JP2020119128A - Information processing device, information processing method and program - Google Patents

Abstract

To provide a mode for efficiently proceeding a business negotiation.SOLUTION: An information processing device includes: a generation part 10 for generating an object business negotiation vector by using a word included in an object business negotiation document inputted by a user and weight associated with the word; a selection part 20 for selecting one or more comparison business negotiation vectors close to the object business negotiation vector from a plurality of comparison business negotiation vectors for the comparison business negotiation document by using a word included in a comparison business negotiation document about a business negotiation and weight associated with the word, stored in the storage parts 40; and an output part 30 for outputting business negotiation information on a business negotiation having the comparison business negotiation vector selected by the selection part 20.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing device, an information processing method, and a program that provide information on business negotiations.

Conventionally, it has been attempted to carry out sales activities efficiently. In Patent Document 1, a first company that manages a sales relationship among a plurality of companies including a first company, which is a business of its own company, and a second company, which is a sales destination of a product processed by the first company. Management information and second management information for managing the sales relationship between the company's sales department and a plurality of companies included in the first management information, and read the first management information and the second management information. Based on, it is a diagram showing the sales relationship between multiple companies, and a diagram in which the correspondence relationship between a given sales department of the company and at least one of the multiple companies can be visually generated and output By doing so, it is proposed that the sales relations of the sales department of the company can be grasped and the sales activities can be carried out efficiently.

JP, 2008-92366, A

The present invention provides a mode in which a business negotiation is efficiently carried out by an approach different from the conventional one.

The information processing apparatus according to the present invention is
A generation unit that generates a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word;
From the plurality of comparison negotiation vectors for the comparison negotiation document, which are stored in the storage unit and included in the comparison negotiation document related to the negotiation and the weight associated with the word, to the target negotiation vector. A selection unit for selecting one or more comparison negotiation vector close to each other,
An output unit for outputting business negotiation content information regarding a business negotiation having the comparative business negotiation vector selected by the selection unit;
May be provided.

In the information processing device according to the present invention,
Each of the comparison negotiation document and the target negotiation document includes a title of the negotiation,
The generation unit may generate the comparison negotiation vector and the target negotiation vector using the title of the negotiation.

In the information processing device according to the present invention,
The output unit may output, as the business negotiation content information, personnel information of a person in charge who has a business negotiation having the comparison business negotiation vector selected by the selection unit.

In the information processing device according to the present invention,
The personnel information of the person in charge may include means for accessing the person in charge.

In the information processing device according to the present invention,
The output unit may output, as the business negotiation content information, the order amount, the budget amount, or the competitive company in the business negotiation having the comparison business negotiation vector selected by the selection unit.

In the information processing device according to the present invention,
The output unit outputs, as the negotiation content information, the related person information related to the person in charge of the other party in the business negotiation having the comparison business negotiation vector selected by the selection unit, and the person associated with the related business person information. You may output the personnel information of.

In the information processing device according to the present invention,
Each of the comparison negotiation document and the target negotiation document includes a negotiation title and a negotiation report,
The generation unit uses the title of the negotiation and the negotiation report, and generates the comparative negotiation vector and the target negotiation vector using the first weight for the title of the negotiation and the second weight for the negotiation report,
The values of the first weight and the second weight may be changed according to the progress of the negotiation.

In the information processing device according to the present invention,
Each of the comparison negotiation document and the target negotiation document includes a negotiation title and a negotiation report,
The generation unit may determine the progress status of the business negotiation from the quantity of the business negotiation report, and generate the comparison business negotiation vector and the target business negotiation vector using the title of the business negotiation, the business negotiation report, and the progress status of the business negotiation.

The information processing method according to the present invention is
A step of generating a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word by the generation unit;
From the plurality of comparison negotiation vectors for the comparison negotiation document stored in the storage unit, using the words included in the comparison negotiation document regarding the negotiation and the weights associated with the word, the selection unit Selecting one or more comparison negotiation vectors close to the target negotiation vector,
A step of outputting, by the output unit, negotiation content information regarding a negotiation having the comparison negotiation vector selected by the selection unit;
May be provided.

