JP2021120803A - Creation support device, creation support method, and creation support program - Google Patents

Abstract

To efficiently recognize specifications about an invitation to bid, and to support creation of an accurate proposal.SOLUTION: A creation support device accepts a new inquiry including information of requirement specifications, supports creation of a proposal including information of bidding specifications corresponding to the new inquiry, and comprises: an arithmetic unit which obtains similarities between contents included in the new inquiry and contents included in past inquiries or past proposals, and uses the similarities to extract contents useful for creation of the proposal from history data associated with the past inquiries; and a providing unit which uses a result of the arithmetic unit to create and provide auxiliary data for use in creation of the proposal.SELECTED DRAWING: Figure 2

Description

The present invention relates to a creation support device, a creation support method, and a creation support program that support the creation of a proposal for bidding with respect to the specifications presented at the time of requesting a bid (Invitation To Bid: ITB).

In a wide variety of industries, the implementation of new projects may involve competitive bidding, in which bids are requested and the contractor of the project is determined from among multiple bidders. Specifically, a plurality of bidders who are participants in the bidding prepare a proposal for the specifications created by the orderer. The ordering party compares the proposals prepared by a plurality of bidders and selects the contractor of the project.

By the way, in the specifications, there are many items and amounts of description contents. In order to create a proposal, bidders need to accurately grasp the necessary information from the large amount of information contained in the specifications. For example, if a bidder makes an omission or mistake in reading the description of the specification, the content of the specification cannot be accurately grasped. In this case, it is not possible to create an accurate proposal. Specifically, if there is an error in the preparation of the proposal, it will be difficult to implement it after receiving the order. In addition, if you make a mistake in making an estimate and the estimated amount is less than the actual cost, you may lose money when you receive an order for the project. On the other hand, if the estimated amount is incorrect more than the actual cost, the estimated amount will be higher than that of other bidders, and the project may not be ordered. Therefore, it is preferable that the bidder accurately grasps the contents of the specification, which has a large amount of description, and then prepares an accurate proposal.

In addition, there may be deficiencies in the specifications. Even in such a case, it is difficult to prepare an accurate proposal. Therefore, bidders are required to identify the deficiencies in the specifications, fill in the deficiencies themselves, and create an accurate proposal.

Against the background described above, bidders are required to efficiently grasp specifications and prepare proposals. Patent Document 1 describes that a past construction plan is used for creating a construction plan. However, the description of Patent Document 1 does not consider how to efficiently grasp the construction condition document for creating the construction plan and how to deal with the case where the construction condition document is inadequate or insufficient.

Japanese Unexamined Patent Publication No. 2019-200499

In view of the above, the present invention efficiently grasps the specifications presented at the time of requesting a bid, and supports the creation of an accurate proposal based on the specifications, a creation support device, a creation support method, and a creation support program. I will provide a.

The creation support device according to the present disclosure is a creation support device that accepts a new inquiry document containing information on the required specifications and supports the creation of a proposal including information on the bidding specifications corresponding to the new inquiry document. Regarding the content contained in the new inquiry, the degree of similarity with the content contained in the past inquiry or the past proposal is obtained, and the similarity is used to make a proposal from the historical data associated with the past inquiry. It is provided with an arithmetic unit that extracts contents useful for creating a document, and a providing unit that generates and provides auxiliary data used for creating a proposal by using the result of the arithmetic unit.

According to the preparation support device, the preparation support method, and the preparation support program of the present disclosure, it is possible to support the efficient grasp of the specifications and the preparation of an accurate proposal.

It is a schematic diagram explaining the bidder and the bidder of a bid. It is a schematic diagram explaining the processing of a bidder of a bid. This is an example of an inquiry. This is an example of a proposal. It is a block diagram explaining the structure of the creation support device. It is the schematic explaining the prediction of the loss risk in the creation support device. It is a flowchart explaining the process of the loss risk prediction in a creation support device. It is a schematic diagram explaining the identification of lack information in a creation support device. It is a flowchart explaining the specific processing of lack information in a creation support apparatus. It is a schematic diagram explaining the identification of the difference part in the inquiry document in the creation support device. It is a flowchart explaining the specific process of the difference part in the inquiry document in the creation support device. Following FIG. 10A, it is a flowchart explaining the specific processing of the difference part in the inquiry document.

The present invention supports the preparation of a proposal for bidding for a specification relating to a request for bidding.

In the present disclosure, the "required specification" is a project specification presented to a bidder at the time of requesting a bid. The required specifications include the contents and conditions regarding the project and transaction desired by the ordering party.

The "inquiry document" is a specification document, specifically, a document of the required specifications. Inquiries include texts and charts.

