Information processing method, program, and information processing device

Abstract

The aim is to provide business support that is both accurate and efficient. [Solution] An information processing method is provided in which an acquisition process is executed by a computer, in which search results are obtained from a vector database based on an input, the vector database stores stored data in a vector-searchable manner, and each of the stored data includes pharmacy product information or pharmacy business information.

Description

[Technical Field]

[0001] The present invention relates to an information processing method, a program, and an information processing device. [Background technology]

[0002] Patent Document 1 describes a system having a pharmaceutical information database. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-196803 Summary of the Invention [Problem to be solved by the invention]

[0004] Pharmacies are required to carry out their work with a high degree of accuracy and efficiency.

[0005] The present invention aims to provide business support that is both accurate and efficient. [Means for solving the problem]

[0006] According to the present invention, there is provided an information processing method in which an acquisition process is executed by a computer, in which search results are acquired from a vector database based on an input, the vector database stores stored data in a vector-searchable manner, and each of the stored data includes pharmacy product information or pharmacy business information.

[0007] According to the present invention, it is possible to provide business support that is both accurate and efficient. [Brief explanation of the drawings]

[0008] [Figure 1]1 illustrates an example of an information processing system 100 (including an information processing device 1 according to an embodiment) of an embodiment. [Figure 2] 1 shows an example of a hardware configuration of an information processing device 1. [Figure 3] 1 illustrates functional blocks realized by the control unit 12. [Figure 4] FIG. 10 is a process flow diagram for explaining a first example (search result presentation) of the information processing method according to the embodiment. [Figure 5] FIG. 10 is a process flow diagram for explaining a second example (presentation via a generative AI model) of the information processing method according to the embodiment. [Figure 6] This is a processing flow diagram for explaining a variation of the second example of the information processing method of the embodiment (presentation via the generative AI model: direct data transmission from the vector DB system to the generative AI model). [Figure 7] FIG. 10 is a process flow diagram illustrating a third example (use of categories) of the information processing method according to the embodiment. [Figure 8] FIG. 10 is a process flow diagram for explaining a modification (use of categories based on supplemental information) of the third example of the information processing method according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Various features shown in the following embodiments can be combined with each other. Furthermore, each feature can be an invention independently.

[0010] Please note that in this disclosure, when multiple configurations or items are arranged with a selective conjunction (e.g., "or"), this means "including at least one" of the multiple options, unless otherwise specified or contradictory. Specifically, for example, when "α or β" is written, this may include either "α" or "β," and may also include both "α" and "β." The same applies to three or more options.

[0011] 1. Configuration of the embodiment 1-1.System Configuration In the embodiment, as an example, an information processing method of the embodiment is provided by an information processing system 100 shown in FIG. 1 (mainly by an information processing device 1). The information processing method of the embodiment acquires search results from a database (vector database) based on an input (e.g., input IN in FIGS. 4 to 8 described later). Hereinafter, the database may be abbreviated as "DB." The information processing system 100 includes an information processing device 1, an input means 2 (e.g., a terminal of a user U), a vector DB system 3, and a generative AI model 4, which are capable of communicating with each other via a communication network 6 (e.g., the Internet).

[0012] The information processing device 1 is provided as a server of any type (e.g., any web server, cloud server, etc.), for example. The information processing device 1 is also a "computer" that executes information processing in the embodiment. The specific configuration of the information processing device 1 will be described later with reference to FIGS. 2 and 3.

[0013] A terminal, which is an example of the input means 2, is operated by a user U. There are no limitations on the type or structure of the user U's terminal, and it may be, for example, a smartphone, tablet PC, notebook PC, or desktop PC. For example, if the user U is a pharmacy member (e.g., a clerk or pharmacist), the terminal may be a terminal installed in the pharmacy store (e.g., a counter terminal), or may be a business terminal assigned to each member with an application program installed.

[0014] The vector DB system 3 in FIG. nThe vector DB system 3 can store vectors in a searchable manner. There are no limitations on the specific implementation method, configuration, processing content, or application technology of the vector DB system 3. The vector DB system 3 can arbitrarily adopt any known vector database and its peripheral system, and may be constructed, for example, on any platform. As an example, the vector DB system 3 of the embodiment includes a vector database main body and any embedding model (e.g., a sentence embedding model, etc.) for vectorizing the input IN. The vector database main body can provide database functions (e.g., data storage, search, editing, management, etc.). Specifically, the vector database main body stores embedded vectors created in advance and stores related data (e.g., text data, etc.) linked to each embedded vector in a searchable manner.

[0015] The generative AI model 4 shown in FIG. 1 is communicatively connected to the information processing device 1 via a communication network 6. The generative AI model 4 has, for example, a neural network trained using a large amount of teacher data. The generative AI model 4 has the function of receiving an arbitrary request from the information processing device 1, generating a response to the request, and outputting the response to the information processing device 1. The term "request" is a broadly encompassing concept of information provided to the generative AI model 4. In many cases, a "request" is roughly synonymous with a prompt, but in the case of a multimodal generative AI model, a request can also include arbitrary attached data (e.g., images, audio, video, etc.) along with the prompt.

[0016] In an embodiment, the request may include "input IN or supplemental information" and "search results," as described below in Figure 5 or Figure 6. Note that when "including input IN or supplemental information" is used, it may include either "input IN" or "supplemental information," or may include both "input IN" and "supplemental information."

[0017] There are no limitations on the configuration of the generative AI model 4. For example, any large-scale language model (LLM) can be adopted, and for example, one having any model architecture can be adopted. For example, any GPT (Generative Pre-trained Transformer) model can be adopted, and specifically, for example, various generative AI models provided by OpenAI (registered trademark) (e.g., ChatGPT (registered trademark), GPT-4o, o1, or o1-mini) may be adopted. Only one type of generative AI model 4 may be adopted, but this is not limited thereto, and a combination of multiple large-scale language models may also be adopted.

