JP5971794B2 - Patent search support device, patent search support method, and program - Google Patents

Description

  The present invention relates to a patent search support apparatus that supports patent search.

  Conventionally, devices for managing the results of patent searches have been developed (see, for example, Patent Document 1).

JP 2007-224004 A

  However, in the conventional patent search, manually classifying related patents and unrelated patents is a very time-consuming work, and if there is an error in the documents of patents once classified as unrelated patents, It was rare that it was reconsidered, and there was a problem that incorrect classification results remained.

A patent search support device according to the first aspect of the present invention includes a related patent document storage unit capable of storing one or more related patent documents, which are related patent documents determined to be related as a result of screening work in patent search, and a patent search The unrelated patent document storage unit that can store one or more unrelated patent documents that are determined to be irrelevant as a result of the sorting operation in the database, and one or more terms from each of the one or more related patent documents To acquire a related patent feature vector, which is a vector of each related patent document using one or more terms, for each related patent document, and to acquire one or more terms from one or more unrelated patent documents A feature vector acquisition unit that acquires, for each unrelated patent document, an unrelated patent feature vector that is a vector of each unrelated patent document using one or more terms, one or more related patent feature vectors, and one or more Non Using the consecutive patent feature vector, the decision part for determining the unrelated patent document corresponding to the unrelated patent feature vector that should belong to the class of the related patent, and the information specifying the unrelated patent document determined by the judgment part are output. And a patent search support device including an output unit.
With this configuration, related patents can be extracted from non-related patent documents. As a result, for example, in a patent search, when a patent to be searched is manually classified as either a related patent or an unrelated patent, it can be a related patent from among patent documents that were mistakenly classified as an unrelated patent. It is possible to obtain information that identifies highly patentable patent documents.

In addition, in the patent search support device of the second invention, in the first invention, the determination unit is a determination target that is a vector of non-related patent documents that is a determination target of whether or not to belong to a related patent class. A patent search support device that determines an unrelated patent document corresponding to an unrelated patent feature vector using one or more unrelated patent feature vectors excluding one or more related patent feature vectors and a judgment target unrelated patent feature vector. is there.
With this configuration, related patents can be extracted from non-related patent documents. As a result, for example, in a patent search, when a patent to be searched is manually classified as either a related patent or an unrelated patent, it can be a related patent from among patent documents that were mistakenly classified as an unrelated patent. It is possible to obtain information that identifies highly patentable patent documents.

In addition, the patent search support device of the third invention has one or more patent documents excluding one or more related patent documents and one or more unrelated patent documents with respect to the first or second invention. A new patent document storage unit that can store new patent documents is further provided, and the feature vector acquisition unit acquires one or more terms from one or more new patent documents, and uses each one or more terms to create each new patent document. A new patent feature vector, which is a vector of, for each new patent document, the determination unit uses one or more related patent feature vectors, or one or more related patent feature vectors and one or more unrelated patent feature vectors The new patent document corresponding to the new patent feature vector that should belong to the related patent class is determined, and the output unit outputs information identifying the new patent document that should belong to the related patent class determined by the determination unit It is a patent search support device.
With this configuration, it is possible to determine whether an unclassified patent document is a related patent or an unrelated patent. Thereby, for example, when a user classifies some patent documents into related patents or non-related patents, the user can acquire information that identifies patent documents that are likely to be related patents from unclassified patent documents.

In addition, in the patent search support device of the fourth invention, for any one of the first to third inventions, the determination unit includes one or more related patent feature vectors, or one or more related patent feature vectors and one or more. To determine the related patent document corresponding to the related patent feature vector that should not belong to the related patent class, and the output section should belong to the related patent class determined by the judgment section It is a patent search support device that outputs information specifying related patent documents that are not.
With this configuration, unrelated patents can be extracted from the patent documents of related patents. As a result, for example, when a patent subject to search is manually classified as either a related patent or an unrelated patent in a patent search, it may be an unrelated patent from among patent documents that were mistakenly classified as related patents. It is possible to obtain information that identifies highly patentable patent documents.

In addition, the patent search support device of the fifth invention includes a related patent document storage unit that can store one or more related patent documents that are related patent documents that are determined to be related as a result of screening work in the patent search; A new patent document storage unit that can store one or more new patent documents that are related patent documents excluding one or more related patent documents and one or more unrelated patent documents, and one or more from each one or more related patent documents A related patent feature vector, which is a vector of each related patent document using one or more terms, for each related patent document, and one or more terms from one or more new patent documents A new patent using a feature vector acquisition unit that acquires a new patent feature vector, which is a vector of each new patent document using one or more terms, for each new patent document, and one or more related patent feature vectors Patent search comprising a determination unit that determines whether a patent feature vector of a class belongs to the class of related patents, and an output unit that outputs information identifying new patent documents that the determination unit has determined to belong to the class of related patents It is a support device.
With this configuration, it is possible to determine whether an unclassified patent document is a related patent. Thereby, for example, when some patent documents are classified into related patents or non-related patents, it is possible to acquire information for identifying patent documents having a high possibility of related patents from unclassified patent documents.

In addition, the patent search support device of the sixth aspect of the invention provides at least one unrelated patent document that is a document of an unrelated patent that is determined to be unrelated as a result of the sorting work in the patent search for the fifth invention. The feature vector acquisition unit further acquires one or more terms from each of the one or more non-related patent documents, and uses each of the one or more terms to store each non-relevant patent document storage unit. An unrelated patent feature vector that is a vector of a patent document is acquired for each unrelated patent document, and the determination unit uses the one or more related patent feature vectors and the one or more unrelated patent feature vectors to create a new patent document. This is a patent search support device that determines whether a patent feature vector of a certain category belongs to a related patent class.
With this configuration, it is possible to determine whether an unclassified patent document is a related patent or an unrelated patent. Thus, for example, if some patent documents are classified as related patents or unrelated patents, patent documents that are likely to be related patents or are likely to be unrelated patents are identified from unclassified patent documents. Information to be acquired.

  According to the patent search support apparatus and the like according to the present invention, it is possible to determine whether a patent document is a related patent or an unrelated patent.

Conceptual diagram of a system including a patent search support device in the first embodiment Block diagram of patent search support apparatus in the embodiment The flowchart which shows operation | movement of the patent search assistance apparatus in the embodiment Flowchart showing the operation of the feature vector acquisition unit in the same embodiment The flowchart which shows operation | movement of the judgment part in the embodiment The figure which shows the example of the document of the patent stored in each patent document storage part in the embodiment The figure which shows the example of the feature vector of each patent document in the embodiment The figure which shows the example of a display of the screen which inputs the information which the reception part in the embodiment receives The figure which shows the example of a display of the screen which displayed the information which the output part in the same embodiment output Flowchart showing the operation of the patent research support device in the second embodiment The figure which shows the example of a display of the screen which displayed the information which the output part in the same embodiment output The figure which shows an example of the external appearance of the computer system in the said embodiment The figure which shows an example of a structure of the computer system in the said embodiment.

  Hereinafter, embodiments of a patent search support apparatus and the like will be described with reference to the drawings. In addition, since the component which attached | subjected the same code | symbol in embodiment performs the same operation | movement, description may be abbreviate | omitted again.