The "bid specification" is the specification of the project planned by the bidder to realize the required specification included in the inquiry. The bid specifications include content and conditions relating to the projects and transactions planned by the bidder. The bidding specifications also include estimates for the project being planned.

A "proposal" is a document of bidding specifications. Proposals include texts and charts.

"Natural language processing" is to have a computer process natural language. In the present disclosure, for example, a corpus generated as a language material in advance is used, and natural language processing is used when extracting feature words by morphological analysis, syntactic analysis, or the like. By "morphological analysis" or "syntactic analysis", a sentence can be divided into words that compose the sentence.

A "characteristic word" is a word that is the subject or purpose of a sentence.

"Similarity" indicates the degree to which one content of a document is similar to another. The degree of similarity is calculated using the inter-word distance obtained based on the difference vector between the feature words in the vector space.

<Overview of creation support device>
The outline of the creation support device will be described with reference to the schematic views shown in FIGS. 1 and 2. The creation support device is used by bidders who are participants in bidding and provides useful data for creating proposals for bidding. As shown in FIG. 1, in bidding and bidding, the ordering party A provides the bidding bidders B to D of the bidding regarding the project with the required specification “inquiry document A”. Each bidder B to D who has received this inquiry prepares "proposals B to D", which are bid specifications, in order to participate in the bidding, and provides them to the ordering party A. The ordering party compares the proposals B to D provided by each bidder and determines the contractor of the project. The contractor determined by this will implement the project according to the proposal. For example, when bidders B to D are plant engineering companies, the project is plant design and construction, etc. In the following, the creation support device according to the embodiment will be described as being used by the bidder B.

As shown in FIG. 2, the bidder B is divided into, for example, an "overall design department" related to the management of the entire design and a plurality of "individual design departments" related to the individual design of each department. In addition, the bidder B associates the plurality of past inquiries B1 to B3 with the "proposal", "loss information", "review", etc. generated in accordance with the past inquiries B1 to B3, and has a history. It is stored and managed as data 121, and it can be read out and used as needed.

The "inquiry form" includes the specifications required in bidding, as in the example shown in FIG. 3A. Further, FIG. 3B is an example of a “proposal” generated based on the “inquiry document” of FIG. 3A. Since the proposal is generated in response to the inquiry, the content of the proposal may correspond to any of the contents of the inquiry. For example, the content of "1. Proposal item of Na compounding device, (1) -1 pretreatment set" in the proposal of FIG. 3B is "1. Na compounding device, (1) -1 before" of the inquiry document of FIG. 3A. It corresponds to "processing".

The "loss information" included in the historical data 121 is information on the loss caused by the execution of the past project. According to the proposal, the history of the loss when the project is implemented can be used, for example, to create a proposal for a future project.

Further, the "review book" included in the history data 121 is information regarding problems and the like that have occurred in the execution of the past project. According to the proposal, the history of problems that occurred when the project was executed can be utilized, for example, in creating a proposal for a future project.

Since the "loss information" and the "review report" are the data generated in connection with the proposal, they will be described as "related data" as necessary.

As shown in FIG. 2, first, the bidder B accepts a new "inquiry form A" generated by the orderer A (S1).

The bidder B uses the "inquiry form A" received in step S1 in the overall design department to grasp the entire design, analyze the history of the past design, and determine the policy regarding the design (S2). ). In addition, the overall design department transmits the information obtained in step S2 to each individual design department.

Subsequently, the bidder B uses the information obtained in step S2 in the individual design department to examine, adjust, design, and prepare an estimate for the content of the proposal (S3). In addition, each individual design department transmits the information obtained in step S3 to the overall design department.

After that, the bidder B integrates the information obtained in step S3 in the overall design department, and carries out the final design, the final estimate creation, and the like (S4).

Further, the bidder B generates a bid specification "proposal B" according to the result of step S4 in the overall design department and provides it to the orderer (S5).

Here, the bidder B uses the creation support device according to the embodiment to (1) predict the loss risk, (2) identify the missing information, and (3) make a difference in the inquiry document, as shown in FIG. To identify. The bidder B uses the information obtained by carrying out these (1) to (3) as "auxiliary data". As a result, the bidder B can accurately generate the "proposal B".

In FIG. 2, the description is given in a format in which "auxiliary data" is used in the overall design department, but the present invention is not limited to this. That is, the individual design department may use the auxiliary data when processing the proposal contents of step S3, such as examination, adjustment, design, and estimation creation.

<Creation support device>
As shown in FIG. 4, the creation support device 1 according to the embodiment includes a reception unit 111, a calculation unit 112, and a provision unit 113. Further, the creation support device 1 transmits / receives data to / from a control unit 11 such as a CPU that executes data processing, a storage unit 12 such as a RAM or ROM that stores data, and an external device via a network. It includes a communication unit 13 to execute, an input unit 14 used for data input, and an output unit 15 used for data output.