[0018] In the embodiment, for convenience of explanation, the information processing device 1, the vector DB system 3, and the generative AI model 4 are described as separate components as an example. However, this is just an example, and it should be noted that any two or more of the information processing device 1, the vector DB system 3, and the generative AI model 4 may be integrated into one system.

[0019] 1-2. Hardware configuration of information processing device 1 As shown in Figure 2, the information processing device 1 has, as an example, a communication unit 10, a memory unit 11, a control unit 12, an output unit 13, and an input unit 14, and these components are electrically connected within the information processing device 1 via a communication bus 15.

[0020] The communication unit 10 may employ, for example, a wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), wired LAN network communication, etc. The communication unit 10 may also employ a configuration in which it is connected to the communication network 6 via a wireless communication means such as wireless LAN network communication, mobile communication such as 3G / LTE / 5G, or Bluetooth (registered trademark) communication. The communication unit 10 may also be configured to use both the wired communication means and the wireless communication means described above.

[0021] The storage unit 11 stores, for example, various programs of the information processing device 1 executed by the control unit 12, and various values ​​such as constants, variables, and setting values. The storage unit 11 may be a storage device such as a solid state drive (SSD), or a random access memory (RAM) that stores temporarily required information (arguments, arrays, etc.) related to program calculations. Furthermore, the information processing device 1 may use an external storage unit (e.g., an external storage medium, cloud, etc.) in addition to the storage unit 11.

[0022] The control unit 12 is configured to execute processing and control related to information processing of the information processing device 1. The control unit 12 can be configured, for example, by a central processing unit (CPU), and in the embodiment, the control unit 12 is an example of a processor capable of executing programs related to each processing step in each processing flow of Figs. 4 to 8 described below. The control unit 12 realizes various functions related to the information processing device 1, for example, by reading out programs stored in the storage unit 11. Furthermore, information processing by software in the information processing device 1 is realized, for example, by the control unit 12 as hardware processing various programs stored in the storage unit 11.

[0023] The output unit 13 is, for example, a display unit of the information processing device 1. The output unit 13 may be included in the housing of the information processing device 1 or may be externally attached. The output unit 13 displays a screen of a graphical user interface (GUI) that can be operated by a user. Any display device such as a liquid crystal display can be used as the output unit 13. It is optional whether or not the information processing device 1 includes the output unit 13. For example, the output of the information processing device 1 may be displayed on a display unit located at a location independent of the location where the information processing device 1 is installed. The output unit 13 may also have a device that outputs audio.

[0024] The input unit 14 is configured to accept, for example, an operation input made by a person accessing the information processing device 1. The person accessing the information processing device 1 may be, for example, the user U in FIG. 1 or a server administrator. The input unit 14 may be included in the housing of the information processing device 1 or may be externally attached. The input unit 14 may be, for example, a touch panel, a switch button, a mouse, or a keyboard. It is optional whether or not the information processing device 1 includes the input unit 14. For example, the information processing device 1 may accept an operation input to the information processing device 1 via an information processing terminal (e.g., a terminal as the input means 2 in FIG. 2) located at a location separate from the location where the information processing device 1 is installed.

[0025] 1-3. Functional configuration of information processing device 1 The functional configuration of the information processing device 1 according to the embodiment will be described with reference to Fig. 3. Information processing by software stored in the storage unit 11 is specifically realized by the control unit 12, which is an example of hardware, to realize each functional unit included in the control unit 12 (e.g., acquisition processing unit 12a, input processing unit 12a1, sorting processing unit 12b, output processing unit 12c, supplementary processing unit 12c1, and model input / output processing unit 12c2).

[0026] The functional units (12a to 12c2) in FIG. 3 are configured to be able to execute the processing steps (S1 to S3b) of the information processing device 1 in the processing flows of FIGS. 4 to 8, which will be described later. The acquisition processing unit 12a is configured to be able to execute acquisition processing S1. The input processing unit 12a1 is configured to be able to execute input processing S1a. The sorting processing unit 12b is configured to be able to execute sorting processing S2. The output processing unit 12c is configured to be able to execute output processing S3. The supplementary processing unit 12c1 is configured to be able to execute supplementary processing S1b, S3a. The model input / output processing unit 12c2 is configured to be able to execute model input / output processing S3b.

[0027] The above components may be implemented by software, but are not limited to this and may also be implemented by hardware. When implemented by software, various functions can be realized by an arbitrary processor (e.g., a CPU) executing a computer program. The program may be stored on a non-transitory computer-readable recording medium, provided as a downloadable file from an external server, or implemented by so-called cloud computing, in which a program stored in an external storage unit is read and functions are realized. When implemented by hardware, various circuits such as an ASIC, FPGA, or DRP can be used. In the embodiments, various information and concepts encompassing such information are handled. These are represented by high and low signal values ​​or quantum bits as a binary bit set consisting of 0s or 1s, and communication and calculations can be performed by the above software or hardware aspects. The software may be executed on a general-purpose OS or a dedicated OS.

[0028] 1-4. Data stored in Vector DB System 3 In the embodiment, as an example, the vector DB body of the vector DB system 3 includes a plurality of stored data DTs. n (n=1, 2, 3, . . . ) are stored in a vector searchable manner. In the embodiment, "vector search" refers to the process of searching a given input IN and each vectorized stored data DT n The input IN functions as a search query, and when an arbitrary input IN is given, information similar to that input IN can be searched for.

[0029] In the embodiment, as an example, each of the stored data DT nincludes "pharmacy product information" or "pharmacy business information." "Pharmacy product information" is various information about various products handled by pharmacies (e.g., pharmaceuticals, quasi-drugs, and other optional products). "Pharmacy business information" is a broad concept that includes various pharmacy operations (e.g., reception, handling prescriptions, accounting, etc.), and can include not only internal pharmacy procedures, but also various information about interactions with outside the pharmacy. Specific examples and variations of stored data will be explained in the section below (3. Specific examples and variations of stored data).