The program according to the invention is
A program for installing in an information processing device,
The information processing device installed with the program
A generation unit that generates a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word;
From the plurality of comparison negotiation vectors for the comparison negotiation document, which are stored in the storage unit and included in the comparison negotiation document related to the negotiation and the weight associated with the word, to the target negotiation vector. A selection unit for selecting one or more comparison negotiation vector close to each other,
An output unit for outputting business negotiation content information regarding a business negotiation having the comparative business negotiation vector selected by the selection unit;
May be provided.

In the present invention, a target generated from a plurality of comparison negotiation vectors stored in the storage unit by using a word included in the target negotiation document input by the user and a weight associated with the word When the selecting unit selects one or more comparison negotiation vectors close to the negotiation vector and outputs the negotiation content information regarding the negotiation having the comparison negotiation vector, it is close to the target negotiation document input by the user. The user can confirm the business negotiation content information having the information, and the user can obtain useful information when proceeding with the business negotiation.

FIG. 1 is a diagram showing a processing flow of a learning process in the information processing apparatus according to the first embodiment of the present invention. FIG. 2 is a diagram showing a processing flow of a similar negotiation calculation process in the information processing apparatus according to the first embodiment of the present invention. FIG. 3 is a schematic block diagram showing the configuration of the information processing apparatus according to the first embodiment of the present invention. FIG. 4 is a diagram showing an example of information output by the information processing apparatus according to the first embodiment of the present invention. FIG. 5 is a flow showing the flow of processing in the information processing apparatus according to the first embodiment of the present invention. FIG. 6 is a diagram showing an aspect of a screen output by the information processing apparatus according to the first embodiment of the present invention. FIG. 7 is a diagram showing an example of functions realized by the information processing device according to the first embodiment of the present invention. FIG. 8 is a diagram showing a processing flow of a learning process in the information processing apparatus according to the second embodiment of the present invention. FIG. 9 is a diagram showing a processing flow of a similar negotiation calculation process in the information processing apparatus according to the second embodiment of the present invention. FIG. 10 is a diagram for explaining that the same wording (wording “fire power”) has different meanings depending on the context.

First Embodiment <<Configuration>>
The information processing device according to the present embodiment may be composed of one device or may be composed of a plurality of devices. Further, when the information processing device is composed of a plurality of devices, each device does not need to be provided in the same space such as the same room, but may be provided in different rooms, different buildings, different areas, and the like. In this embodiment, "and/or" may be used for description, but the word "or" includes the meaning of "and". That is, the word A or B is a concept including A alone, B alone, and A and B.

As illustrated in FIG. 3, the information processing apparatus stores a word and a weight associated with the word in association with each other, and associates the word included in the comparative negotiation document regarding the business negotiation with the word. A storage unit 40 that stores a plurality of comparison negotiation vectors for the comparison negotiation document generated by using the weight, a word included in the target negotiation document input by the user, and a weight associated with the word. , A selection unit 20 for selecting one or more comparison negotiation vectors close to the target negotiation vector, and a negotiation content regarding a negotiation having the comparison negotiation vector selected by the selection unit 20. And an output unit 30 that outputs information. The target negotiation document generated by the generation unit 10 may also be stored in the storage unit 40.

In the present application, a comparison negotiation vector “close” to the target negotiation vector means a comparison negotiation vector whose similarity is equal to or more than a threshold value when compared with the target negotiation vector. Therefore, “selecting one or more comparison negotiation vectors close to the target negotiation vector” means selecting a comparison negotiation vector whose similarity is equal to or more than a threshold value with reference to the target negotiation vector. .. This similarity may be determined using the inner product of the target negotiation vector and the comparison negotiation vector. Each of the target negotiation vector and the comparison negotiation vector may be a feature vector.

Each of the comparison negotiation document and the target negotiation document may include the title of the negotiation. The generation unit 10 may generate the comparative negotiation vector and the target negotiation vector using the title of the negotiation.

The output unit 30 may output, as the business negotiation content information, the personnel information of the person in charge who has the business negotiation having the comparison business negotiation vector selected by the selection unit 20. The personnel information may include the name and contact information of the person in charge, as well as information regarding the position.

The personnel information of the person in charge may include means for accessing the person in charge. A telephone number including an e-mail address and an extension number may be used as the access means.