In the creation support device 1, the creation support program P stored in the storage unit 12 is executed, so that the control unit 11 executes processing as the reception unit 111, the calculation unit 112, and the provision unit 113.

In addition to the creation support program P, the storage unit 12 has a corpus 122 formed by historical data 121 regarding the history of past inquiries and proposals, and words extracted from past inquiries and proposals by morphological analysis. And the inquiry form 123 received from the ordering party are stored.

The creation support device 1 may be composed of a plurality of computers. For example, a part of the data may be stored in another device and configured to read the data. Specifically, the history data 121 and the corpus 122 may be stored in an external storage device, and the creation support device 1 may read data from these storage devices and use the data.

The history data 121 is data in which an inquiry document is associated with a plurality of documents created based on the inquiry document. The history data 121 can include data related to a plurality of projects carried out in the past by the bidder B who uses the creation support device 1. For example, the history data 121 is based on the "inquiry form" acquired by the bidder B in the past, the "proposal form" generated based on the inquiry form, and the proposal form, as described above with reference to FIG. It is the data to be associated with the "related data" generated in the above. For example, by associating the inquiry document, the proposal, and the related data with the same identification information, the original inquiry document can be identified from the proposal and the related data. It is also possible to identify the relevant data generated based on the inquiry from the inquiry. As described above, since the history data 121 is data related to the projects implemented in the past, the proposal included in the history data 121 is not only created by the bidder B but also adopted. That is, the proposal included in the history data 121 is a proposal that was able to win in the past bid. By using the information about the project that was able to win the past bid included in the history data 121 in this way, the creation support device 1 is useful for creating a valid proposal for a new inquiry and implementing the project. It becomes possible.

In addition to the above-mentioned "loss information" and "review report" using FIG. 2, the "related data" may include other information generated based on the inquiry document. Other information is, for example, "minutes" of a meeting or the like held when a project is implemented in response to an inquiry. Alternatively, the other information is a "report" created when the project was implemented in response to the inquiry. In addition, other information may include a "work report" created for each work related to project implementation, a "trouble report" created for troubles that occur during project implementation, and the like.

The corpus 122 is, for example, a structure of words extracted by morphological analysis using a character string included in the history data 121. Further, the corpus 122 may be generated by using data such as an existing glossary or a dictionary. The generation of the corpus 122 may be executed in the creation support device 1. Further, the creation support device 1 may acquire and use the corpus 122 generated by an external device.

The reception unit 111 receives a new target inquiry form from the ordering party. Further, the reception unit 111 stores the received inquiry document 123 in the storage unit 12. Here, the received inquiry document 123 will be described as an inquiry document A.

The calculation unit 112 uses a similarity measurement algorithm or a difference measurement algorithm to extract the contents necessary for generating a proposal for a new inquiry by natural language processing.

Specifically, the arithmetic unit 112 extracts words from the new inquiry book received by the reception unit 111 by using techniques such as morphological analysis and syntactic analysis. Further, the calculation unit 112 specifies a feature word for a new inquiry document in consideration of the frequency of appearance of each word, the importance, a predetermined stop word, and the like. Further, the calculation unit 112 also extracts the feature words of the history data 121 in the same manner, and obtains the similarity between the feature words of the new inquiry document and the feature words of the history data 121 by using the similarity measurement algorithm. .. In the similarity measurement algorithm, as described above, the similarity is calculated using the inter-word distance obtained based on the difference between the feature words in the vector space. After that, the calculation unit 112 extracts necessary information according to the degree of similarity.

The providing unit 113 outputs auxiliary data based on the information extracted by the calculation unit 112. The bidder B can accurately create the proposal by using the auxiliary data provided by the providing unit 113.

As described above in FIG. 2, according to the creation support device 1, (1) prediction of loss risk, (2) identification of insufficient information, and (3) identification of differences in the inquiry document are carried out. Hereinafter, each process in the creation support device 1 will be specifically described.

<< 1. Loss Risk Forecast >>
The “prediction of loss risk” in the creation support device 1 will be described with reference to the conceptual diagram of FIG. 5 and the flowchart shown in FIG. If there is a loss in a past project, it is more likely that a similar loss will occur if a similar project is carried out. Whether or not it is similar to the project in which the past loss occurred can be judged by comparing the new inquiry document with the past inquiry document by the "similarity measurement algorithm". Therefore, the creation support device 1 compares the old and new inquiries, and provides information on the loss caused in the project based on the similar inquiries in the past as auxiliary data because it may occur in the new project. As a result, the bidder B can prepare a new proposal by referring to the loss that has occurred in the past so that the same loss does not occur.