[0030] Here, in the embodiment, as an example, each of the stored data DT n contains embedding vectors, and each embedding vector is, for example, stored data DT n The embedded vector is a vector representation of part or all of the stored data. In the embodiment, the associated data linked to the embedded vector includes, for example, data indicating the contents of pharmacy product information or pharmacy business information. Regarding the "vector representation of part of stored data," for example, when a "Q&A system" is provided in the embodiment, one piece of stored data may be configured to include a pair of question text and answer text. In this case, the functionality of the Q&A system can be fulfilled by vector-representing only the question text portion of the stored data. This is because, when a question is given, a similar question embedding vector is searched for by vector search, and associated data (answer text) corresponding to the question embedding vector is provided, thereby realizing a Q&A system. Regarding the "vector representation of all stored data," the question text and the answer text may be concatenated and represented as a vector. In this case, when a question is given, a similar question / answer embedding vector may be searched for by vector search, and associated data corresponding to the question / answer embedding vector (e.g., only the answer portion or the entire question / answer) may be provided. It should be noted that such a "partial or full vector representation" is not limited to Q&A systems, but can be adopted in any search system (e.g., recommendation systems, etc.).

[0031] It should be noted that the present invention is not limited to providing one embedding vector for one piece of stored data, and that it is also possible to provide multiple embedding vectors for one piece of stored data. For example, one piece of text data (document data) may be converted into word embedding vectors, different parts of the document may be vectorized, or the entire document may be vectorized.

[0032] 2. Information processing method of the embodiment An example of an information processing method according to an embodiment will be described with reference to Figures 4 to 8. Here, an example in which a "Q&A system" is provided according to an embodiment will be described. Note that in the information processing method according to the embodiment, the information processing device 1 is mainly used as the "computer" that executes each process, but any process can also be executed by a computer other than the information processing device 1.

[0033] 2-0. Data storage (step S100) As a common feature of the process flows of FIGS. 4 to 8, in step S100, the vector DB system 3 stores the stored data DT n In this case, each stored data DT n Regarding the embedding vector (stored data DT n A process of creating a vector representation of a part or all of the data is executed, and in this process, an embedded model installed in the vector DB system 3 can be used. Note that there are various well-known techniques for constructing various vector databases by vectorizing any data and storing it in a database, and the embodiments are not limited to specific techniques, so a description thereof will be omitted.

[0034] 2-1. First example (search operation) The processing flow in FIG. 4 is an example in which the search results of the vector DB system 3 are directly output. As shown in FIG. 4, in the first example, as an example, an acquisition process S1, a sort process S2, and an output process S3 are executed by a computer (information processing device 1). More specifically, in the first example shown in FIG. 4, an input IN is first received by an input means 2 (e.g., a terminal of a user U) (step S0). The input IN is input information, and in a Q&A system, for example, a question from the user U. Note that the input IN is a broad concept that includes any input, but in the embodiment, a more specific example will be described in the section below (5. Specific examples of input IN in the embodiment).

[0035] Next, an acquisition process S1 is executed. In the acquisition process S1 of the embodiment, as an example, a search result is acquired from the vector DB system 3 based on the input IN. In the embodiment, as an example, the acquisition process S1 includes an input process S1a. The input process S1a transmits the input IN given to the information processing device 1 to the vector DB system 3 as raw data (e.g., text) or after performing any processing (e.g., conforming to interface specifications such as an API, satisfying data compatibility, etc.).

[0036] When the vector DB system 3 of the embodiment receives an input IN, it executes a vectorization process S101 (e.g., sentence embedding or word embedding, etc.). This vectorizes the input IN. Next, the vector DB system 3 executes a similarity search process S102. In the similarity search process S102, any calculation method for similarity search (e.g., cosine similarity, vector inner product value, Euclidean distance, Manhattan distance, etc.) can be adopted, and there are no limitations.

[0037] Next, the vector DB system 3 executes the search result output process S103. In this process, the vector DB system 3 vectorizes the input IN and compares it with the embedded vectors (similarity calculation), thereby identifying embedded vectors with relatively high similarity (e.g., extracting the highest similarity or a predetermined number of vectors in descending order of similarity), and can output related data linked to the identified embedded vectors. This mechanism makes it possible to search for information that is highly similar or related to the input IN. For example, a Q&A system can output appropriate related data (answers) to questions.

[0038] The number of search results acquired from the vector DB system 3 may be, but is not limited to, one (highest similarity). In one embodiment, a plurality of search results are acquired from the vector DB system 3. The "number of acquired search results" in this case is not limited, but may be, for example, 2 to 100, or may be 2, 3, 4, 5, 10, 20, 30, 40, 50, or 100, or may be within any two numerical ranges listed here. The number of acquired results may be a fixed setting value, or may be prepared as a variable parameter. In this case, the number may be variably set by, for example, a number specification included in the input IN from the input means 2 (e.g., an input operation from the input means 2) or by an operation via the input unit 14 of the information processing device 1.

[0039] Next, in the embodiment, as an example, a sorting process S2 is further executed by the computer (information processing device 1). In the sorting process S2 of the embodiment, as an example, multiple search results acquired from the vector DB system 3 are sorted in order of similarity with the input IN. As an example, the sorting process S2 may employ a process of sorting the multiple search results acquired by similarity and then narrowing down the results to a predetermined number of results with the highest similarity. The number of results to be narrowed down may be any number between 2 and 50, or may be 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 40, or 50, or may be within any two numerical ranges listed here. The number of results to be narrowed down may be a fixed setting value or may be prepared as a variable parameter. In this case, the number may be variably set, for example, by specifying the number in the input IN from the input means 2 (e.g., by input operation from the input means 2) or by operation via the input unit 14 of the information processing device 1. Alternatively, as another example, the sorting process S2 may simply sort the acquired search results in order of similarity, without narrowing down the number of results.

[0040] In the embodiment, the information processing device 1 executes the sorting process S2, but this is just an example. As another example, a process similar to the sorting process S2 may be provided as an internal process of the vector DB system 3.