The output unit 30 may output, as the negotiation content information, the order amount, the budget amount, and/or the competitors in the negotiation having the comparison negotiation vector selected by the selection unit 20.

When selecting the comparative negotiation vector in the selection unit 20, filtering may be performed according to the type of industry, or the comparative negotiation vector may be selected from a population of industries that match or are similar to the industry input by the user. However, such filtering is not always necessary. If such a mode that filtering is not applied is adopted, it is advantageous in that the contents of business negotiations in different industries can be referred to. In particular, it is useful that the person in charge in a different industry and the user have an interview so that the opportunity to provide a business based on broader knowledge can be obtained by the collaboration of the human resources, which has not been possible in the past. .. From this point of view, it may be possible to perform control so as to select a comparative negotiation vector from a population of a business type different from the business type input by the user when performing filtering. It may be registered in the storage unit 40 in advance as a list as to whether they are similar industries or different industries.

The output unit 30 outputs, as the negotiation content information, the related person information related to the person in charge of the other party in the business negotiation having the comparison negotiation vector selected by the selection unit 20, and the person in charge associated with the related person information. You may output the personnel information of. The related person-in-charge information may include the name and contact information of the person in charge of the other party, as well as information regarding the post. The person in charge associated with the person in charge of the other party may be, for example, the boss of the person in charge of the other party.

Each of the comparison negotiation document and the target negotiation document may include the title of the negotiation and the negotiation report. The generation unit 10 may generate the comparative negotiation vector and the target negotiation vector using the title of the negotiation and the negotiation report, and using the first weight for the title of the negotiation and the second weight for the negotiation report. The first weight and the second weight may have different values.

The generation unit 10 may determine the progress status of the business negotiation from the quantity of the business negotiation report, and generate the comparative business negotiation vector and the target business negotiation vector using the business title, the business negotiation report, and the business progress of the business negotiation. The quantity may be determined based on the date, or may be increased by "1" each time the date is different. For example, if there are three business negotiation reports on different dates, the quantity of the business negotiation may be “3”.

Even if the generation unit 10 has an artificial intelligence function and creates a feature vector of a target word from the closeness of words used in a large amount of text data (for example, words used on the WEB). Good. The proximity of words is determined by the artificial intelligence function of the generation unit 10 by being inferred from how the words are used (the context of the words, etc.) on the website (including SNS sites, etc.) on the WEB (learning). However, it may be updated at any time or as appropriate. Information about the closeness of words may be stored in the storage unit 40.

The generation unit 10 may generate the weight of the word in the document based on the frequency of use of the word used in the document regarding the negotiation. When determining the weight, consider whether or not the word is a commonly used word, lower the weight for "particles" (for example, "0"), and increase the weight for words that are rarely used. Good. Words that are frequently used in business negotiation documents may be given lower weight.

When a certain amount of record data is accumulated in the storage unit 40, the past record data previously input may be deleted. In addition, actual data older than a certain period (for example, 5 years) may be deleted. When a vector is generated based only on actual data within a certain period (for example, within 5 years) as described above, it is advantageous in that a vector can be generated using relatively new actual data.

As an example of the artificial intelligence function, a classifier using a machine learning method may be used. This classifier may determine the adoption variable to be used and its coefficient by machine learning technology so as to generate a vector with high accuracy from past performance data, for example. Then, by applying the determined adoption variable and its coefficient to the target business negotiation information, similar business negotiation information may be selected by the selection unit 20. The artificial intelligence function may perform regression analysis, decision tree analysis, or the like. In the regression analysis, a predetermined function for generating a vector with high accuracy is determined from past performance data, and by incorporating information about a target business negotiation into this function, a vector related to the business negotiation is generated. In the decision tree analysis, the hiring variables and their coefficients to be used are determined so that the probability that the target event will occur is the most accurate (to generate a vector with high accuracy), and the hiring variables and coefficients that have been used are used. By incorporating information about the target business negotiation into the decision tree, a vector about the business negotiation is generated. With respect to the machine learning technique, various models can be adopted, for example, a logistic regression model, a random forest model, a tree model, etc. can be adopted.

Each member such as the generation unit 10, the selection unit 20, the output unit 30, and the storage unit 40 of the present embodiment may be realized by one or a plurality of IC chips or electronic modules, or realized by a circuit configuration. May be. In addition, one electronic member may serve as a control unit and serve the functions of the generation unit 10, the selection unit 20, and the output unit 30.