(Embodiment 1)
In the present embodiment, a patent document classified as an unrelated patent or an unclassified patent document is a related patent document by using the result of manually classifying the patent document into a related patent and an unrelated patent. The patent search support apparatus 1 that determines a possible patent document and outputs the result will be described.

  FIG. 1 is a conceptual diagram of a system including a patent search support apparatus 1 according to the present embodiment. In FIG. 1, a patent research support device 1 and one or more user terminals 3 are connected via a network 100. The network 100 is a wired or wireless communication line, such as the Internet, an intranet, a LAN (Local Area Network), a public telephone line, or the like. The user terminal 3 may be any terminal that can be connected to the network 100. For example, the user terminal 3 may be a desktop personal computer, a notebook personal computer, a smartphone, or a PDA.

  FIG. 2 is a block diagram of the patent search support apparatus 1 in the present embodiment. The patent search support apparatus 1 includes a reception unit 101, a related patent document storage unit 102, an unrelated patent document storage unit 103, a new patent document storage unit 104, a feature vector acquisition unit 105, a determination unit 106, and an output unit 107.

  The receiving unit 101 receives information specifying one or more related patent documents and information specifying one or more unrelated patent documents. Related patent documents are documents related to related patents. An unrelated patent document is a document of an unrelated patent. Related patents are patents related to the technology or invention to be investigated. An unrelated patent is a patent that is not related to the technology to be searched or the invention to be searched.

  The information specifying the related patent document received by the receiving unit 101 is information for specifying the patent document that is determined to be related as a result of the sorting operation in the patent search conducted manually. The information specifying unrelated patent documents received by the receiving unit 101 is information for specifying patent documents that are determined to be irrelevant as a result of screening work in a patent search conducted manually.

  The patent search is a search for a document related to a certain technology or a certain invention. For example, a patent search is a prior art search, an infringement prevention search, an invalidation data search, or the like. Prior art search refers to investigating whether prior art exists before filing a patent application. Infringement prevention investigation refers to investigating whether or not to infringe other patents before putting goods or services on the market. The invalidation document search is a search for invalidating the registered patent of another person. A patent document is information relating to an application document such as a patent filed at the JPO. The types of patent documents are published patent gazette, patent gazette, utility model published gazette, utility model registered gazette, published patent gazette, published utility model gazette, republished patent gazette, republished utility model gazette, and the like. The country where patent documents are issued is not limited to Japan, the United States, China, Europe, South Korea, etc. The language and data format of patent documents are not limited. The information specifying the patent document may be anything as long as it is information that can specify one patent document. The information for specifying the patent document may be, for example, the patent document itself or a patent document ID for uniquely specifying the patent document. The patent document ID may be an application number, a publication number, a registration number, or the like.

  In addition, the reception unit 101 stores related patent documents in the related patent document storage unit 102. The receiving unit 101 stores unrelated patent documents in the unrelated patent document storage unit 103. The receiving unit 101 stores the new patent document in the new patent document storage unit 104. New patent documents are unclassified patent documents excluding related patent documents and unrelated patent documents. The receiving unit 101 may receive one or more new patent documents, and obtains a new patent document by acquiring a patent document that does not include the received related patent document and unrelated patent document from a patent document storage unit (not shown). It may be a patent document. The patent document storage unit (not shown) may be an external patent search device or the like.

  When receiving the patent document ID, the receiving unit 101 may acquire a patent document from a patent document storage unit (not shown). At that time, the reception unit 101 may acquire patent documents for each patent document ID, or may acquire all patent documents at once. Note that a patent document storage unit (not shown) stores one or more patent documents.

  The accepting unit 101 may accept the determination result output by the output unit 107 (the determination result will be described later), and also accept each patent document reclassified according to the determination result. Also good. Each patent document that has been reclassified according to the determination result is one or more patent documents that are classified again by the user with reference to the determination result after the output unit 107 outputs the determination result.

  The accepting unit 101 usually accepts information specifying each patent document transmitted from the user terminal 3 via the network 100, but accepts information input from an input device such as a keyboard, mouse, touch panel, wired or It may be reception of information transmitted via a wireless communication line or reception of information read from a recording medium such as an optical disk, a magnetic disk, or a semiconductor memory.

  The related patent document storage unit 102 can store one or more related patent documents that are related patent documents. The related patent here is usually a patent that is determined to be related as a result of manual selection in a patent search, but may be a patent that is automatically determined to be related. Here, storage is a concept including long-term storage using a non-volatile recording medium and temporary storage using a volatile recording medium. The related patent document storage unit 102 is preferably a nonvolatile recording medium, but can also be realized by a volatile recording medium. The same applies to other storage units described later. The process for storing the related patent documents in the related patent document storage unit 102 is not limited. Normally, the related patent document storage unit 102 stores the related patent document received by the receiving unit 101. For example, the related patent document storage unit 102 stores the related patent document via a recording medium. Alternatively, related patent documents transmitted via a communication line or the like may be stored in the related patent document storage unit 102. The process of storing information in the storage unit is not limited, and the same applies to other recording media.

  The unrelated patent document storage unit 103 can store one or more unrelated patent documents that are unrelated patent documents. The unrelated patents here are usually patents that are determined to be unrelated as a result of manual selection in patent searches, but may be patents that are automatically determined not to be related.

  The new patent document storage unit 104 can store one or more new patent documents. A new patent document is a patent document excluding one or more related patent documents and one or more unrelated patent documents. The new patent document does not overlap with the patent documents stored in the related patent document storage unit 102 and the unrelated patent document storage unit 103.

  The feature vector acquisition unit 105 acquires one or more terms from one or more related patent documents, and uses the one or more terms to obtain related patent feature vectors, which are vectors of the related patent documents, Obtain for each document. The feature vector acquisition unit 105 normally acquires related patent feature vectors from two or more related patent documents. Furthermore, the feature vector acquisition unit 105 acquires one or more terms from one or more of each unrelated patent document, and uses the one or more terms to obtain an unrelated patent feature vector that is a vector of each unrelated patent document. Acquire for each unrelated patent document. Note that the feature vector acquisition unit 105 normally acquires unrelated patent feature vectors from two or more unrelated patent documents. The feature vector acquisition unit 105 acquires one or more terms from one or more new patent documents, and uses each of the one or more new patent documents to generate a new patent feature vector, which is a vector of each new patent document, for each new patent document. You may get to.