[0041] Next, as an example in the embodiment, an output process S3 is further executed by the computer (information processing device 1). In the output process S3 in the first example of the embodiment, as an example, the search results are directly output. That is, a predetermined number of search results provided by the sort process S2 are transmitted to the input means 2, and if the input means 2 is a terminal of the user U, for example, they are displayed on the display UI screen of this terminal.

[0042] According to the Q&A system provided by the first example above (Figure 4), when an input IN (question) is given, it is possible to search the vector DB system 3 for answers that are highly relevant to the input IN (i.e., information that is likely to be an appropriate answer).

[0043] Table 1 below provides a specific example of the information processing method of the embodiment.

[0044] [Table 1]

[0045] In the example shown in Table 1, each stored data D T However, it includes an embedding vector that embeds each text data in the question column of Table 1 (the vector itself is omitted in Table 1), and related data linked to each embedding vector (the text data in the answer column of Table 1).

[0046] In Table 1, each embedding vector (each question vector) corresponding to each text data in the question column is stored in the vector DB system 3. Each embedding vector is stored in the stored data DT n It is a vector representation of a part (question part only) of (question column + answer column).

[0047] In Table 1, each embedding vector based on each text data in the question column is linked to related data (each text data in the answer column). In the example of Table 1, the text data in the answer column includes pharmacy product information or pharmacy business information.

[0048] Table 1 shows an example of the results of a similarity search (cosine similarity search) between an input vector obtained by vectorizing (sentence embedding) an input text and each embedding vector (question vector). The numerical data shown in "distance" indicates the distance between the input vector and each embedding vector, and sorting process S2 sorts them in order of closest distance (i.e., highest similarity).

[0049] 2-2. Second example (via generative AI) Each processing flow in Fig. 5 or Fig. 6 shows a second example of an embodiment. In this second example, the search results of the vector DB system 3 are output via a generative AI model. The example in Fig. 5 and the example in Fig. 6 differ in the transmission path of the search results from the vector DB system 3. In this second example, the DB search function and the generative AI model are integrated.

[0050] First, a description will be given with reference to Fig. 5. In the second example, the information processing method, for example, causes a computer (information processing device 1) to execute an acquisition process S1, a sorting process S2, and an output process S3, but the processing content of the output process S3 differs from that of the first example (see Fig. 4). The data storage (step S100), input reception (step S0), acquisition process S1, each process (S101 to S103) of the vector DB system 3, and sorting process S2 are the same as those of the first example (see Fig. 4).

[0051] Next, the computer (information processing device 1) is further caused to execute output processing S3, and in the output processing S3 in the second example, as an example, the search results are output via the generative AI model 4. There is no limitation on how the search results are given to the generative AI model 4, but in the example of Figure 5, the information processing device 1 receives the search results from the vector DB system 3 and transmits the search results (and other information) to the generative AI model 4.

[0052] In one example of the embodiment, in the output process S3, supplemental information or input IN is provided to the generative AI model along with the search results, thereby outputting the generated results obtained from the generative AI model. Specifically, in FIG. 5, the output process S3 includes a supplemental process S3a and a model input / output process S3b.

[0053] In the supplemental processing S3a, the computer (information processing device 1) acquires supplemental information. The "supplemental information" may be various information useful for providing effective output presentation. In FIG. 5, as an example, in order to acquire personalized supplemental information, necessary information (e.g., user-specific information, information linked to a login account, etc.) is acquired from the input means 2. Specific examples and variations of supplemental information will be described later in the section (4. Explanation of "supplemental information" in the embodiment).

[0054] In the model input / output process S3b, the computer (information processing device 1) provides a request to the generative AI model, thereby outputting the generation results obtained from the generative AI model. A "request" refers to any information provided to the generative AI model with the intention of requesting generation. In the embodiment, the request includes supplementary information, input IN, and search results. The prompt included in the request may include a statement of "how the supplementary information, input IN, and search results should be considered," or may include any other conditional statement (e.g., other context generation conditions, etc.).

[0055] Upon receiving the request, the generation AI model 4 outputs a generation result (AI model answer) based on the information included in the request. In the model input / output process S3b, when the computer (information processing device 1) receives the generation result, it outputs the generation result either as is or after appropriate processing. In other words, the generation result is transmitted to the input means 2, and if the input means 2 is a terminal of the user U, for example, it is displayed on the display UI screen of this terminal.

[0056] According to the Q&A system provided by the second example (Fig. 5) above, when an input IN (question) is given, the vector DB system 3 is searched for answers that are highly relevant to it (i.e., information that is likely to be an appropriate answer), and the search results can be output via the generative AI model 4. In addition, since the vector DB system 3 outputs appropriate search results for the input IN (question), and these search results are taken into consideration in the generation process S200, there is an advantage in that hallucination of the generative AI model 4 can be suppressed.

[0057] FIG. 6 shows a further variation of the second example. In the example of FIG. 6, the search result output process S103a links the vector DB system 3 and the generation AI model 4 so that the search results (e.g., only a predetermined number of search results with high similarity) are transmitted directly from the vector DB system 3 to the generation AI model 4. Also, the sorting process S2 is omitted. In the example of FIG. 6, the acquisition process S1 is executed by a computer (information processing device 1), but the "search results acquired from the vector DB system 3" are provided to the generation AI model 4, not to the information processing device 1. The generation AI model 4 executes the generation process S200 based on the search results from the vector DB system 3 and the request (including the input IN and supplemental information) from the information processing device 1. Such a configuration can also be employed as one variation of the embodiment.

[0058] 2-3. Third example (using categories) In the embodiment, as an example, the stored data DT n In the acquisition process S1, the stored data DT similar to the input may be categorized and stored. n Although there are various possible specific examples and implementation methods of such "categories," one example will be described below.

[0059] 7 shows a third example (using categories) of the information processing method according to the embodiment. This is similar to the processing flow in FIG. 5, but differs in the content of data storage (step S100) and includes an additional category search process S101a.