Next, a more specific mode will be described. In the following, a description will be given focusing on a mode in which a business negotiation name is used as business negotiation information, but this is merely an example and the present invention is not limited to this. As the negotiation information, a project name may be used, or a negotiation report such as a sales report may be used.

[Learning process]
The learning process is shown in FIG. Such learning may be performed every day, or may be performed at an appropriate timing such as once a week or once a month.

In the morpheme analysis (S11) in FIG. 1, all negotiation information such as negotiation names stored in the negotiation database, which is a kind of the storage unit 40, is subjected to morphological analysis and divided into word units. Of the words obtained as a result, part-of-speech information may be referred to and only nouns may be left. Furthermore, words that do not contribute to the intended degree of similarity may be excluded based on a preset excluded word list. In this morphological analysis, negotiation information such as the negotiation name and the exclusion word list may be input, and the word list existing in all past negotiations may be output.

In the word vectorization (S12) in FIG. 1, a word included in the negotiation information such as the negotiation name may be converted into a 200-dimensional numerical vector using Word2Vec or the like. Compare with a word vector that has been numerically vectorized from the WEB site such as Wikipedia (registered trademark) Japanese version, the word included in the word vector generated from the information from the WEB site adopts that vector, and Words that are not included may be word vectors vectorized from business negotiation information such as business negotiation names. Thus, the information on the WEB site may be prioritized over the business negotiation information such as the business negotiation name, but the business negotiation information such as the business negotiation name may be prioritized over the information on the WEB site. In this word vectorization, a word vector generated from a WEB site such as the word list and Wikipedia Japanese version may be input, and a table of words×200-dimensional numerical values may be output.

In the word importance calculation (S13) in FIG. 1, the importance of the word in each negotiation name may be calculated by TFIDF or the like. The degree of importance may be increased if the business negotiation information such as the business negotiation name includes many target words, and the degree of importance may be decreased if the target word is also included in other business negotiation information. In this word importance calculation, a word list existing in all past business negotiations may be input, and a business negotiation×word importance matrix and an IDF table for each word may be output.

In the negotiation vectorization (S14) in FIG. 1, the negotiation vector such as the negotiation name may be vectorized using the word vector and the word importance to generate the negotiation vector. The negotiation vector generated in this way may be stored in the storage unit 40. In this negotiation vectorization, a table of words×200-dimensional numerical value and an importance matrix of negotiations×word may be input and a table of negotiations×200-dimensional numerical value may be output.

In the negotiation vectorization, the following calculation formula may be used.
Is TF (Term Frequency),
Is IDF (Inverse Document Frequency). Also, t is a word vector, n _t,b is the number of appearances of the word t in the negotiation d,
Is the sum of the number of appearances of all the words in the negotiation d, N is the total number of negotiations, and df(t) is the number of negotiations in which the word t appears.

In the weighted word vectorization (S15) in FIG. 1, the IDF obtained during the word importance calculation may be multiplied as a weight to generate a weighted word vector. The weighted word vector thus generated may be stored in the storage unit 40. In this weighted word vectorization, a word×200-dimensional numerical value table and a word-by-word IDF table may be input, and a word×200-dimensional numerical value table (weighted) may be output.

[Similar business negotiation calculation process]
FIG. 2 shows a similar negotiation process. Such calculation may be started when the user inputs negotiation information such as a negotiation name.

In the morphological analysis (S21) in FIG. 2, the newly registered business negotiation information such as the business negotiation name is morphologically analyzed and divided into word units. Of the words obtained as a result, part-of-speech information may be referred to and only nouns may be left. Furthermore, words that do not contribute to the degree of similarity may be excluded based on a preset excluded word list. In this morphological analysis, negotiation information such as a new negotiation name may be input and a word list existing in the new negotiation may be output.

In the new negotiation vectorization (S22) in FIG. 2, the weighted word vectors of the words that appear in the negotiation information such as the new negotiation name may be added together to vectorize the new negotiation name. In this new negotiation vectorization, a word list and a table of words×200-dimensional numerical values (weighted) existing in the new negotiation may be input, and the new negotiation vector (200-dimensional) may be output. By thus generating the new negotiation vector using the weighted word vector, the accuracy in the similarity calculation can be improved.