  The term acquired from the related patent document by the feature vector acquisition unit 105 may be, for example, a predetermined part of speech (for example, a noun), a predetermined part of speech, or a technical term. There may be other terms. If the term is a technical term, for example, the technical term is stored in a recording medium (not shown), and the term can be acquired by determining whether the technical term is included in the related patent document. The term may be acquired by using an algorithm for extracting technical terms from a document. For the latter method, see, for example, the following document. Hirokazu Ohata, Hiroshi Nakagawa, “Extraction of technical terms by different number of words”, Information Processing Society of Japan Research Report, 2000-NL-136, p. 119-126. Nakagawa Hiroshi, Mori Yasunori, Yumoto Yasuaki, “Terminology Extraction Based on Appearance Frequency and Connection Frequency”, Natural Language Processing, Vol. 10 No. 1, p. 27-45, January 2003. When acquiring the term, the feature vector acquisition unit 105 may use the TF value or the TF-IDF value as the importance level, and may acquire the term except for the less important term. Further, the documents that are targeted when calculating the IDF value may be all documents stored in the related patent document storage unit 102, the unrelated patent document storage unit 103, and the new patent document storage unit 104, When the processing is not performed on the patent document, all documents stored in the related patent document storage unit 102 and the unrelated patent document storage unit 103 may be used. The term of low importance may be a term whose importance value is a threshold value or less, or may be a predetermined number of terms arranged in order of importance. The threshold may be a predetermined value, for example, or may be a value obtained by multiplying the maximum value of importance by a number smaller than 1 (for example, 0.9 or 0.8). The predetermined number may be a predetermined number, for example, and a value smaller than 1 (for example, 0.01 or 0.001) is added to the total number of terms acquired by the feature vector acquisition unit 105. It may be a multiplied number. The feature vector acquisition unit 105 acquires a term from a character string including at least a part of a patent document. The character string including at least a part of a patent document may be, for example, the full text described in the patent document, a character string identified by a heading, or a combination of a plurality of headings. It may be a character string. The heading may be anything as long as it identifies an area in a patent document, such as the name of an invention, a claim, or a summary. In addition, although the feature vector acquisition part 105 described about the method of acquiring a term from a related patent document, it is the same also when acquiring a term from an unrelated patent document and a new patent document.

  Each related patent feature vector acquired by the feature vector acquisition unit 105 is usually a vector having dimensions corresponding to all the terms acquired by the feature vector acquisition unit 105. The element of the related patent feature vector may be, for example, the number of occurrences of the term, the TF value of the term, or the TF-IDF value of the term, and the term exists in the patent document. It may be a number indicating whether or not to be performed, specifically “1 (present)”, “−1 (not present)”, or the like. In addition, although the feature vector acquisition part 105 described about the method of acquiring a related patent feature vector, it is the same also when acquiring an unrelated patent feature vector and a new patent feature vector.

  The feature vector acquisition unit 105 may acquire a feature vector from the entire patent document (full text), or may acquire a feature vector from a part of the patent document. The part of the patent document is, for example, only an abstract, claims only, an abstract and claims, a description only, or an embodiment of the specification.

  Note that the feature vector acquisition unit 105 holds a related term dictionary (not shown), and may determine that the related terms stored in the related term dictionary are the same term and acquire the feature vector. . Here, the related term dictionary holds, for example, one or more related term groups having two or more synonyms and synonyms. For example, the related term group is “database, database, DB, repository, dictionary”. Then, the feature vector acquisition unit 105 processes two or more terms included in one related term group as the same term. The feature vector acquisition unit 105 can usually be realized by an MPU, a memory, or the like. The processing procedure of the feature vector acquisition unit 105 is usually realized by software, and the software is recorded in a recording medium such as a ROM. However, it may be realized by hardware (dedicated circuit).

  The determination unit 106 uses the one or more related patent feature vectors and the one or more unrelated patent feature vectors to generate an unrelated patent document corresponding to the unrelated patent feature vector that should belong to the related patent class. You may decide. When determining the unrelated patent document corresponding to the unrelated patent feature vector that should belong to the class of the related patent, the determination unit 106 determines the class of the related patent feature vector and the unrelated patent feature vector to be determined. It is possible to determine whether the unrelated patent feature vector to be determined belongs to the class of the related patent, the class of the related patent feature vector, the class of the unrelated patent feature vector, and the unrelated patent feature vector to be determined. It may be used to determine whether the unrelated patent feature vector to be determined belongs to the related patent class. The determination unit 106 normally uses two or more related patent feature vectors. The determination unit 106 normally uses two or more unrelated patent feature vectors. The same applies to other cases described below. In addition, the determination unit 106 selects one or more related patents from unrelated patent documents corresponding to a determination target unrelated patent feature vector that is a vector of non-related patent documents to be determined whether or not to belong to a related patent class. The determination may be made using one or more unrelated patent feature vectors excluding the feature vector and the determination target unrelated patent feature vector. It should be noted that the determination unit 106 outputs an unrelated patent document corresponding to a determination target unrelated patent feature vector, which is a vector of unrelated patent documents to be determined whether or not to belong to a related patent class, as one or more related patents. The determination may be made using two or more unrelated patent feature vectors including the feature vector and the determination target unrelated patent feature vector. The determination unit 106 belongs to the class of related patents using one or more related patent feature vectors, or one or more related patent feature vectors and one or more unrelated patent feature vectors. A new patent document corresponding to the new patent feature vector to be determined may be determined. The determination unit 106 uses one or more related patent feature vectors, or one or more related patent feature vectors and one or more unrelated patent feature vectors, to deal with related patent feature vectors that should not belong to the related patent class. Related patent documents may be determined.

  A class is a set of one or more feature vectors calculated. Classes do not contain different types of feature vectors. The phrase “not including different types of feature vectors” means, for example, that a class of related patent feature vectors does not include a feature vector that is not a related patent feature vector. The determination unit 106 may determine the complement of the related patent feature vector class as an unrelated patent set, and may determine the complement of the unrelated patent feature vector class as a related patent set. The complementary set of the union of the related patent feature vector class and the unrelated patent feature vector class may be determined as a patent document that is neither a related patent nor an unrelated patent. The class may be a set that is a result of classification by a learning device using machine learning described later.

  Here, the determination method described above is not limited using the feature vector. For example, the determination unit 106 may determine using vector similarity or may use machine learning. Hereinafter, the determination unit 106 uses (1) a method for determining a related patent document from one or more unrelated patent documents using a vector, (A) a method for determining using a vector similarity, and (B) machine learning It will be described in two parts: a method of making a determination using Also, (2) a method for determining a new patent document corresponding to a new patent feature vector belonging to a class of related patents, and (3) a method for determining an unrelated patent document from one or more related patent documents (1) ) And will be described in two parts. The determination unit 106 may classify new patent documents into three or more classes.

(1) Method for determining related patent document from one or more unrelated patent documents (A) Judgment using feature vector similarity Determining using feature vector similarity means unrelated patent feature to be judged Calculating a COS scale (cosine scale) of a vector, a representative vector of a class of related patent feature vectors, and a COS scale of a representative vector of a class of unrelated patent feature vectors; It may be determined that the class belongs to a class including a representative vector having a large COS measure. Further, a COS scale between the unrelated patent feature vector to be determined and one or more representative vectors of the class of the related patent feature vector is calculated, and the COS scale between the unrelated patent feature vector and one of the representative vectors is a threshold value. If larger, the determination unit 106 may determine that the unrelated patent is a related patent. Since the COS scale calculation method is a known technique, the description thereof is omitted. Note that the similarity calculation method may use a Pearson correlation coefficient or a deviation pattern similarity. Since the Pearson correlation coefficient and the deviation pattern similarity are well-known techniques, description thereof is omitted. Further, the class of the related patent feature vector may be 1, or may be 2 or more. When creating two or more classes, for example, feature vectors may be clustered using division optimization clustering (such as k-means method), or feature vectors may be clustered using hierarchical clustering (such as shortest distance method). Alternatively, the feature vectors may be clustered using other known clustering methods. Note that details of each clustering method are well-known techniques, and thus description thereof is omitted. The same applies to the class of unrelated patent feature vectors. The representative vector may be a feature vector that appears most frequently in the class, or may be an average vector of the class. The average vector may be a normal vector average, or a unit vector average calculated by treating all as unit vectors. If it is determined that the unrelated patent feature vector belongs to the class of the related patent feature vector, the patent document of the unrelated patent feature vector is determined as the related patent document.