[0060] First, in step S0, a user U provides an arbitrary input IN to a computer (information processing device 1) via the input means 2. For example, in a Q&A system, this input IN is a question. Next, the computer (information processing device 1) receives this input IN and transmits it to the vector DB system 3 (step S1a), and the input IN is vectorized in the embedded model of the vector DB system 3 (step S101).

[0061] Next, in the category search process S101a, the vectorized input IN (e.g., the embedding vector of a question sentence) is searched (similarity searched) in a prepared category vector space to identify a category corresponding to the vectorized input IN. Here, in the embodiment, the vector space corresponds to a set (group) of vectors related to predetermined data so that a similarity search can be performed. The category vector space is an embedding vector space in which each embedding vector corresponding to each category name is stored, and in the example of FIG. 7, it is created and stored in advance in step S100. In the category search process S101a of the embodiment, as an example, one category with the highest similarity may be identified. Alternatively, multiple categories with similarity (e.g., cosine similarity) equal to or greater than a predetermined value (e.g., 0.85) may be obtained, and the category with the highest similarity may be selected from among them.

[0062] Furthermore, in the category search process S101a, based on the appropriate category (the category specified this time) specified by the search, a plurality of embedding vector spaces (stored data DT n The vector space corresponding to the specified category is selected from the vector spaces (corresponding to the category). Each embedding vector space stores various information belonging to the specified category in a vector searchable manner.

[0063] Next, in the similarity search process S102, a search (similarity search) based on the vector expression of the input IN is performed in the embedded vector space selected in the category search process S101a. In this way, the vector search may be performed in multiple stages (category search → stored data search). As a modified example, an implementation may be adopted in which each embedded vector space has further detailed categories as smaller categories (minor classifications).

[0064] More comprehensively, one or more embedded vector spaces are provided in which stored data is stored in a vector searchable manner, and each embedded vector space is configured to be searchable by category. For example, each of the multiple embedded vector spaces may be associated with a different category. Alternatively, within each embedded vector space, stored data may be grouped into categories using any method.

[0065] In the example of FIG. 7, the vector DB system 3 is provided with multiple embedded vector spaces, and different data is stored in each embedded space. This is just one example, and there are no limitations on the specific implementation method. For example, when the vector DB system 3 is built on a cloud platform, multiple vector databases may be built switchably within a single platform. Alternatively, a configuration may be adopted in which vector databases on different cloud platforms or on-premise environments can be used across the board.

[0066] FIG. 8 is a partial modification of the processing flow of FIG. 7. The "category" can also be treated as an example of "supplementary information." In the example of FIG. 8, category information based on the supplementary information acquired in the supplementary processing S1b is provided to the vector DB system 3. The vector DB system 3 executes a category determination process (step S101a) based on the supplementary information (category information). The processing thereafter can be the same as the processing flow of FIG. 7.

[0067] As another example, when the "supplementary information" is a fixed value based on settings, the "category" may be selected (generated) using an arbitrary AI model. The above procedure assumes vectorization, but as another example, a prompt including the input IN and a "category classification instruction" may be given to an arbitrary AI model (e.g., the generation AI model 4). This allows the generation AI model 4 to respond with a category that is likely to be related to the input IN. Furthermore, supplementary information may be given together, which has the advantage of improving classification accuracy.

[0068] In the example of Figure 7 or Figure 8, categorization is used in the second example (using a generative AI model: see Figure 5 or Figure 6), but this is just an example. Categorization may also be used in the first example (see Figure 4).

[0069] Table 2 below shows a specific example of using categories. The information in the question column and answer column is pre-classified into each category in the Category column.

[0070] [Table 2]

[0071] According to the information processing method of the above-described embodiment, it is possible to provide pharmacy business support that is both accurate and efficient. That is, by utilizing the vector DB system 3, accurate information can be efficiently searched. Furthermore, according to the second and third examples, when using the generation AI model 4, there is also the advantage that hallucination can be suppressed.

[0072] 3. Examples and variations of stored data In step S100 in FIG. 4 to FIG. 8, data of any type and any quantity is stored in the vector DB system 3 as the stored data DT n This point will be explained below.

[0073] Overview In the embodiment, as an example, each stored data includes pharmacy product information or pharmacy business information. For convenience, the following uses the grouping in Table 3 below, and specific information corresponding to each group is listed as an example.

[0074] [Table 3]

[0075] Although Table 3 refers to "pharmacy side (other than the patient)," "others other than the patient" may include, for example, pharmacy staff, pharmacists inside or outside the pharmacy, doctors (when making inquiries, etc.), other medical professionals, or external organizations (e.g., screening organizations for national health insurance, business system providers, etc.) (all of these are not limited to natural persons). It should be noted that the group classifications in Table 3 are merely examples, and that the stored data that can be employed in the embodiments is not limited to Table 3. The groups in Table 3 are not necessarily completely exclusive, and some of the information exemplified below may belong to both one group and another. Furthermore, any two or more pieces of information exemplified below may be combined and employed as a single stored data set. Therefore, it should be noted that each stored data set may include information that corresponds to both pharmacy product information and pharmacy business information.

[0076] For example, if you want to provide a Q&A system, create question and answer information for each information in each group GD1 to GD4, and store the data DT n The question and answer information may be created in any format, such as FAQs (Frequently Asked Questions), a collection of questions and answers, a dialogue format, or an interview format. Alternatively, when providing a recommendation system or the like, by using the information in each of groups GD1 to GD4 as stored data, it becomes possible to search for information related to the input IN (e.g., keyword) according to a similarity search.

[0077] 3-2.Pharmacy product information In the embodiment, the "pharmacy product" may include medicines, quasi-drugs, or any other products handled by pharmacies. In the embodiment, as an example, the pharmacy product information includes at least one of the following (D1) and (D2): (D1) At least one of information about the pharmacy product itself, patient information related to the pharmacy product, prescription information, and instruction information (see group GD1 below) (D2) Administrative information about pharmacy products (see group GD2 below)

[0078] 3-2-1. Group GD1: Pharmacy product information directly related to patients Specifically, in the group GD1, at least the following information (D1a) or (D1b) can be arbitrarily adopted.