In the similarity calculation (S23) in FIG. 2, the similarity may be calculated by taking the inner product of the new negotiation vector and the already generated (past) negotiation vector. In this similarity calculation, a table of new negotiation vector and negotiation×200-dimensional numerical value may be input, and a table of similarity for each negotiation may be output.

In the similar business negotiation presentation (S24) in FIG. 2, the selection unit 20 may determine a presentation candidate by a threshold value from those having a high similarity to the new business negotiation vector, or a predetermined number of candidates may be selected in descending order of the similarity. It may be determined as a presentation candidate. It is also possible to exclude, from the presentation candidates, the new business negotiation and the business negotiation with the same registrant. The similar business negotiation list obtained as a result may be presented. In this similar negotiation presentation, a table of the degree of similarity for each negotiation may be input, and a similar negotiation list may be output.

As in this embodiment, by performing word distributed expression and weighting based on appearance frequency and vectorizing business negotiation information such as business negotiation name, unlike keyword search, similarity can be achieved even if characters do not match. Can be evaluated.

As an example, when the negotiation name “automation of the factory” is input, the result including the name of the similar negotiation and the distance as shown in FIG. 4 is output. “,” “IoT introduction”, etc. will be listed.

"Method"
An example of the information processing method using the information processing apparatus according to the present embodiment will be described mainly with reference to FIG. It should be noted that an example is shown here, and all the aspects described in the above “configuration” and all the aspects described in the following “effect” can be applied in the “method”.

When the user inputs a new business negotiation from an input unit such as a terminal of a personal computer, information about the new business negotiation is registered in a business negotiation database, which is a kind of the storage unit 40. The input of information may be performed by voice.

The generation unit 10 accesses the negotiation database every predetermined time (for example, every 10 to 30 seconds, typically every 20 seconds), and confirms whether a new negotiation has been input. Note that the present invention is not limited to such an aspect, and when a new business negotiation is input, the generation unit 10 may automatically obtain information regarding the new business negotiation. When adopting the mode of accessing the negotiation database every predetermined time and acquiring the information of the new negotiation, the information of the new negotiation can be obtained by the simple cooperation with the negotiation database and the maintenance is easy. It is beneficial in terms.

When it is confirmed that the new negotiation has been input, the generation unit 10 extracts information about the new negotiation and calculates a similar negotiation similar to the new negotiation. Then, for example, the selection unit 20 selects a certain number of similar business negotiations (for example, the top five) or similar business negotiations having a degree of similarity equal to or more than a threshold value.

When the similar business negotiation is selected by the selection unit 20 in this manner, information regarding the selected similar business negotiation is transmitted to the user via the electronic mail. The link information is included in the e-mail, and the information as shown in FIG. 6 may be viewed by clicking the link information. In FIG. 6, the customer name, the P code that is the project code, the PM indicating the person in charge, the order amount, and the competitor are shown. The PM indicating the person in charge includes the link information, and clicking the link information may allow the information regarding the person in charge to be browsed. The information about the person in charge may include a facial photograph in addition to the access means. Being able to check such a facial photograph is useful in that it allows the person in charge to meet smoothly when actually having a meeting with the person in charge and that the user can judge whether or not he/she knows himself/herself from the facial photograph. Is. The information regarding the person in charge may include the year of joining the person in charge, the title, the qualification for holding, and the like. The user can determine whether or not the person in charge can easily access the person in charge when the mode that includes the position is adopted, and in the case that the mode that includes the year of hire is adopted, the person in charge can be in charge through the acquaintance of the person in charge. It is also possible to access the person.

All or part of the information about the business negotiation registered in the business negotiation database is acquired at a fixed time (for example, midnight, 2:00 am as an example). It should be noted that the information already acquired in the storage unit 40 may not be acquired, and only the information regarding the negotiation that has not been acquired yet (only the difference) may be acquired by the generation unit 10.

Then, the generation unit 10 performs machine learning to vectorize the negotiation using the information about the negotiation registered in the negotiation database thus acquired. By using the information thus vectorized, a business negotiation having a vector similar to the new business negotiation described above is selected.