(B) Judgment using machine learning Judgment using machine learning means learning using a learning device such as a neural network, SVM (Support Vector Machine) or SVR (Support Vector Regression), and learning is completed. On the other hand, the unrelated patent feature vector to be judged is classified to determine whether it belongs to the related patent feature vector class. A neural network is a learning model that aims to express some characteristics of brain function by computer simulation. There are various types of models and methods in the neural network, any of which may be adopted. For example, as the type of neural network, a perceptron may be employed, back-property may be employed, or a Boltzmann machine may be employed. Details of each model relating to the neural network are well known in the art, and a description thereof is omitted. SVM is a learning model that learns a classification pattern using teacher data, sets boundary lines for classification, and performs classification. Details of the SVM are well known in the art, and a description thereof will be omitted. SVR is a learning model that learns classification patterns using teacher data and classifies them into three or more classes. The details of SVR are well-known techniques, and thus the description thereof is omitted. When the determination unit 106 makes a determination using machine learning, the feature vectors of the patent documents stored in the related patent document storage unit 102 and the unrelated patent document storage unit 103 are learned as teacher data. The feature used by the determination unit 106 for learning is a value corresponding to each term acquired by the feature vector acquisition unit 105. The features used by the determination unit 106 for learning may include a patent classification. When the patent classification is included in the feature, numerical information that uniquely identifies the patent classification is set for the patent classification, and learning is performed using the value. After the learning is completed, for example, when the unrelated patent feature vector is determined to be the related patent feature vector, the determination unit 106 determines the patent document of the unrelated patent feature vector as the related patent document.

(2) Method for determining a new patent document corresponding to a new patent feature vector belonging to the class of related patents (A) Judging using the similarity of feature vectors Judging using the similarity of feature vectors A COS measure of a new patent feature vector, a representative vector of a class of related patent feature vectors, and a COS measure of a representative vector of a class of unrelated patent feature vectors are calculated, and the new patent feature vector to be judged is a COS measure. Is determined to belong to a class including a representative vector having a large.

(B) Judgment using machine learning Judgment using machine learning means learning using a learning device such as a neural network or SVM, and a new patent feature vector to be judged is determined for a learning device that has completed learning. The classification is to determine whether the class belongs to a related patent feature vector or an unrelated patent feature vector.

(3) Method for determining unrelated patent document from one or more related patent documents (A) Judgment using feature vector similarity Determining using feature vector similarity is related patent feature vector to be judged And a COS measure of the representative vector of the class of the related patent feature vector and a COS measure of the representative vector of the class of the non-related patent feature vector, and a representative vector having a large COS measure of the related patent feature vector to be determined Is determined to belong to the class included.

(B) Judgment using machine learning Judgment using machine learning means learning using a learning device such as a neural network or SVM, and the related patent feature vector to be judged is determined for the learning device that has completed learning. The classification is to determine whether it belongs to the class of unrelated patent feature vectors.

  The determination unit 106 can usually be realized by an MPU, a memory, or the like. The processing procedure of the determination unit 106 is usually realized by software, and the software is recorded on a recording medium such as a ROM. However, it may be realized by hardware (dedicated circuit).

  The output unit 107 outputs information specifying the related patent document determined by the determination unit 106. Further, the output unit 107 outputs information specifying a new patent document that should belong to the related patent class determined by the determination unit 106. Further, the output unit 107 outputs information for specifying related patent documents that should not belong to the related patent class determined by the determination unit 106.

  Here, output means display on a display, projection using a projector, printing with a printer, sound output, transmission to an external device, storage on a recording medium, other processing device or other program, etc. It is a concept including delivery of processing results. Further, the output unit 107 normally transmits the determination result to the user terminal 3 through the network 100. The output unit 107 can be realized by output device driver software, or output device driver software and an output device. As a result, it is possible to find a document of a patent that is originally a related patent but erroneously classified as a non-related patent, and it is possible to obtain only a related patent document from an unclassified patent document.

  FIG. 3 is a flowchart showing an example of the operation of the patent search support apparatus 1 in the present embodiment. Hereinafter, the operation will be described with reference to FIG.

  (Step S301) The reception unit 101 stores a folder storing one or more related patent documents from the user terminal 3, a folder storing one or more unrelated patent documents, and one or more new patent documents. Determine whether the folder is accepted. If each folder is received, the process proceeds to step S302. If not received, step S301 is repeatedly executed.

  (Step S302) The receiving unit 101 stores the received related patent document in the related patent document storage unit 102, stores the received non-related patent document in the non-related patent document storage unit 103, and newly receives the received new patent document. It is stored in the patent document storage unit 104.

  (Step S303) The feature vector acquisition unit 105 acquires a related patent feature vector from the related patent document stored in the related patent document storage unit 102. Details of the acquisition method of the feature vector acquisition unit 105 will be described later with reference to the flowchart of FIG.

  (Step S304) The feature vector acquisition unit 105 acquires an unrelated patent feature vector from the unrelated patent documents stored in the unrelated patent document storage unit 103. Details of the acquisition method of the feature vector acquisition unit 105 will be described later with reference to the flowchart of FIG.

  (Step S <b> 305) The feature vector acquisition unit 105 acquires a new patent feature vector from the new patent document stored in the new patent document storage unit 104. Details of the acquisition method of the feature vector acquisition unit 105 will be described later with reference to the flowchart of FIG.

  (Step S306) The determination unit 106 calculates an average vector from all the related patent feature vectors. If the dimensions of the vectors are different, the dimensions are unified to the related patent feature vector having the largest number of dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S307) The determination unit 106 calculates an average vector from all unrelated patent feature vectors. If the dimensions of the vectors are different, the dimensions are unified to the unrelated patent feature vector having the largest number of dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S308) The determination unit 106 substitutes 1 for the counter m.

  (Step S309) The determination unit 106 determines whether or not the m-th unrelated patent document is stored in the unrelated patent document storage unit 103. If the m-th unrelated patent document is stored in the unrelated patent document storage unit 103, the process proceeds to step S310. Otherwise, the process proceeds to step S312.

  (Step S310) The determination unit 106 determines whether the mth unrelated patent document is a related patent. Details of the determination by the determination unit 106 will be described later with reference to the flowchart of FIG.

  (Step S311) The determination unit 106 increments the counter m by 1, and returns to step S309.

  (Step S312) The determination unit 106 substitutes 1 for a counter n.

  (Step S313) The determination unit 106 determines whether or not the nth new patent document is stored in the new patent document storage unit 104. If the n-th new patent document is stored in the new patent document storage unit 104, the process proceeds to step S314, and if not, the process proceeds to step S316.

  (Step S314) The determination unit 106 determines whether the nth new patent document is a related patent. Details of the determination by the determination unit 106 will be described later with reference to the flowchart of FIG.

  (Step S315) The determination unit 106 increments the counter n by 1, and returns to step S313.

  (Step S316) The determination unit 106 substitutes 1 for the counter k.