[0079] (D1a) Information about the pharmacy product itself Product information (e.g., product name, name or abbreviation, ingredients, target patient's condition (symptoms, symptoms, effects, efficacy), usage, dosage, product shape, dosage form, packaging, etc.) Regulations or classifications (e.g., pharmaceutical classifications, schedule classifications, approval information, etc.) ·Safety information (e.g. side effects, contraindications, precautions, interactions, storage methods, etc.) Economic information (e.g., drug prices, availability of generic and original drugs, insurance coverage, etc.)

[0080] (D1b) Patient information, prescription information, or instruction information related to pharmacy products Questions and answers about pharmacy products from any person (e.g., patient, pharmacy staff, pharmacist, doctor, etc.) Relationship between subjective information provided by patients and pharmacy products Relationship between patient objective information and pharmacy products -Items evaluated by pharmacists during medication instruction -Instructions from pharmacists to patients Information on prescriptions

[0081] 3-2-2. Group GD2: Management information on pharmacy products, etc. Specifically, in group GD2, at least the following information can be arbitrarily adopted. Information regarding product inventory management (e.g., inventory data, order date, supplier, quantity, delivery date, storage, and other records)

[0082] 3-3.Pharmacy business information In the embodiment, as an example, the pharmacy business information includes at least one of the following (D3) and (D4). (D3) Information on patient care operations (see group GD3 below) (D4) At least one of information regarding internal pharmacy operations and information regarding collaboration with outside pharmacies (see Group GD4 below)

[0083] 3-3-1. Group GD3: Information on patient care Specifically, in group GD3, at least the following information can be arbitrarily adopted. Patient reception Prescription handling services Correspondence between prescribed medication and patient condition (e.g., chief complaint, symptoms, medical condition, etc.) Dispensing operations (e.g., dispensing equipment, dispensing records, etc.) Audit work (e.g. prescription audits, dispensed drug audits, etc.) Medication guidance Accounting services · Receipt work Handling of various patient information (e.g., medication history, health information, or personal information) (e.g., acquisition, searching within the system, management, presentation, disposal, etc.)

[0084] 3-3-2. Group GD4: Information related to internal or external business operations of the pharmacy Specifically, in group GD4, at least the following information (D4a) or (D4b) can be arbitrarily adopted.

[0085] (D4a) Information primarily related to internal pharmacy operations (see (D4a1) or (D4a2) below) (D4a1) Information about business operations within the pharmacy Information regarding pharmacy product ordering, delivery, or inventory ·Questions and answers about basic pharmacy information ·Questions and answers about pharmacy procedures Opinions or impressions regarding the status of pharmacy operations, problems, areas for improvement, etc. ·Questions and answers about quantitative data for pharmacies ·Answering questions about pharmacy equipment, supplies, and supplies - Latest and past information on various standards (e.g., various laws and regulations including the Pharmaceutical Affairs Act, public standards, administrative guidelines, industry guidelines, specifications, standards, industry practices, and their revision history, etc.)

[0086] (D4a2) Pharmacy systems and equipment related information - Information related to any business system or equipment (e.g., operation methods, maintenance methods, vendor or manufacturer contact points, past response history, management history, etc.) - Answers to questions or manuals regarding how to operate any business systems (e.g., prescription computers, electronic medication history systems, etc.) · Answers to questions or manuals regarding how to use any business software (e.g., electronic medication history software, inventory management software, etc.) · Questions and answers or manuals regarding the operation of any pharmacy equipment (e.g., dispensing equipment, inspection equipment, packaging machines, etc.) Reception work (e.g., how to enter data into the pharmacy system, etc.) ·Various operational procedures and manuals within the pharmacy · Maintenance, operation, manuals, etc. of systems and equipment within the pharmacy -Information on how to operate the business system introduced at the pharmacy

[0087] (D4b) Information on collaboration with outside pharmacies (e.g., information on dispensing fees, etc.) Questions about the dispensing fee schedule Questions regarding calculation requirements for dispensing fees Questions regarding the points for specific dispensing procedures Dispensing fee revision information Questions about dispensing fee claims Questions about drug price standards Questions about errors in dispensing fee claims · Questions regarding various inquiries Information regarding prescription applications to review agencies · Application or notification to a public health center, welfare bureau, or city, ward, town, or village office ·Required items for setting up prescription computers, etc. Information about the classification of pharmacy stores (e.g., based on quantitative information about pharmacies, which classification of dispensing fees your pharmacy falls into) Information on how to set the relevant classification as the default setting on the receipt computer

[0088] 3-4.Other examples of classification of stored data In the above, pharmacy product information and pharmacy business information are listed with a number of specific pieces of information, with examples of group classifications given in Table 3. However, please note that the classifications in Table 3 are merely examples.

[0089] As an example, some of the information exemplified above includes "important information that should not be referred to as erroneous information (e.g., drug information, quasi-drug information, prescription handling information, dispensing fee information, drug price standard information, standard information of laws and regulations, etc.)." Therefore, according to the embodiment, "stored data DT n It also provides that "including, by way of example, regulated or compliant information."

[0090] As another example, some of the information exemplified above have in common the fact that they are information relating to the exchange of information with any person related to the pharmacy (e.g., pharmacy clerk, pharmacist, patient, doctor, system vendor, business partner, etc.). For this reason, according to the embodiment, the "stored data DT n However, as an example, it also provides an aspect that "includes information regarding communication with persons related to the pharmacy."

[0091] As another example, some of the information exemplified above has in common the fact that it is information related to pharmacy store management (e.g., business procedures, information related to systems or equipment within the pharmacy, classification of pharmacy stores, etc.). For this reason, according to the embodiment, the "stored data DT n However, as an example, "pharmacy store operation information is also included" is also provided.

[0092] When preparing multiple stored data of different information types based on any two or more (or all) types of information exemplified in the embodiment, these stored data may be stored in a single vector DB main body so that they can be searched for vectors. This allows for unified management of different information types, enabling cross-sectional searches.