As shown in FIG. 6, first, the reception unit 111 receives a new inquiry (S11). In addition, the reception unit 111 stores the new inquiry document 123 received in the storage unit 12. In FIG. 5, the new inquiry form received is referred to as "inquiry form A".

The calculation unit 112 identifies the feature word of the new inquiry document received in step S11 (S12). Specifically, the arithmetic unit 112 extracts a word from a new inquiry book by using morphological analysis, syntactic analysis, and the like, and identifies a feature word. The calculation unit 112 uses the corpus 122 stored in the storage unit 12 to extract the feature words.

Further, the calculation unit 112 selects one of the past inquiries included in the history data 121 (S13). The method of selecting the inquiries in step S13 is not limited, but in order to select the inquiries in order, for example, the identification numbers may be selected in descending order or ascending order.

Using the feature word specified in step S12, the calculation unit 112 obtains the similarity between each content of the new inquiry document received in step S11 and each content of the past inquiry document selected in step S13 (). S14). In addition, the calculation unit 112 uses the feature words extracted from the past inquiries for the similar calculation. The feature words extracted from the past inquiries are also those extracted by using the corpus 122, but those extracted in the past and stored in the storage unit 12 may be used.

Further, each content for which the similarity is calculated in step S14 is an element constituting an inquiry document such as "sentence a", "sentence b", "sentence c", and "drawing k" in FIG. .. In step S14, since the similarity is calculated using the feature words, here, only the "sentence" that can include the feature words is defined as the "content of the inquiry document", and the "drawing" is referred to as the "inquiry document". It shall not be included in "Content". For example, the calculation unit 112 uses paragraph-based sentences as targets for calculating the degree of similarity.

The calculation unit 112 determines whether or not the processing of steps S13 and S14 has been completed for all the past inquiries included in the history data 121 (S15).

If all the inquiries have not been completed (NO in S15), the calculation unit 112 repeats the processes of steps S13 to S15. Here, when repeating step S13, a past inquiry document that has not been selected in the previous processing is selected.

When all the inquiries are completed (YES in S15), the calculation unit 112 extracts the past inquiries similar to the new inquiries by using the similarity calculated in the processing of each step S14 (YES in S15). S16). For example, the calculation unit 112 uses the similarity of each content to select the inquiry document having the most content having a similarity equal to or higher than a predetermined value. FIG. 5 shows an example in which "inquiry document B2" is selected from a plurality of past inquiry documents B1 to B3.

The calculation unit 112 extracts the loss information associated with the past inquiry document extracted in step S16 from the history data 121 (S17). In FIG. 5, the loss information associated with the extracted past inquiries is referred to as “loss information B2”. This will identify loss information for projects similar to the new inquiry.

The providing unit 113 provides the loss information extracted in step S17 as auxiliary data (S18). For example, the providing unit 113 may provide all of the extracted loss information as auxiliary data. Alternatively, the providing unit 113 may provide only the portion of the extracted loss information that is strongly related to the new inquiry document as auxiliary data.

In the above description, the number of inquiries to be selected in step S16 is not limited. For example, the creation support device 1 may select a plurality of inquiries. In that case, the calculation unit 112 generates auxiliary data using a plurality of loss information corresponding to each inquiry.

The creation support device 1 thus provides the loss information as auxiliary data. As a result, the bidder B who uses the creation support device 1 can prevent in advance the risk of loss incurred in the project when creating a proposal based on the new inquiry.

<< 2. Identifying missing information >>
The “identification of insufficient information” in the creation support device 1 will be described with reference to the conceptual diagram of FIG. 7 and the flowchart shown in FIG. Information is often lacking in new inquiries. When a proposal is generated using such an inquiry, it may not be possible to create an accurate proposal due to lack of information. On the other hand, past information may be used to supplement the lack of information in new inquiries. For example, if the description of the inquiry is vague, it is difficult to specify what kind of proposal should be prepared from such an inquiry. In such a case, the creation support device 1 uses the similarity measurement algorithm to extract the same description as the vague description of the new inquiry from the past inquiry, and the proposal associated with this inquiry. And related data can be used to create new proposals. That is, as described above, since the proposal included in the historical data 121 was able to win the bid in the past, the proposal, the inquiry document associated with this proposal, and the related data are accurately proposed. It can be effectively used as information for creating a book.

Therefore, the creation support device 1 compares the old and new inquiries, considers that the information associated with similar inquiries in the past may be necessary for a new project, and provides it as auxiliary data. As a result, the bidder B can create a proposal without any shortage by referring to the past information in creating a new proposal.

The process of "identifying the missing information" executed by the creation support device 1 will be specifically described. In the flowchart of FIG. 8, the above-described processing using FIG. 6 will be omitted by using the same reference numerals as those attached to the processing of FIG. Further, in the actual processing in the creation support device 1, these processings may be the same processing. Specifically, these processes may be executed as common processes for the above-mentioned "prediction of loss risk" and "identification of insufficient information", and may be executed separately only for different processes. Further, in FIG. 8, a process different from the process of FIG. 6 is distinguished by underlining the reference numerals.

As shown in FIG. 8, when the processes of steps S11 to S16 are executed and a past inquiry document similar to the new inquiry document is extracted, the calculation unit 112 associates the past inquiry document extracted in step S16 with the past inquiry document. The past proposals to be made are extracted (S21). In FIG. 7, the extracted past proposals are referred to as “proposals B2”.

The calculation unit 112 obtains the degree of similarity between each content of the new inquiry document received in step S11 and each content of the past proposal document extracted in step S21 (S22). The calculation unit 112 uses feature words extracted from past proposals for similar calculations. The feature words extracted from the past proposals are also those extracted using the feature words extracted in step S12 and the corpus 122, but the ones extracted in the past and stored in the storage unit 12 are used. It may be something to do. Here, too, the “content” is an element that constitutes the inquiry or proposal, and is limited to the “sentence” that can include the characteristic words, and does not include the “drawing”. For example, the calculation unit 112 uses paragraph-based sentences as targets for calculating the degree of similarity.

The calculation unit 112 uses the similarity to extract contents similar to the contents of the new inquiry from the past proposals (S23). Specifically, from the past proposals, the contents that are equal to or higher than the predetermined contents, which are determined to have a high degree of similarity to the contents of the new inquiry, are extracted. As a result, it is possible to extract the part that is strongly related to the new inquiry from the past proposals.

Further, the calculation unit 112 extracts the content related to the content extracted in step S23 from the related data (S24). In the example of FIG. 7, the content related to the content extracted from the “proposal B2” is extracted from the “loss information B2” and the “review document B2” associated with the “proposal B2”. For example, in the history data 121, each content of the proposal B2 is associated with the content of the related data that is strongly related in advance, and this association is used to relate the content of the proposal B2 and the content that is highly relevant. Extract from the data. As a result, it is possible to extract the part that is strongly related to the new inquiry from the past related data.

The providing unit 113 provides the content extracted from the past proposal in step S23 and the content extracted from the related data in step S24 as auxiliary data (S25). At this time, by including in the auxiliary data which content of the new inquiry document is strongly related to each content, it is possible to facilitate the utilization of the content provided as the auxiliary data. Further, when a plurality of inquiries are selected in step S16, the calculation unit 112 shall generate auxiliary data in step S25 using the plurality of proposals and related data corresponding to each inquiries.

The creation support device 1 thus provides the contents included in the proposal and the related data as auxiliary data. As a result, the bidder who uses the creation support device 1 can use the past information to supplement the lacking information of the new inquiry when creating the proposal based on the new inquiry. ..

<< 3. Identification of differences in the inquiry form >>
The “identification of the difference in the inquiry document” in the creation support device 1 will be described with reference to the conceptual diagram of FIG. 9 and the flowcharts shown in FIGS. 10A and 10B. There may be differences in the content described in different parts of the new inquiry. For example, the terms that should be unified in different paragraphs in the inquiry document may be different, or the terms that should be unified in the text and the drawing may be different in the inquiry document. In other words, there may be inconsistencies in the content of the inquiry. When a proposal is generated using such an inquiry, there is a possibility that an accurate proposal cannot be prepared due to inconsistency in information. For such problems, it may be possible to use the past history to predict inconsistencies in new inquiries. Therefore, the creation support device 1 uses an algorithm for measuring the degree of difference to identify the possibility of a contradiction in the new inquiry document, which is specified by comparing the new inquiry document with the past inquiry document and the proposal document. , Provided as auxiliary data. As a result, the bidder B can prepare a proposal that is not affected by the contradiction in the inquiry document by referring to the past information in the preparation of the new proposal. In addition, in the proposal, it is possible to point out the contradiction of the inquiry document and urge the ordering party to clarify or correct the description of the inquiry document.

The process of "identifying the difference in the inquiry document" executed by the creation support device 1 will be specifically described. In the flowchart of FIG. 10A, the above-described processing using FIG. 6 will be omitted by using the same reference numerals as those attached to the processing of FIG. Further, in the actual processing in the creation support device 1, these processings may be the same processing. Specifically, even if these processes are executed as common processes for "prediction of loss risk" and "identification of differences in inquiries" described above, and only different processes are executed separately. good. Further, in FIGS. 10A and 10B, the processing different from the processing of FIG. 6 is distinguished by underlining the reference numerals.