Note that in the mode shown in FIG. 7, general operations such as negotiation similarity learning, new negotiation monitoring and comment posting, process monitoring to monitor whether new negotiation monitoring and comment posting are functioning, and password setting are managed. It is shown that operational management is performed.

"effect"
Next, the effects of the present embodiment having the above-described configuration, which are not described yet, will be mainly described.

From the plurality of comparison negotiation vectors stored in the storage unit 40, the target negotiation vector generated by using the word included in the target negotiation document input by the user and the weight associated with the word When the selecting unit 20 selects one or more close comparative negotiation vectors and outputs the negotiation content information regarding the negotiation having the comparative negotiation vector, the information close to the target negotiation document input by the user is selected. The user can confirm the negotiation content information that the user has, and the user can obtain useful information when proceeding with the negotiation. Further, by vectorizing, unlike the keyword search, it is possible to access negotiation information having high similarity even if characters do not match.

If the comparison negotiation document and the target negotiation document each include the title of the negotiation, and the aspect of generating the comparison negotiation vector and the target negotiation vector using the title of the negotiation (that is, the negotiation name) is adopted, the title of the negotiation is used as the basis. In addition, the user can access the business negotiation information having high similarity. In particular, it is useful to employ such an aspect in a company in which rules for naming business talks are given. Further, when the comparison negotiation vector and the target negotiation vector are generated using only the title of the negotiation, it is very useful in that the vector such as the feature vector can be created easily and quickly. In the case of using the title of the business talk, it is excellent in that the title can be the same even if the business talk progresses and the system can be simplified. Also, when using other business negotiation titles and other business negotiation information such as business negotiation reports, the weights (relative weight values) of business negotiation titles and other business negotiation information may be changed according to the progress status of business negotiations. Often, in the initial stage of the business negotiation, the first weight W1 for the title of the business negotiation is set to be heavier than the second weight W2 for other business negotiation information, and the business negotiation progresses (for example, when a predetermined number of business negotiations are made, or When a certain period has elapsed from the start), the first weight W1 for the title of the negotiation may be changed to be lighter than the second weight W2 for the other negotiation information.

When the aspect in which the personnel information of the person in charge who has the business negotiation having the comparison negotiation vector selected by the selection unit 20 is output as the negotiation content information is adopted, the person in charge can be accessed by using the personnel information. , The user can directly ask the person in charge about the contents of the business negotiation that the user wants to confirm. From this point of view, it is very useful when the personnel information of the person in charge includes access means to the person in charge.

When the mode of outputting the order amount or the budget amount in the business negotiation having the comparison business negotiation vector selected by the selection unit 20 as the business negotiation content information is adopted, the sense of scale of the business negotiation being compared can be grasped, and The user can determine whether or not it is close to the business negotiation that is underway. Further, in the case of adopting a mode of outputting the competitors in the negotiation having the comparison negotiation vector selected by the selection unit 20 as the negotiation content information, the user determines whether or not the competitor is the same as the negotiation in progress. I can judge. When the competitors are the same, the person in charge is asked what kind of correspondence the competitor has taken in the past business negotiations (comparative business negotiations), or obtained from the information stored in the storage unit 40. You can As an example, the user can understand whether the competitors will compete in terms of technology or price.

As the negotiation content information, the related person information related to the other person in charge in the negotiation having the comparison negotiation vector selected by the selection unit 20 is output, and the personnel information of the person associated with the related person information is output. In the case of adopting the mode described above, it is possible to compare the business negotiation that the user is promoting with the person in charge of the other party in the past business negotiation (comparative business negotiation) or the person in charge associated with the person in charge of the other party. When the boss of the person in charge of the other party is registered as the person in charge who is associated with the person in charge of the other party, the boss of the person in charge of the other party of the business negotiation that he is proceeding with matches the boss in the past business negotiation (comparative business negotiation) In this case, it is possible to refer to the business negotiations of the competitors in the past business negotiations.