[0093] 4. Explanation of "Supplementary Information" in the embodiment 4-1. Examples of supplementary information In the embodiment, "supplementary information" is used in the supplementary process S3a of the output process S3 or the supplementary process S1b of the acquisition process S1 (see FIG. 8). The "supplementary information" can be various information that is useful for the information processing device 1 to provide effective output presentation depending on the usage situation.

[0094] For example, when a Q&A system is provided according to an embodiment, attributes of the user U who is asking the question may be used as supplemental information. This allows the generative AI model 4 to generate an appropriate answer, for example. Furthermore, if the user U is a pharmacist, the premise of the question (e.g., patient information when providing a prescription drug, or the individual specific circumstances of each pharmacy, etc.) may be used as supplemental information. This allows the generative AI model 4 to generate a more appropriate answer that takes into account the premise information (context), for example.

[0095] For the sake of convenience, the following will be categorized into groups H1 to H4 in Table 4, and specific information that can be mainly used in each group will be listed as an example. However, please note that the grouping is just an example, and each piece of information in each group may be diverted to another group depending on the intended use.

[0096] [Table 4]

[0097] Specifically, in group H1, at least the following information can be arbitrarily adopted. Management information such as patient ID Patient's gender or age, etc. - Patient's allergy history, current illness history, past medical history, history of side effects from medications, history of concomitant medications, etc. - Information regarding precautions for patients regarding medication (e.g., drug name, dosage, administration, medication history, etc.) Other medical information of the patient Patient's medication history

[0098] Specifically, in group H2, at least the following information can be arbitrarily adopted. - Stock of medicines used · Wholesale of employed pharmaceuticals - Dispensing record of the drug used (e.g. number of patients, etc.) - Expected dispensing of the adopted medicine (e.g., number of expected patients, etc.) - Appropriate inventory of used pharmaceuticals

[0099] Specifically, in group H3, at least the following information (H3a) or (H3b) can be arbitrarily adopted. (H3a) Pharmacy management information (e.g., patient ID, staff ID, etc.), store local information (e.g., handover notes related to a specific patient, etc.), etc. (H3b) Attribute information of the inquirer (e.g., medical professionals or medical staff) (e.g., indicators of job proficiency, years of employment, etc.)

[0100] Specifically, in group H4, at least the following information (H4a) or (H4b) can be arbitrarily adopted. (H4a) Various information about the pharmacy where the system will be implemented (e.g., staff information, partner information, system vendor information, etc.) (H4b) Background information of the pharmacy (e.g., quantitative information such as the number of prescriptions filled at the pharmacy, the classification of dispensing fees to which the pharmacy belongs, etc.)

[0101] 4-2. How to obtain supplementary information In the embodiment, as an example, as shown in Figures 5 to 8 as "obtaining necessary information," supplementary information is obtained by communication with the outside (e.g., input means 2) in supplementary processing S1b and S3a. There are no limitations on the route by which supplementary information can be obtained, and it should be noted that there are various variations, including the examples of (R1) to (R3) below.

[0102] (R1) For example, information linked to a user U can be used as supplementary information. For example, information linked to the login account of each user U can be used as supplementary information. The supplementary information may include, for example, differences in the attributes of the user U who made the input IN, such as their job (e.g., pharmacist or medical office worker), responsibilities, years of service, years of experience, etc. The supplementary information may also be information that the user U can set themselves (e.g., language used, etc.).

[0103] (R2) The supplementary information may be acquired in real time or non-real time (offline). For example, it may be any information that can be acquired when the user U uses the system provided by the information processing device 1. For example, inquiry information from any pharmacy staff, external organization, patient, etc. acquired by the user U via a chatbot or email, etc. may be used as the supplementary information.

[0104] (R3) Information may be acquired automatically or manually. Automatically acquired information may be information acquired, for example, via an arbitrary database or via another arbitrary system (including another AI model). Manually acquired information may be information entered by the user U himself or herself or by another person who is a conversation partner, for example, via a UI such as a chatbot.

[0105] 5. Explanation of input IN in the embodiment 5-1.Example of input IN It should be noted that input IN (see FIGS. 4-8) of an embodiment may include a wide range of information directly or indirectly related to the products or operations of the pharmacy.

[0106] For example, when a Q&A system is provided by the information processing method of the embodiment, the input IN may be, for example, a question keyword or question sentence related to a pharmacy product or pharmacy operations. In this case, the content of the question varies depending on which user U (e.g., a pharmacy staff member in a broad sense, such as a pharmacy clerk or a pharmacist) asks the question and in what situation. The content of the question varies widely, for example, whether the user wants to know information about a specific pharmacy product (e.g., medicines, etc.), how to treat patients, pharmacy procedures or how to operate a business system, or how to proceed with general prescription work or interactions with external parties. The information processing method of the embodiment can meet these diverse needs. Furthermore, when a recommendation system is provided by the information processing method of the embodiment, the input IN may be, for example, any term, phrase, or sentence related to a pharmacy product or pharmacy operations, thereby satisfying needs such as searching for information related to a certain pharmacy product or pharmacy operations.

[0107] Some examples of input IN information types (type, content) are shown below. Prescription information (e.g., paper or electronic prescriptions) Patient Information Pharmacy product information (e.g., information on medicines, quasi-drugs, and other products sold by pharmacies) Patient information requests (e.g., questions, consultations, searches, recommendation requests, etc., the same applies to the following items) · Pharmacy staff requesting information Information requests or guidance from pharmacists Prescription medications and patient conditions Drug name or abbreviation

[0108] 5-2. Input IN acquisition route, etc. A variety of variations can also be adopted for the acquisition route or format of the input IN. As an example in the embodiment, the input IN may include at least one of the following (A1) to (A3). (A1) Input operation by user U (A2) Information output from a computer (any computer) (A3) Information obtained through the digitization of paper media

[0109] Regarding (A1) above, the input means 2 may accept manual text input operations or file uploads by the user U via, for example, any user interface (eg, chatbot, etc.), and use these as input IN.