Further, as shown in FIG. 10A, the calculation unit 112 selects one of the past proposals included in the history data 121 (S31). The method of selecting the proposals in step S31 is not limited, but since the proposals are selected in order, they may be selected in descending or ascending order of the identification numbers, for example. The processes after step S31 may be executed after the process of step S12 is completed, or may be executed in parallel with the processes of steps S11 and S12.

The calculation unit 112 obtains the similarity between each drawing of the new inquiry document received in step S11 and each drawing of the past proposal selected in step S31 (S32). Usually, in the inquiry document, the drawing is image data having only two-dimensional information, and in the proposal document, it is three-dimensional CAD data created based on the image data of the inquiry document. Therefore, the calculation unit 112 needs to be able to obtain the degree of similarity even in drawings of such different formats. Although details are omitted, for example, there is an example in which the degree of similarity is obtained by comparing headings representing the features of the drawings. Another example is to obtain the degree of similarity by extracting the features of the drawing and comparing the features with each other. In extracting the features of the drawing, the 2D CAD is automatically generated from the 3D CAD data created based on the image data of the inquiry document, and it is used as learning data to efficiently extract the features. The accuracy can be improved.

The calculation unit 112 determines whether or not the processing of steps S31 and S32 has been completed for all the past proposals included in the history data 121 (S33).

If all the proposals have not been completed (NO in S33), the calculation unit 112 repeats the processes of steps S31 to S33. Here, when repeating step S31, a past proposal that has not been selected in the previous processing is selected.

When all the proposals are completed (YES in S33), the calculation unit 112 uses the similarity calculated in the process of step S32 to draw a drawing of the past proposal similar to each drawing of the new inquiry. Extract (S34). Specifically, the drawing having the highest degree of similarity to the drawing of the new inquiry is extracted from the past proposals. As a result, it is possible to extract a drawing that is strongly related to the drawing of the new inquiry from the past proposals. FIG. 9 is an example in which “drawing p” of “proposal B2” is extracted as a drawing similar to “drawing k”. Note that FIG. 9 shows an example in which only "drawing p" is extracted as a drawing similar to "drawing k", but since "inquiry document A" includes a plurality of drawings, in step S34, a plurality of drawings are present. Drawings can be extracted.

Subsequently, the calculation unit 112 extracts a sentence associated with the drawing extracted in step S34 from the past proposal (S35). Here, the text associated with the drawing in the proposal is text such as the title of the drawing and the description of the drawing. In the example shown in FIG. 9, it is assumed that "sentence n" is extracted as the sentence associated with "drawing p" in "proposal B2". When a plurality of drawings are extracted in step S34, the calculation unit 112 extracts sentences associated with the extracted drawings from the past proposals.

In addition, the calculation unit 112 extracts a past inquiry document associated with the past proposal document including the drawing extracted in step S34 (S36). In the example shown in FIG. 9, the “inquiry document B2” associated with the “proposal document B2” is extracted. When a plurality of drawings are extracted in step S34, the calculation unit 112 extracts past inquiries associated with the proposal including the drawings for each of the extracted drawings.

Next, the calculation unit 112 obtains the similarity between each sentence of the new inquiry document received in step S11 and each sentence of the past inquiry document extracted in step S36 (S37). The calculation unit 112 uses the feature words extracted in step S12 and the feature words extracted from the past inquiries to calculate the similarity. The feature words extracted from the past inquiries are also those extracted by using the corpus 122, but those extracted in the past and stored in the storage unit 12 may be used. In FIG. 9, for example, the degree of similarity between each sentence of the “inquiry document A” and each sentence of the “inquiry document B2” is obtained. Here, when a plurality of past inquiries are extracted in step S36, the calculation unit 112 obtains the similarity between each sentence of the "inquiry book A" and each sentence of the extracted plurality of past inquiries. ..

After that, as shown in FIG. 10B, the calculation unit 112 uses the similarity obtained in step S37 to extract sentences similar to each sentence of the new inquiry document from the past inquiry documents (S38). For example, the calculation unit 112 extracts a sentence having a degree of similarity with the sentence of the new inquiry document of a predetermined value or more from the past inquiry document. Since the inquiry document contains a plurality of sentences, the calculation unit 112 may extract a plurality of sentences as sentences similar to the sentences of the new inquiry document from one past inquiry document. In FIG. 9, for example, the calculation unit 112 extracts the “sentence k” of the “inquiry book B2” as a sentence similar to the “sentence b” of the “inquiry book A”. Further, when the similarity with each sentence of a plurality of past inquiries is obtained in S37, the calculation unit 112 extracts similar sentences from each of the inquiries. That is, the calculation unit 112 may extract a plurality of sentences from a plurality of past inquiries.