Each of the comparative negotiation document and the target negotiation document includes the title of the negotiation and the negotiation report, and when the aspect of generating the comparison negotiation vector and the target negotiation vector by using the title of the negotiation and the negotiation report is adopted, the title of the negotiation is adopted. Since the vector (feature vector) is generated by using both the business negotiation report and the business negotiation report, it is possible to judge the similarity by paying attention to the content of the business negotiation and to present the similar business negotiation with higher accuracy. When adopting a mode in which the relative values of the weight of the negotiation title and the weight of the negotiation report are changed according to the progress of the negotiation, the title of the negotiation and the weight of the negotiation report can be changed according to the progress of the negotiation. It is possible to present similar business negotiations with higher accuracy.

In the case of adopting a mode in which the progress of the negotiation is judged from the quantity of the negotiation report, and the comparative negotiation vector and the target negotiation vector are generated using the title of the negotiation, the negotiation report and the progress of the negotiation, the title of the negotiation and the negotiation Since the vector (feature vector) is generated in consideration of the progress status of the business talk in addition to the report, it is useful in that a similar business talk that matches the progress status of the business talk that the user is promoting can be presented. Further, by using the quantity of the business negotiation report, it is useful in that the progress status of the business negotiation can be easily captured as information.

In addition to or in place of the quantity of the business negotiation report, the progress status of the business negotiation is judged using the length (year/month/day) elapsed from the start of the business negotiation, and the feature vector is generated using the progress status of the business negotiation. You may.

Second Embodiment Next, a second embodiment of the present invention will be described.

In the present embodiment, as shown in FIG. 8, the step of “word vectorization by language model” (S88) is performed in the learning process, and in the similar negotiation calculation process as shown in FIG. The step of "vectorizing negotiation name" (S92) and the step of "vectorizing new negotiation name by language model" (S98) are performed. In this embodiment, any configuration adopted in the first embodiment can be adopted in the second embodiment. The same reference numerals are given to the members described in the first embodiment for description.

In the morphological analysis (S81), word vectorization (S82), word importance calculation (S83) and weighted word vectorization (S85) in FIG. 8, the morphological analysis (S11), word vectorization (S12), The same processing as the word importance calculation (S13) and the weighted word vectorization (S15) is performed.

In the word vectorization (S88) by the language model in FIG. 8, words of negotiation information such as past negotiation names are input to the language model (neural network) that has been learned in advance, and dimensions are added to add feature vectors (dimension addition). (Feature vector) may be generated, and for example, a 1024-dimensional numerical vector may be generated. In the word vectorization by this language model, a word list and a learned language model (learned deep learning model) may be input and a table of words×1024-dimensional numerical values may be output.

When the negotiation vectorization using a language model (S88) is used, an external model for vectorizing a word may be constructed in advance from a large number of documents by deep learning. By applying this external model, a table of additional dimensions (for example, 1024-dimensional numerical values) may be output.

In the negotiation vectorization (S84) in FIG. 8, the negotiation name may be vectorized using the word vector and the word importance. Further, a table in which dimensions are added may be generated by vectorizing business negotiation information such as business negotiation name using a word vector created from a language model and linearly joining the two. In this negotiation negotiation vectorization, for example, a table of words×200-dimensional numerical values, a negotiation×word importance matrix, and a table of words×1024-dimensional numerical values may be input, and a table of negotiations×1224-dimensional numerical values may be output.

In FIG. 9, similar negotiations will be detected using the feature vector with the added dimension. In the morphological analysis (S91), the new negotiation vectorization by weighted word vector (S92) and the similar negotiation presentation (S94) in FIG. 9, the morphological analysis (S21), the new negotiation vectorization (S22) and the similar negotiation presentation in FIG. The same processing as (S24) is performed.

In the new negotiation vectorization by the language model in FIG. 9 (S98), the words appearing in the negotiation information such as the new negotiation name may be sequentially input to the language model learned in advance to obtain the word vector. The business negotiation information such as a new business negotiation name may be vectorized by adding them. In the new negotiation vectorization by the language model, the word list and the learned language model existing in the negotiation information such as the new negotiation may be input, and the new negotiation vector (1024 dimensions) by the language model may be output.

In the similarity calculation (S93) in FIG. 9, a 1224-dimensional new business negotiation vector in which the outputs of the new business negotiation vectorization using the weighted word vector and the new business negotiation vectorization using the language model are linearly joined and already generated (past) business negotiations The similarity may be calculated by taking the dot product of the vectors. In this similarity calculation, a new negotiation vector, a new negotiation vector based on a language model, and a table of negotiations×1224-dimensional numerical values may be input, and a similarity table for each negotiation may be output.