[0110] Regarding (A2) above, the information processing system 100 of the embodiment may be a standalone system or may be linked to any other system. For example, when the information processing system is linked to another system, any data transmitted from the other system may be treated as input IN. In this case, an interface (API) connecting the information processing device 1 to the other system serves as the input means 2. Specifically, for example, the "information output from a computer" may be information (e.g., patient information, medication history information, etc.) acquired from any computer system (e.g., electronic medical record system, electronic medication history system, etc.). As another example, any information (e.g., information on pharmaceuticals or quasi-drugs, etc.) acquired from any database system (e.g., drug database system, etc.) may be adopted as input IN.

[0111] Regarding (A3) above, information obtained by digitizing paper media (e.g., prescriptions, etc.) may be used as input IN. As an example, the paper media may be imaged using any terminal (e.g., a scanner or the camera function of a smartphone, etc.), and text information may be extracted from the image by applying OCR. The image may be converted into a PDF, and the OCR-generated text-embedded PDF may be used as input IN. There is no limitation to paper media, and examples include prescriptions, drug instructions, drug information sheets, medicine envelopes, and dispensing statements.

[0112] The data format of the input IN is not limited to text, and may be non-text (e.g., audio, image, video, etc.) as long as it is compatible with the vector DB system 3. In particular, image search may be used to identify identical or similar pharmacy products based on, for example, any pharmacy product image (e.g., medicine, its box, manual, etc.). The same applies to the generative AI model 4.

[0113] 6. Supplementary Notes, etc. Various embodiments are illustrated below. The embodiments shown below can be combined with each other.

[0114] [Appendix 1] An information processing method, comprising: causing a computer to execute an acquisition process; In the acquisition process, search results are acquired from a vector database based on the input. The vector database stores the stored data in a vector searchable manner; An information processing method, wherein each of the stored data includes pharmacy product information or pharmacy business information.

[0115] [Appendix 2] 10. The information processing method according to claim 1, each of the stored data includes an embedding vector; each said embedding vector is a vector representation of some or all of the stored data; An information processing method in which the associated data linked to the embedded vector includes data indicating the contents of the pharmacy product information or the pharmacy business information.

[0116] [Appendix 3] 10. The information processing method according to claim 1 or 2, further executing a sorting process by the computer; In the sorting process, a plurality of search results obtained from the vector database are arranged in order of similarity with the input.

[0117] [Appendix 4] An information processing method according to any one of Supplementary Note 1 to Supplementary Note 3, further causing the computer to perform output processing; An information processing method, wherein the output process outputs the search results directly or via a generative AI model.

[0118] [Appendix 5] 5. The information processing method according to claim 4, An information processing method in which the output processing outputs the generated results obtained from the generative AI model by providing supplementary information or the input to the generative AI model together with the search results.

[0119] [Appendix 6] An information processing method according to any one of Supplementary Note 1 to Supplementary Note 5, The stored data is categorized and stored, In the acquisition process, the stored data similar to the input is searched for using the category.

[0120] [Appendix 7] An information processing method according to any one of Supplementary Note 1 to Supplementary Note 6, The information processing method, wherein the input includes at least one of the following (A1) to (A3): (A1) User Input (A2) Information output from a computer (A3) Information obtained through the digitization of paper media

[0121] [Appendix 8] An information processing method according to any one of Supplementary Note 1 to Supplementary Note 7, An information processing method, wherein the pharmacy product information includes at least one of the following (D1) and (D2): (D1) At least one of information on the pharmacy product itself, patient information related to the pharmacy product, prescription information, and instruction information (D2) Management information about pharmacy products

[0122] [Appendix 9] An information processing method according to any one of Supplementary Note 1 to Supplementary Note 8, An information processing method, wherein the pharmacy business information includes at least one of the following (D3) and (D4): (D3) Information on patient care (D4) At least one of information regarding internal pharmacy operations and information regarding collaboration with outside pharmacies

[0123] [Appendix 10] A program that causes at least one computer to execute the information processing method according to any one of Supplementary Note 1 to Supplementary Note 9.

[0124] [Appendix 11] An information processing device that executes each process of the information processing method according to any one of Supplementary notes 1 to 9.

[0125] Although the embodiments have been described above, they are presented as examples and are not intended to limit the scope of the invention. The novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made. The embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0126] 1: Information processing equipment 2: Input method 3: Vector DB system 4: Generative AI model 6: Communication network 10: Communications Department 11: Storage section 12: Control section 12a: Acquisition processing unit 12a1: Input processing section 12b: Sorting processing section 12c: Output processing section 12c1: Supplementary processing section 12c2: Model input / output processing section 13: Output section 14: Input section 15: Communication bus 100: Information Processing Systems

Claims

1. An information processing method, The acquisition process and the output process are performed by a computer; In the acquisition process, search results are acquired from a vector database based on the input. The vector database stores the stored data in a vector searchable manner; The stored data includes stored data of pharmacy product information and stored data of pharmacy business information, The type of pharmacy product information includes product information; The types of pharmacy business information include information related to dispensing and information related to reception or accounting, The stored data is categorized and stored in the vector database; In the acquisition process, the category is used to search for the stored data similar to the input; The categories are predefined to classify the pharmacy product information and the pharmacy business information, The categories include a category related to dispensing and a category related to reception or accounting, In the output process, the search results are output via a generating AI model, An information processing method in which the output processing outputs the generated results obtained from the generative AI model by providing the input together with the search results to the generative AI model.

2. 2. The information processing method according to claim 1, In the output process, the supplemental information and the input are provided to the generative AI model together with the search results, thereby outputting the generated results obtained from the generative AI model; An information processing method, wherein the supplemental information includes user attribute information about a job.

3. A program configured to cause at least one computer to execute the information processing method according to claim 1 or 2.

4. An information processing device that executes each process of the information processing method according to claim 1 or 2.

Images