In addition, the calculation unit 112 calculates the degree of similarity between the text in the past inquiry document extracted in step S36 and the text associated with the past drawing extracted from the past proposal in step S35 (S39). The calculation unit 112 also uses the feature words extracted by using the corpus 122 to calculate the similarity. As in the case described above, the feature word used in step S39 may be one extracted in the past and stored in the storage unit 12. In FIG. 9, the calculation unit 112 obtains, for example, the degree of similarity between each sentence of the “inquiry document B2” and the “sentence n” of the “proposal document B2”.

Subsequently, the calculation unit 112 uses the similarity obtained in step S39 to select a sentence in the past inquiry document that is similar to the sentence associated with the past drawing extracted in step S35 (S40). For example, the calculation unit 112 extracts from the past inquiry document a sentence having a degree of similarity with the sentence associated with the past drawing of a predetermined value or more. In the case of the example of FIG. 9, it is said that the "sentence n" associated with the past drawing "drawing p" and the "sentence k" in the "inquiry book B2" are similar, and the "sentence k" from the "inquiry book B2" to the "text". k ”is selected.

Further, the calculation unit 112 uses the sentence selected in step S40 as a sentence to be replaced with a drawing of a new inquiry document for determining the contradiction (S41). In the case of the example of FIG. 9, the sentence is a sentence in which the “sentence k” is replaced with the “drawing k”. Although FIG. 9 illustrates an example of extracting only one sentence, it is possible to extract a plurality of sentences from a plurality of inquiries.

Subsequently, the calculation unit 112 evaluates the contradiction between similar contents of the new inquiry document by using the difference degree measurement algorithm (S42). Specifically, the calculation unit 112 evaluates the contradiction between the contents of the new inquiry document A by using the feature words of each sentence in the new inquiry document A. The feature word used in this case is also the one used by using the corpus 122. At this time, with respect to the drawing, inconsistency with other sentences is evaluated by using the feature words of the replaced sentence in step S41. In the example of FIG. 9, the "drawing k" of the "inquiry document A" is replaced with the "sentence k" of the "inquiry document B2", and the sentence b is considered to be similar to the sentence k, and the contradiction is evaluated. In addition, for evaluation of contradiction, for example, Multi-NLI is used.

The providing unit 113 generates auxiliary data using the list obtained in descending order of the evaluation of the contradiction in step S42 (S43).

Further, the providing unit 113 provides the auxiliary data generated in step S42 (S43). In the example of the auxiliary data shown in FIG. 9, the contradiction between "sentence a" and "sentence c" is "74.2%", and the contradiction between "sentence a" and "sentence q" is "64.6%". The contradiction between "sentence b" and "drawing k" is cited as "63.8%". By replacing a drawing with a sentence, whether it is a contradiction between sentences or a contradiction between drawings, it is evaluated in the same way as a contradiction between sentences.

At this time, the providing unit 113 can determine the display method according to the degree of contradiction. Auxiliary data with different colors may be generated for values with a high possibility of inconsistency and for values with a low possibility of inconsistency. Further, the providing unit 113 may include the auxiliary data only when the possibility of the contradiction becomes more than a predetermined value.

In this way, the creation support device 1 replaces the drawing with the text included in the past inquiry document and uses it for calculating the similarity, evaluates the contradiction between the drawing and the text, and provides it as auxiliary data. As a result, the bidder who uses the creation support device 1 grasps the contradiction even if the contradiction is included in the contradiction when creating the proposal based on the new contradiction, and responds to this contradiction. It becomes possible.

As described above, according to the creation support device according to the embodiment, when creating a proposal for the required specifications in bidding, auxiliary data effective for creating the proposal is provided. As a result, the user of the creation support device can accurately create the proposal regardless of the deficiency or deficiency in the inquiry document.

<Modification example 1>
When the calculation unit 112 selects a past inquiry or a past proposal to be compared from the history data 121, the calculation unit 112 uses the past inquiry or the past inquiry generated by the same ordering party as the new inquiry. You may want to select the associated past proposal. Past inquiries generated by the same orderer are likely to have similar content trends. Therefore, it may be possible to efficiently extract necessary information by preferentially processing the past inquiries generated by the same orderer or the proposals associated with the past inquiries.

For example, the history data 121 associates the past inquiries with the identification information of the creator of the past inquiries. Then, when calculating the similarity, the calculation unit 112 calculates only the similarity with the past inquiry or the past proposal associated with the identification information of the creator of the new inquiry.

Further, for example, when calculating the similarity, the calculation unit 112 may place another order when calculating the similarity with the past inquiry or the past proposal associated with the identification information of the creator of the new inquiry. Give a higher weight compared to the inquiries or proposals associated with the person's identification information. As a result, the information of the creator of the new inquiry is preferentially extracted.

<Modification 2>
The calculation unit 112 has described an example in which a feature word is used by using natural language processing when a past inquiry document or a past proposal document similar to a new inquiry document is similar from the history data 121. On the other hand, the calculation unit 112 may use a trained model in which the relationship between a plurality of past inquiries and the similarity between the inquiries is learned. By using this trained model, the calculation unit 112 can make a new inquiry document extract a similar past inquiry document from a plurality of past inquiry documents. A neural network, a support vector machine, or the like can be used for learning the model.

Further, the calculation unit 112 may use a trained model in which the relationship between the past inquiry document and the loss amount generated in the past inquiry document is learned. By using this trained model, the calculation unit 112 can estimate the amount of loss for the new inquiry for the new inquiry. Further, the providing unit 113 uses the estimated loss amount as auxiliary data. The bidder can prepare a proposal with less loss by using the estimated loss amount as a reference for preparing the proposal. At this time, it is also effective to learn the relationship of the loss amount of the contents of the past inquiries, not the whole of the past inquiries. For example, in the creation support device 1, by creating auxiliary data that the amount of loss is large for a certain content in the new proposal, care should be taken not to cause a loss when creating the proposal. Can be encouraged.

The creation support device of the present disclosure is useful in preparing a proposal for a specification, for example, because it participates in bidding.

1 Creation support device 11 Control unit 111 Reception unit 112 Calculation unit 113 Provision unit 12 Storage unit 121 History data 122 Corpus 123 Inquiry document 13 Communication unit 14 Input unit 15 Output unit P Creation support program

Claims (8)

It is a creation support device that accepts new inquiries containing information on required specifications and supports the creation of proposals containing information on bidding specifications corresponding to the new inquiries.
Regarding the content included in the new inquiry, the degree of similarity with the content contained in the past inquiry or the past proposal is obtained, and the similarity is used from the historical data associated with the past inquiry. , A calculation unit that extracts useful content for creating proposals,
A providing unit that generates and provides auxiliary data used for creating a proposal by using the result of the calculation unit, and a providing unit.
Creation support device equipped with. The calculation unit
The degree of similarity between each content contained in the new inquiry document and each content contained in a plurality of past inquiries is determined, and a past inquiry document similar to the new inquiry document is selected and selected. Using the past inquiries, the proposal corresponding to the past inquiries, or the related data corresponding to the past inquiries, the contents not included in the new inquiries are extracted.
The providing part
Output the extracted contents as auxiliary data,
The creation support device according to claim 1. The content is a sentence composed of a plurality of words.
The calculation unit uses feature words extracted from the words included in the sentence by using a corpus that defines the relationship between a plurality of words, in calculating the similarity of each content.
The creation support device according to claim 2, wherein the creation support device is characterized in that. The calculation unit
Find the degree of similarity between the drawings included in the new inquiry and each drawing included in the past proposals.
Replacing the drawings contained in the past proposal similar to the drawings contained in the new inquiry and the text contained in the corresponding past inquiries with the drawings contained in the new inquiry,
Evaluate the contradictions of each content in the new inquiry,
The providing part
The creation support device according to claim 1, wherein auxiliary data is generated and output using the evaluation results. The content is a sentence or drawing composed of a plurality of words.
The calculation unit uses feature words extracted from the words contained in each sentence using a corpus that defines the relationship between a plurality of words, and replaces each drawing in calculating the similarity of each content. The creation support device according to claim 4, wherein the feature words extracted by using the corpus are used from the words included in each sentence included in the past inquiry document. The past inquiry document associates the identification information of the creator of the past inquiry document,
When calculating the similarity, the calculation unit preferentially calculates the similarity with the past inquiry or the past proposal associated with the identification information of the creator of the new inquiry, and extracts the contents. The creation support device according to any one of claims 1 to 5. It is a creation support method that accepts new inquiries containing information on required specifications and supports the creation of proposals containing information on bidding specifications corresponding to the new inquiries.
Regarding the content included in the new inquiry, the degree of similarity with the content contained in the past inquiry or the past proposal is obtained, and the similarity is used from the historical data associated with the past inquiry. , Steps to extract useful content for creating proposals,
Steps to generate auxiliary data to be used for creating proposals using the extraction results,
Creation support method with. It is a creation support program that accepts new inquiries containing information on required specifications and supports the creation of proposals containing information on bidding specifications corresponding to the new inquiries.
On the computer
Regarding the content included in the new inquiry, the degree of similarity with the content contained in the past inquiry or the past proposal is obtained, and the similarity is used from the historical data associated with the past inquiry. , Steps to extract useful content for creating proposals,
Steps to generate auxiliary data to be used for creating proposals using the extraction results,
Creation support program that realizes.

Images