The number of dimensions can be increased by adopting the mode as in this embodiment. Therefore, it is possible to more appropriately deal with so-called polysemous words, and it is possible to add "depth" to the meaning of the wording even in the same wording from the relationship with the words before and after. That is, by considering the context, even the same character can be vectorized according to the meaning according to the context. In the aspect shown in FIG. 10, even with the same word “fire power”, the first region means “fire power” of a firearm, the second region means “fire power” of thermal power generation, and the third region means “fire power” during cooking. Although it has been shown to mean "fire power", the feature vector can be generated in consideration of such difference in meaning.

The description of each of the above-described embodiments and the disclosure of the drawings are merely examples for explaining the invention described in the claims, and the scope of the claims is defined by the description of the above-described embodiments or the disclosure of the drawings. The described invention is not limited.

10 generation unit 20 selection unit 30 output unit 40 storage unit

Claims (10)

A generation unit that generates a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word;
From the plurality of comparison negotiation vectors for the comparison negotiation document, which are stored in the storage unit and included in the comparison negotiation document related to the negotiation and the weight associated with the word, to the target negotiation vector. A selection unit for selecting one or more comparison negotiation vector close to each other,
An output unit for outputting business negotiation content information regarding a business negotiation having the comparative business negotiation vector selected by the selection unit;
An information processing device comprising: Each of the comparison negotiation document and the target negotiation document includes a title of the negotiation,
The information processing apparatus according to claim 1, wherein the generation unit generates the comparison negotiation vector and the target negotiation vector using the title of the negotiation. The information processing apparatus according to claim 1, wherein the output unit outputs, as the business negotiation content information, personnel information of a person in charge who has a business negotiation having the comparison business negotiation vector selected by the selection unit. The information processing apparatus according to claim 3, wherein the personnel information of the person in charge includes access means to the person in charge. The information according to any one of claims 1 to 4, wherein the output unit outputs, as the negotiation content information, an order amount, a budget amount, or a competitor in a negotiation having the comparison negotiation vector selected by the selection unit. Processing equipment. The output unit outputs, as the negotiation content information, the related person information related to the person in charge of the other party in the business negotiation having the comparison business negotiation vector selected by the selection unit, and the person associated with the related business person information. The information processing apparatus according to claim 1, wherein the information processing apparatus outputs the human resource information. Each of the comparison negotiation document and the target negotiation document includes a negotiation title and a negotiation report,
The generation unit uses the title of the negotiation and the negotiation report, and generates the comparative negotiation vector and the target negotiation vector using the first weight for the title of the negotiation and the second weight for the negotiation report,
7. The information processing apparatus according to claim 1, wherein the values of the first weight and the second weight are changed according to the progress of negotiations. Each of the comparison negotiation document and the target negotiation document includes a negotiation title and a negotiation report,
The said production|generation part judges the progress of a negotiation from the quantity of a negotiation report, and produces the said comparative negotiation vector and the said target negotiation vector using the title of a negotiation, the negotiation report, and the progress of a negotiation. The information processing apparatus according to any one of 1. A step of generating a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word by the generation unit;
From the plurality of comparison negotiation vectors for the comparison negotiation document stored in the storage unit, using the words included in the comparison negotiation document regarding the negotiation and the weights associated with the word, the selection unit Selecting one or more comparison negotiation vectors close to the target negotiation vector,
A step of outputting, by the output unit, negotiation content information regarding a negotiation having the comparison negotiation vector selected by the selection unit;
An information processing method comprising: A program for installing in an information processing device,
The information processing device installed with the program
A generation unit that generates a target negotiation vector using a word included in the target negotiation document input by the user and a weight associated with the word;
From the plurality of comparison negotiation vectors for the comparison negotiation document, which are stored in the storage unit and included in the comparison negotiation document related to the negotiation and the weight associated with the word, to the target negotiation vector. A selection unit for selecting one or more comparison negotiation vector close to each other,
An output unit for outputting business negotiation content information regarding a business negotiation having the comparative business negotiation vector selected by the selection unit;
A program that comprises: