig. 8 shows functional blocks related to the data acquisition processing of this embodiment and is a diagram showing an example of a flowchart related to the data acquisition processing. Note that the functional blocks show an example and are not limited to the functions described in this embodiment.

[0053] In step S800, the processing unit of the computer device 200 reads a database configuration list (e.g., database configuration list 400) from the memory 204. At this time, the processing unit may read an intermediate file stored on a local device instead of from the memory 204, or may read a database configuration list designated by an administrator instead of a configuration list of all databases.

[0054] In step S801, the processing unit inputs a question from the user. A possible question is, "Please tell me the details of the complaint made by customer ID 1234," but the question is not limited to this.

[0055] In step S802, the processing unit generates a prompt (for example, prompt 500) from the database configuration list and the question content.

[0056] In step S803, the processing unit inputs a prompt to the language model 210 (first language model).

[0057] In step S804, the processing unit outputs the structure of the database to be referenced and the acquisition method as output from the language model 210.

[0058] In step S805, the processing unit reads the database based on the database configuration and the acquisition method. For example, in the case of SQL, the SQL statement { SELECT complaint_content FROM customer_complaint WHERE customer_id=1234;} is output as the acquisition method, and in step S805, the processing unit executes a program based on the SQL statement. Although not described in detail, even in cases other than SQL, the execution is performed by another program, and data is acquired by an appropriate program based on the data format. In addition, the content acquired from the database is not limited to pinpointed data acquired by specifying identification information such as a customer ID, but may also be data acquired within a certain range by specifying keywords such as product names.

[0059] In step S806, the processing unit generates a prompt (for example, prompt 900 described below) using the contents of the database acquired in step S805 as reference information together with the question content, and inputs the generated prompt to the language model 210 (second language model). At this time, the reference information may be input to the database information 902 in the prompt 900, or may be input to either the command statement 901 or the question content 903. Furthermore, the first language model used in step S803 and the second language model used in step S806 may be the same language model or different language models.

[0060] In step S807, the processing unit outputs the answer (output result) from the language model 210 via the input / output interface 202.

[0061] [Prompt 900] Figure 9 shows an example of a prompt (prompt 900). Command statement 901 contains an instruction such as, for example, "Please understand the question and the database information and generate an answer." Database information 902 contains some or all of the database contents acquired in step S805 as reference information. Question content 903 contains the question, "Please report the details of the complaint for customer ID 1234."

[0062] [Screen 1000] FIG. 10 shows an example of a user interface (screen 1000). Screen 1000 includes an input section 1001 and an output section 1002, and the output content changes depending on the input content. In response to the input content "Please report the details of the complaint for customer ID 1234," the output content displayed is "The details of the complaint for customer ID 1234 are as follows. I recently purchased your product P-25, but it won't turn on. I think it's a manufacturing defect, so I would like to have it replaced." This output content can be obtained by referencing the complaint content, retrieving from a database the contents of a database in which the complaint content is likely written, retrieving from the database contents the contents of a database linked to the customer ID and product name from the database contents, and finally outputting the contents retrieved from the database by language model 210.

[0063] [Behavior during actual work] As described in detail above, according to this embodiment, it is possible to generate an accurate answer from the question asked by the questioner, and not only can it understand the intention of the question asked by the questioner and obtain data in the database, but it can also obtain an accurate database according to the question content, thereby suppressing hallucination of the language model.

[0064] (III) Supplementary Notes The above-described embodiment includes, for example, the following contents.

[0065] In the above-described embodiment, the present invention has been described as being applied to a data acquisition method, but the present invention is not limited to this and can be widely applied to various other systems, devices, methods, and programs. For example, a program according to the present invention may be a program that causes a computer device connected to one or more language models to execute the following processes: generate a prompt for identifying a database containing information related to the content of a user's question; input the prompt into the language model; and receive at least one of the location of the database, the name of the database, the format of the database, and a column of the database from the language model as a response to the prompt.

[0066] In the above-described embodiments, some or all of the programs may be installed from a program source into a device, such as a computer, that implements the computer system. The program source may be, for example, a program distribution server connected via a network or a computer-readable recording medium (e.g., a non-transitory recording medium). In the above description, two or more programs may be implemented as a single program, or one program may be implemented as two or more programs.

[0067] Furthermore, in the above-described embodiment, the screens shown and described are merely examples, and any design may be used as long as the information received is the same.

[0068] Furthermore, the screens shown and described in the above-described embodiment are merely examples, and any design may be used as long as the information presented is the same.

[0069] In the above-described embodiment, the output of information is not limited to display on a display screen, but may be audio output from a speaker, output to a file, printed on paper or the like by a printer, projected onto a screen or the like by a projector, or in other forms.

[0070] In addition, in the above description, information such as programs, tables, files, etc. that realize each function can be stored in a storage device such as a memory, a hard disk, or an SSD (Solid State Drive), or in a recording medium such as an IC card, an SD card, or a DVD.

[0071] The above-described embodiment has the following characteristic configurations, for example.

[0072] (1) A data acquisition method executed by one or more computer devices (e.g., computer device 200, computer device capable of communicating with computer device 200) connected to one or more language models (e.g., language model 210), the method including: generating a prompt (e.g., prompt 500) for identifying a database (e.g., a database on a local device, a database on a device on network 100) containing information related to the content of a user's question (e.g., see step S602); inputting the prompt into a language model (e.g., a first language model) (e.g., see step S603); and receiving at least one of the database location, database name, database format, and database column information from the language model as an answer to the prompt (e.g., see step S604).

[0073] In the above configuration, the computer device receives, as a response to the prompt, at least one of the location, name, format, and column of a database containing information related to the question, thereby narrowing down existing databases containing information related to the question. For example, with the above configuration, the computer device can obtain information related to the question from existing databases, thereby providing an accurate answer without building a dedicated database. Furthermore, with the above configuration, for example, since there is no need to build a dedicated database, storage capacity can be reduced. Furthermore, with the above configuration, for example, the computer device can narrow down existing databases containing information related to the user's question, thereby reducing the processor load when obtaining information related to the user's question.

[0074] (2) The prompt includes at least one of the following configuration information of the database from which data is to be acquired: the location of the database, the name of the database, the format of the database, and information about the columns of the database (see, for example, FIGS. 4 and 5).

[0075] According to the above configuration, since the prompt includes configuration information of the database from which data is to be acquired, for example, an administrator can appropriately manage the information provided to the user by adjusting the configuration information of the database. Also, for example, if the prompt includes information about a column of the database, the language model can acquire the structure and meaning of the database from the information about the column of the database, thereby more appropriately narrowing down the databases that include information related to the content of the user's question.

[0076] (3) The computer device acquires information related to the user's question from the database based on at least one of the database location, database name, database format, and database column information received from the language model (e.g., see step S805); the computer device generates a prompt (e.g., prompt 900) including the question, information related to the question, and an instruction to understand the question and the information related to the question and generate an answer (e.g., see step S806); the computer device inputs the prompt into a language model (e.g., a second language model) (e.g., see step S806); and the computer device receives the answer to the question from the language model (e.g., see step S807).

[0077] According to the above configuration, for example, the language model understands the content of a question from a user and information related to the content of the question and generates an answer, so that hallucination of the language model can be suppressed.

[0078] (4) The database format is the database extension and / or the type of database structure (e.g., "format").

[0079] According to the above configuration, for example, the format of the database location can be changed depending on the format of the database, so that data can be acquired from any type of database.

[0080] (5) A storage device for storing the database from which data is acquired is provided in either or both of a computer device (e.g., computer device 200) and a computer device (e.g., devices 101, 102, 103, 104, 105, 106) connected to the computer device via a network.

[0081] In the above configuration, for example, a database stored in a storage device accessible by the computer device can be used as a target for data collection.

[0082] Furthermore, the above-described configurations may be modified, rearranged, combined, or omitted as appropriate within the scope of the present invention.

[0083] It should be understood that items included in a list in the format "at least one of A, B, and C" can mean (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C). Similarly, items listed in the format "at least one of A, B, or C" can mean (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C).

[0084] 100...network, 200...computer device, 210...language model

Claims

1. A data acquisition method executed by one or more computer devices connected to one or more language models, the data acquisition method including: the computer device generating a prompt to identify a database containing information related to a user's query; the computer device inputting the prompt into the language model; and the computer device receiving at least one of the following information from the language model as an answer to the prompt: the location of the database, the name of the database, the format of the database, and column information of the database.

2. The data acquisition method according to claim 1, wherein the prompt includes at least one of the following configuration information of the database from which data is acquired: the location of the database, the name of the database, the format of the database, and information about columns of the database.

3. A data acquisition method as described in claim 1 or claim 2, comprising: a computer device acquiring information related to the user's question from the database based on at least one of the database location, database name, database format, and database column information received from the language model; a computer device generating a prompt including the question, information related to the question, and an instruction to understand the question and the information related to the question and generate an answer; a computer device inputting the prompt into a language model; and a computer device receiving an answer to the question from the language model.

4. The data acquisition method according to claim 1, wherein the database format is a database extension and / or a type of database structure.

5. The data acquisition method according to claim 1, wherein the storage device for storing the database from which data is acquired is provided in either or both of the computer device and a computer device connected to said computer device via a network.

6. A computer device connected to one or more language models, comprising: a generation unit that generates a prompt to identify a database containing information related to the content of a user's question; an input unit that inputs the prompt into the language model; and a receiving unit that receives at least one of the following information from the language model as an answer to the prompt: the location of the database, the name of the database, the format of the database, and information on columns of the database.

7. A data acquisition system connected to one or more language models, comprising: a generation unit that generates a prompt to identify a database containing information related to a user's question; an input unit that inputs the prompt into the language model; and a receiving unit that receives at least one of the following information from the language model as an answer to the prompt: the location of the database, the name of the database, the format of the database, and information on columns of the database.

Images