  (Step S317) The determination unit 106 determines whether or not the k-th related patent document is stored in the related patent document storage unit 102. If the k-th new patent document is stored in the related patent document storage unit 102, the process proceeds to step S318, and if not, the process proceeds to step S320.

  (Step S318) The determination unit 106 determines whether the nth related patent document is an unrelated patent. Details of the determination by the determination unit 106 will be described later with reference to the flowchart of FIG.

  (Step S319) The determination unit 106 increments the counter k by 1, and returns to step S317.

  (Step S320) The output unit 107 outputs the result determined by the determination unit 106, and ends.

  In step S306 in the flowchart of FIG. 3, the determination unit 106 calculates an average vector from all the related patent feature vectors. However, the determination unit 106 configures two or more groups within which the difference in distance between all related patent feature vectors is within a certain range (COS scale is equal to or greater than a threshold), and calculates an average vector of the related patent feature vectors for each group. May be.

  In step S307, the determination unit 106 calculates an average vector from all unrelated patent feature vectors. However, the determination unit 106 configures two or more groups within which the difference in distance between all unrelated patent feature vectors is within a certain range (COS scale is equal to or greater than a threshold value), and calculates an average vector of unrelated patent feature vectors for each group. It may be calculated. Further, in the flowchart of FIG. 3, the feature vector of the unrelated patent document may not be calculated.

  FIG. 4 is a flowchart illustrating an example of the operation of the feature vector calculation (steps S302, S303, and S304) in FIG. The feature vector calculation process will be described below with reference to FIG. In FIG. 4, X is assigned a related patent document when called from step S302, and when called from step S303, an unrelated patent document is assigned, and when called from step S304. Suppose that a new patent document is assigned. The following description will be made assuming that the call is made from step S302 in which the related patent document is assigned to X. Step S303 and step S304 are to be read respectively.

  (Step S401) The feature vector acquisition unit 105 assigns 1 to the counter p.

  (Step S <b> 402) The feature vector acquisition unit 105 determines whether the p-th related patent document is stored in the related patent document storage unit 102. If the p-th related patent document is stored in the related patent document storage unit 102, the process proceeds to step S403. If the p-th related patent document is not stored, the process returns to the upper process.

  (Step S403) The feature vector acquisition unit 105 acquires all terms from the p-th related patent document.

  (Step S404) The feature vector acquisition unit 105 adds the term acquired in step S403 so as not to overlap with a term storage unit (not shown). Note that a term storage unit (not shown) continues to add terms until the processing is completed.

  (Step S405) The feature vector acquisition unit 105 calculates the TF-IDF value of the term acquired in step S403. Note that the feature vector acquisition unit 105 holds a related term dictionary (not shown), determines that the related terms stored in the related term dictionary are the same term, and calculates the TF-IDF value of the term. You may do it.

  (Step S406) The feature vector acquisition unit 105 creates a vector having TF-IDF values of terms as elements in the order in which the terms stored in a term storage unit (not shown) are stored. In addition, the element of the term which is not described in the p-th related patent document is set to 0.

  (Step S407) The feature vector acquisition unit 105 stores the feature vector created in step S406 in a feature vector storage unit (not shown) in association with the type of patent document. The type of patent document is information indicating whether it is a related patent, a non-related patent, or a new patent.

  (Step S408) The feature vector acquisition unit 105 increments the counter p by 1, and returns to step S402.

  FIG. 5 is a flowchart showing an example of the operation of the determination of FIG. 3 (steps S310, S314, and S317). Hereinafter, the determination process will be described with reference to FIG. Note that, when called from step S310, the m-th unrelated patent document is substituted for Y, and when called from step S314, the n-th new patent document is substituted and called from step S317. In this case, it is assumed that the kth related patent document is assigned. In Z, a related patent is substituted when called from step S310, a related patent is substituted when called from step S314, and an unrelated patent is substituted when called from step S317. It is assumed that The following description is based on the assumption that the non-related patent document is assigned to Y and the related patent is assigned to Z, and is called from step S310. Step S314 and step S317 are to be read respectively.

  (Step S501) The determination unit 106 calculates a COS measure between the average vector of related patents and the feature vector of the m-th unrelated patent document. When the dimensions of the vectors are different, the dimensions are unified into a vector with many dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S502) The determination unit 106 calculates a COS measure between the average vector of unrelated patents and the feature vector of the m-th unrelated patent document. When the dimensions of the vectors are different, the dimensions are unified into a vector with many dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S503) The determination unit 106 determines whether an average vector having a large COS measure is an average vector of related patents. If it is the average vector of the related patents, the process proceeds to step S504, and if it is not the average vector of the related patents, the process returns to the upper process.

  (Step S504) The determination unit 106 stores in the determination result storage unit (not shown) the mth unrelated patent document in association with the patent document ID that there is a possibility of a related patent.

  In the case where the determination unit 106 forms two or more groups within which the difference in distance between all the related patent feature vectors is within a certain range and calculates the average vector of the related patent feature vectors for each group, FIG. In the flowchart, if the COS measure of the patent feature vector of the unrelated patent and the average vector of the related patent feature vectors of any group is larger than the threshold, the determination unit 106 determines that the unrelated patent is the related patent. Also good. In such a case, in the flowchart of FIG. 5, when the patent feature vector of the unrelated patent and the average vector having the largest COS scale are the average vectors of the related patent feature vectors of any group, the determination unit 106 Such unrelated patents may be determined as related patents.

  Further, when the determining unit 106 forms two or more groups within which the distance difference between all the related patent feature vectors is within a certain range and calculates the average vector of the related patent feature vectors for each group, FIG. In the flowchart, when the COS measure between the patent feature vector of the new patent document and the average vector of the related patent feature vectors of any group is larger than the threshold, the determination unit 106 determines that the new patent document is a related patent. Also good. On the other hand, when the COS measure between the patent feature vector of the new patent document and the average vector of the related patent feature vectors of any group is smaller than the threshold, the determination unit 106 determines that the new patent document is an unrelated patent. good.

  Further, when the determining unit 106 forms two or more groups within which the distance difference between all the related patent feature vectors is within a certain range and calculates the average vector of the related patent feature vectors for each group, FIG. In the flowchart, if the COS measure of the patent feature vector of the related patent and the average vector of the related patent feature vectors of all the groups is smaller than the threshold, the determination unit 106 may determine the related patent as an unrelated patent. . Further, in the flowchart of FIG. 5, the patent feature vector of the unrelated patent need not be used.

  Hereinafter, a specific operation of the patent research support device 1 in the present embodiment will be described. As shown in FIG. 6, the user of the user terminal 3 operates a pointing device, a keyboard, or the like, and related patent folders storing related patent documents and unrelated patent folders storing non-related patent documents. And a new patent document folder in which new patent documents are stored, and then the “execute” button is clicked. Then, the accepting unit 101 accepts the network 100. When the receiving unit 101 receives the data of each folder, the related patent documents in the related patent folder are transferred to the related patent document storage unit 102, and the unrelated patent documents in the unrelated patent folder are set to the unrelated patent document storage unit 103. The new patent documents in the new patent folder are stored in the new patent document storage unit 104 (steps S301 and S302). In this specific example, it is assumed that the patent documents stored in each patent document storage unit are as shown in FIG. 7A shows related patent documents stored in the related patent document storage unit 102, FIG. 7B shows non-related patent documents stored in the non-related patent document storage unit 103, FIG. 7C shows a new patent document stored in the new patent document storage unit 104. Each table in FIG. 7 has a large number of records including a patent document ID, an invention name, a summary text, a claim text, a background art text, and the like. Here, the patent document ID is a public number here. Further, although patent documents are shown here as records in a table, it goes without saying that the data structure does not matter.

  The feature vector acquisition unit 105 is stored in the related patent document stored in the related patent document storage unit 102, the unrelated patent document stored in the unrelated patent document storage unit 103, and the new patent document storage unit 104. A feature vector is created from the new patent document (steps S303 to S305). Hereinafter, details of such processing will be described.

  The feature vector acquisition unit 105 acquires “patent document ID: JP2011-AAAAAA”, which is the first related patent document, from the related patent document storage unit 102 (steps S401 and S402). The feature vector acquisition unit 105 obtains a series of nouns “data management device”, “display unit”, “search keyword”, “data management system”, “data management system”, from the acquired “patent document ID: JP2011-AAAAAA”. “Electronic device”, “location information”, “current location”, “computer”, and the like are acquired as terms. The feature vector acquisition unit 105 adds the acquired term to a term storage unit (not shown) (steps S403 and S404). The feature vector acquisition unit 105 calculates a TF-IDF value for each acquired term as “data management device: 0.0264”, “display unit: 0.1628”, or the like (step S405). Then, the feature vector acquisition unit 105 creates a vector having the TF-IDF value as an element and stores it in a feature vector storage unit (not shown) (steps S406 and S407). Then, the feature vector acquisition unit 105 moves the process to the next patent document (step S408). Hereinafter, the feature vector acquisition unit 105 similarly creates feature vectors for all patent documents in the related patent document storage unit 102, the unrelated patent document storage unit 103, and the new patent document storage unit 104. The feature vector of each patent document acquired by the feature vector acquisition unit 105 is assumed to be as shown in FIG. The table of FIG. 8 has a large number of records having patent document IDs, patent document types, and vector elements corresponding to terms stored in a term storage unit (not shown).

  The determination unit 106 calculates an average vector of feature vectors of related patents and an average vector of feature vectors of unrelated patents (steps S306 and S307).

  The determination unit 106 acquires an unrelated patent feature vector of “patent document ID: JP2012-DDDDDD” which is the first unrelated patent document from a feature vector storage unit (not shown) (steps S308 and S309). The determination unit 106 calculates a COS measure between the acquired unrelated patent feature vector and the average vector of the related patent feature vectors, and a COS measure between the average vector of the unrelated patent feature vectors (steps S501 and S502). It is assumed that the non-related patent feature vector of “Patent Document ID: JP2012-DDDDDD” has a larger COS scale with the average vector of the related patent feature vectors. The determination unit 106 stores “patent document ID: JP2012-DDDDDD” in association with information indicating that there is a possibility of a related patent in a determination result storage unit (not shown) (steps S310, S503, and S504). Note that the determination unit 106 stores nothing if the COS scale with the average vector of unrelated patent feature vectors is larger. Then, the determination unit 106 moves the process to the next unrelated patent feature vector (step S311). Thereafter, the determination unit 106 similarly determines all remaining unrelated patent documents stored in a feature vector storage unit (not shown).

  The determination unit 106 acquires a new patent feature vector of “patent document ID: JP2012-GGGGGG”, which is the first new patent document, from a feature vector storage unit (not shown) (steps S312 and S313). The determination unit 106 calculates a COS measure between the acquired new patent feature vector and the average vector of the related patent feature vectors, and a COS measure between the average vector of the unrelated patent feature vectors. It is assumed that the new patent feature vector of “Patent Document ID: JP2012-GGGGGG” has a larger COS scale with the average vector of unrelated patent feature vectors. If the COS scale with the average vector of unrelated patent feature vectors is larger, the determination unit 106 moves the processing to the next new patent feature vector without storing anything (step S315). In addition, when the COS scale with the average vector of the related patent feature vectors is larger, the determination unit 106 is information indicating that there is a possibility of a related patent in the determination result storage unit not illustrated with “patent document ID”. And stored in association with each other (step S314). Thereafter, the determination unit 106 similarly determines all remaining new patent documents stored in a feature vector storage unit (not shown).

  The determination unit 106 acquires a related patent feature vector of “patent document ID: JP2011-AAAAAA”, which is the first related patent document, from a feature vector storage unit (not shown) (steps S316 and S317). The determination unit 106 calculates a COS measure between the acquired related patent feature vector and the average vector of the related patent feature vectors, and a COS measure between the average vector of the unrelated patent feature vectors. It is assumed that the unrelated patent feature vector of “Patent Document ID: JP2011-AAAAAA” has a larger COS scale with the average vector of unrelated patent feature vectors. The determination unit 106 stores “patent document ID: JP2011-AAAAAA” in association with information indicating that there is a possibility of an unrelated patent in a determination result storage unit (not shown) (step S318). Note that the determination unit 106 stores nothing when the COS scale with the average vector of the related patent feature vectors is larger. Then, the determination unit 106 moves the process to the next related patent feature vector (step S319). Thereafter, the determination unit 106 similarly determines all remaining unrelated patent documents stored in a feature vector storage unit (not shown).

  When the output unit 107 detects that the “patent document ID” is stored in the determination result storage unit (not shown), the information indicating whether there is a possibility of the “patent document ID” and the related patent document or the unrelated patent document. Are transmitted to the user terminal 3 via the network 100 (step S320). On the user terminal 3, “JP 2012-DDDDDD”, “JP 2005-FFFFFF”, and the like are displayed as patent documents that may be related patents in the unrelated patent document storage unit 103. In addition, “JP 2011-AAAAAA” or the like is displayed as a patent document that may be an unrelated patent in the related patent document storage unit 102. In addition, “JP 2011-HHHHHH” or the like is displayed as a patent document that may be a related patent in the new patent document storage unit 104. On the user terminal 3, it is displayed as shown in FIG. By viewing the display in FIG. 9, for example, the user confirms again a patent document that has been erroneously classified as an unrelated patent.

  As described above, according to the patent search support apparatus 1 of the present embodiment, it is possible to check related patent documents that are mistakenly classified as unrelated patent documents when a patent search is performed, thereby avoiding a search omission due to a human error. can do. In addition, it is possible to check unrelated patent documents that have been mistakenly classified as related patent documents during a patent search, thereby reducing work time. Further, if related patents and non-related patents are classified to some extent, it is possible to determine whether an unclassified new patent document is a related patent document, so that the time required for subsequent patent searches can be shortened.

  In the present embodiment, the case where the patent search support apparatus 1 includes the new patent document storage unit 104 has been described. However, the patent search support apparatus 1 may not include the new patent document storage unit 104. In the case where the new patent document storage unit 104 is not provided, the patent search support device 1 may not perform the process related to the new patent document.

  In the present embodiment, the determination unit 106 of the patent search support apparatus 1 determines whether the patent document stored in the related patent document storage unit 102 is a patent document that may be an unrelated patent. Although the case has been described, the determination unit 106 may not determine from the patent documents stored in the related patent document storage unit 102 whether the document is a patent document that may be an unrelated patent. If it is not determined from the patent documents stored in the related patent document storage unit 102 whether the patent document is a potentially unrelated patent, the related documents are stored in the unrelated patent document storage unit 103. It may be determined whether the document is a patent document that may be a patent, a patent document that may be a related patent and a new patent document from the patent documents stored in the unrelated patent document storage unit 103. It may be determined whether or not the patent document is a related patent document stored in the storage unit 104.

  Further, in the present embodiment, the determination unit 106 of the patent search support apparatus 1 determines that the determination target feature vector belongs to the class of the unrelated patent feature vector, but the determination unit 106 This is a patent document that may be a related patent from the patent documents stored in the unrelated patent document storage unit 103 using only the related patent feature vector class without using the class of the unrelated patent feature vector. It may be determined whether it is a patent document that has a possibility of a related patent from the patent documents stored in the new patent document storage unit 104, and is stored in the related patent document storage unit 102. It may be determined whether the document is a patent document that may be an unrelated patent. When the non-related patent feature vector class is not used, a feature vector that does not belong to the related patent feature vector class may be used as an unrelated patent. If a class of unrelated patent feature vectors is not used, a method using machine learning cannot be used. Therefore, a related patent is used by using the similarity of a related patent feature vector with one or more representative vectors of classes. It is determined whether or not it belongs to a feature vector class. The determination as to whether or not belonging to a class may be made based on, for example, whether or not the similarity between the representative vector and the feature vector to be determined is equal to or greater than a predetermined threshold. The determination may be made based on whether or not the feature vector class is clustered.

  Moreover, the software which implement | achieves the patent search assistance apparatus 1 in this Embodiment is the following programs. That is, the program includes a related patent document storage unit that can store one or more related patent documents that are determined to be related as a result of screening work in the patent search, and a related result of screening work in the patent search. A computer having access to an unrelated patent document storage unit capable of storing one or more unrelated patent documents, which are documents of unrelated patents determined not to be used, and one or more terms from each of the one or more related patent documents; Obtain a related patent feature vector, which is a vector of each related patent document using one or more terms, for each related patent document, and acquire one or more terms from each of one or more unrelated patent documents A feature vector acquisition unit that acquires, for each unrelated patent document, an unrelated patent feature vector that is a vector of each unrelated patent document using one or more terms, and one or more related patent feature vectors And one or more unrelated patent feature vectors, a determination unit for determining related patent documents from one or more unrelated patent documents, and an output unit for outputting information specifying the related patent documents determined by the determination unit It is a program to make it function as.

(Embodiment 2)
In the present embodiment, using a result obtained by manually classifying a patent document into a related patent and an unrelated patent, a patent that may be a related patent document or an unrelated patent document from an unclassified patent document, etc. The patent search support apparatus 2 that determines the document and outputs the result will be described.

  FIG. 1 is a conceptual diagram of a system including a patent search support apparatus 1 according to the present embodiment. By replacing the patent search support apparatus 1 with a patent search support apparatus 2, the concept of a system including the patent search support apparatus 2 is illustrated. The figure.

  FIG. 2 is a block diagram of the patent search support apparatus 1 according to the present embodiment. By replacing the determination unit 106 with the determination unit 201 and the output unit 107 with the output unit 202, FIG. To do. The patent research support apparatus 2 includes a reception unit 101, a related patent document storage unit 102, an unrelated patent document storage unit 103, a new patent document storage unit 104, a feature vector acquisition unit 105, a determination unit 201, and an output unit 202. Since the receiving unit 101, the related patent document storage unit 102, the unrelated patent document storage unit 103, the new patent document storage unit 104, and the feature vector acquisition unit 105 are the same as those in the first embodiment, the description thereof is omitted.

  The determination unit 201 uses one or more related patent feature vectors to determine whether or not the patent feature vector of the new patent document belongs to the related patent class. In addition, the determination unit 201 uses one or more related patent feature vectors and one or more unrelated patent feature vectors to determine whether the patent feature vector of the new patent document is a related patent class or an unrelated patent. Decide whether it is a class. Since the method using the feature vector and the concept of class are the same as those of the determination unit 106 of the first embodiment, description thereof is omitted. The determination unit 201 determines whether the new patent document belongs to a class of related patent feature vectors or a class of unrelated patent feature vectors. Further, the determination unit 201 may determine whether or not it belongs to the class of the related patent feature vector, and may determine whether or not it belongs to the class of the unrelated patent feature vector. The determination unit 201 may determine that a new patent document that does not belong to one class belongs to the other class. The determination method of the determination unit 201 is substantially the same as that of the determination unit 106, but the determination unit 106 handles the determination result differently. For example, when the determination unit 106 determines whether or not the new patent document is a related patent, if it is determined that the new patent document is not a related patent, the determination result is not used. Determines that the patent indicated by the new patent document is an unrelated patent. The determination unit 201 can usually be realized by an MPU, a memory, or the like. The processing procedure of the determination unit 201 is usually realized by software, and the software is recorded on a recording medium such as a ROM. However, it may be realized by hardware (dedicated circuit).

  The output unit 202 outputs the class determined by the determination unit 201 for each new patent document. The output unit 107 normally transmits the determination result to the user terminal 3 through the network 100. The output unit 107 can be realized by output device driver software, or output device driver software and an output device. Thereby, related patent documents or unrelated patent documents can be classified and acquired from unclassified patent documents.

  FIG. 10 is a flowchart showing an example of the operation of the patent search support apparatus 1 in the present embodiment. Hereinafter, the operation will be described with reference to FIG. Steps S301 to S307 are the same as those in the first embodiment, and a description thereof will be omitted.

  (Step S601) The determination unit 201 assigns 1 to a counter n.

  (Step S <b> 602) The determination unit 201 determines whether the nth new patent document is stored in the new patent document storage unit 104. If the nth new patent document is stored in the new patent document storage unit 104, the process proceeds to step S603, and if not, the process proceeds to step S607.

  (Step S603) The determination unit 201 calculates a COS measure between the average vector of related patents and the feature vector of the nth new patent document. When the dimensions of the vectors are different, the dimensions are unified into a vector with many dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S604) The determination unit 201 calculates a COS measure between the average vector of unrelated patents and the feature vector of the nth new patent document. When the dimensions of the vectors are different, the dimensions are unified into a vector with many dimensions. 0 is assigned to the dimension element that has been increased during unification.

  (Step S605) The determination unit 201 stores the type of the average vector patent document having the larger COS scale in association with the nth new patent document ID in a determination result storage unit (not shown).

  (Step S606) The determination unit 201 increments the counter n by 1, and returns to step S313.

  (Step S607) The output unit 202 outputs the result determined by the determination unit 201, and ends.

  Hereinafter, a specific operation of the patent research support device 2 in the present embodiment will be described. Since operations other than the determination unit 201 and the output unit 202 are the same as those in the first embodiment, description thereof will be omitted, and the related patent document storage unit 102, the unrelated patent document storage unit 103 in FIG. Data is stored in the patent document storage unit 104, and the feature vector of each patent document in FIG. 8 is acquired.

  The determination unit 201 acquires a new patent feature vector of “patent document ID: JP 2012-GGGGGG”, which is the first new patent document, from a feature vector storage unit (not shown) (steps S601 and S602). The determination unit 106 calculates a COS measure between the acquired new patent feature vector and the average vector of the related patent feature vectors, and a COS measure between the average vector of the unrelated patent feature vectors (steps S603 and S604). It is assumed that the new patent feature vector of “Patent Document ID: JP2012-GGGGGG” has a larger COS scale with the average vector of unrelated patent feature vectors. The determination unit 106 stores “patent document ID: JP 2012-GGGGGG” in association with information indicating that there is a possibility of an unrelated patent in a determination result storage unit (not shown). (Step S605). If the COS scale with the average vector of the related patent feature vectors is larger, the determination unit 106 may have a related patent in the determination result storage unit (not shown) “Patent Document ID: JP2012-GGGGGG”. And stored in association with information indicating that there is. Then, after the determination unit 201 moves the process to the next new patent feature vector (step S606), the determination unit 201 also applies to all the remaining new patent documents stored in the feature vector storage unit (not shown). Judgment.

  When the output unit 202 detects that the “patent document ID” is stored in a determination result storage unit (not shown), the output unit 202 outputs “patent document ID” and information indicating whether the document is a related patent document or an unrelated patent document. (Step S607). In a specific example, “JP2011-HHHHHH” or the like is a patent document that may be a related patent in the new patent document storage unit 104, and JP2012-GGGGGG or “JP2005-IIIIII” is new. Output as unrelated patent documents in the patent document storage unit 104. The output unit 202 transmits to the user terminal 3 via the network 100. When the user terminal 3 receives the determination result transmitted by the output unit 202, the user terminal 3 performs the display shown in FIG. The user confirms the patent document by looking at the display of FIG.

  As described above, according to the patent research support device 2 of the present embodiment, if a related patent and an unrelated patent are classified to some extent, an unclassified new patent document is a related patent document or an unrelated patent document. Therefore, patent search can be conducted efficiently.

  Moreover, the software which implement | achieves the patent search assistance apparatus 1 in this Embodiment is the following programs. That is, the program includes a related patent document storage unit that can store one or more related patent documents that are determined to be related as a result of screening work in the patent search, and a related result of screening work in the patent search. Non-related patent document storage unit that can store one or more non-related patent documents that are determined to be non-related patents, and related patents excluding one or more related patent documents and one or more non-related patent documents A computer having access to a new patent document storage that can store one or more new patent documents that are documents of the same, obtain one or more terms from each of the one or more related patent documents, and use the one or more terms A related patent feature vector, which is a vector of each related patent document, is acquired for each related patent document, and one or more terms are acquired from each of one or more unrelated patent documents, and the one or more terms are used. An unrelated patent feature vector, which is a vector of unrelated patent documents, is acquired for each unrelated patent document, and one or more terms are acquired from each of one or more new patent documents, and A new patent feature vector that is a vector of a new patent document is obtained for each new patent document by using a feature vector acquisition unit, one or more related patent feature vectors, and one or more unrelated patent feature vectors. This is a program for functioning as a decision unit that determines whether the class to which a patent feature vector belongs is a related patent class or an unrelated patent class, and an output unit that outputs the class determined by the decision unit for each new patent document. .

  In each of the above embodiments, the case where the patent search support apparatuses 1 and 2 include the reception unit 101 has been described. However, the patent search support apparatuses 1 and 2 may not include the reception unit 101. If the receiving unit 101 is not provided, the patent search support apparatuses 1 and 2 in which patent documents are stored in advance in the related patent document storage unit 102, the unrelated patent document storage unit 103, and the new patent document storage unit 104 It may be.

  In each of the above embodiments, the case where the receiving unit 101 of the patent search support apparatuses 1 and 2 receives a new patent document has been described. However, the receiving unit 101 may not receive a new patent document. When the receiving unit 101 does not receive a new patent document, a related patent document and an unrelated patent document received by the receiving unit 101 from a patent document storage unit that stores three or more patent documents outside or not illustrated. Patent documents other than the above may be acquired and stored in the new patent document storage unit as new patent documents.

  Further, although cases have been described with the above embodiments where they are server / client systems, each of the above devices may be a server device in a server / client system or may be a stand-alone device. In the case of a stand-alone device, the receiving unit 101 may receive information input from an input device such as a keyboard, a mouse, or a touch panel. The output unit 107 and the output unit 202 may perform output for display on a display, projection using a projector, printing with a printer, and the like.

  In each of the above embodiments, each component may be configured by dedicated hardware, or a component that can be realized by software may be realized by executing a program. For example, each component can be realized by a program execution unit such as a CPU reading and executing a software program recorded on a recording medium such as a hard disk or a semiconductor memory.

  In each of the above programs, the function realized by the program does not include a function that can be realized only by hardware. For example, functions that can be realized only by hardware such as a modem and an interface card in an acquisition unit that acquires information, an output unit that outputs information, and the like are not included in the functions realized by the program.

  FIG. 12 is a schematic diagram showing an example of the external appearance of a computer that executes the above-described programs to realize the present invention according to the above-described embodiment. The embodiment described above can be realized by computer hardware and a computer program executed on the computer hardware.

  In FIG. 12, a computer system 1100 includes a computer 1101 including a CD-ROM drive 1105 and an FD drive 1106, a keyboard 1102, a mouse 1103, and a monitor 1104.

  FIG. 13 is a diagram showing an internal configuration of the computer system 1100. In FIG. 13, in addition to the CD-ROM drive 1105 and the FD drive 1106, a computer 1101 is connected to an MPU 1111, a ROM 1112 for storing a program such as a bootup program, and an MPU 1111. And a RAM 1113 that provides a temporary storage space, a hard disk 1114 that stores application programs, system programs, and data, an MPU 1111, a bus 1115 that interconnects the ROM 1112, and the like. The computer 1101 may include a network card (not shown) that provides connection to the LAN.

  A program for causing the computer system 1100 to execute the functions of the present invention according to the above embodiment is stored in the CD-ROM 1121 or the FD 1122, inserted into the CD-ROM drive 1105 or the FD drive 1106, and transferred to the hard disk 1114. May be. Instead, the program may be transmitted to the computer 1101 via a network (not shown) and stored in the hard disk 1114. The program is loaded into the RAM 1113 when executed. The program may be loaded directly from the CD-ROM 1121, the FD 1122, or the network.

  The program does not necessarily include an operating system (OS) or a third-party program that causes the computer 1101 to execute the functions of the present invention according to the above-described embodiment. The program may include only a part of an instruction that calls an appropriate function (module) in a controlled manner and obtains a desired result. How the computer system 1100 operates is well known and will not be described in detail.

  The present invention is not limited to the above-described embodiments, and various modifications are possible, and it goes without saying that these are also included in the scope of the present invention.

  For example, patent search support can reduce the time to classify related patents and non-related patents, and if there is an error in the documents of patents that have been classified as non-related patents, there is an opportunity to reconsider them. It is useful as a device.

DESCRIPTION OF SYMBOLS 1, 2, Patent search support apparatus 101 Reception part 102 Related patent document storage part 103 Unrelated patent document storage part 104 New patent document storage part 105 Feature vector acquisition part 106, 201 Judgment part 107, 202 Output part

Claims (6)

A related patent document storage unit capable of storing one or more related patent documents, which are related patent documents determined to be related as a result of the screening work in the patent search;
A non-related patent document storage unit capable of storing one or more non-related patent documents, which are non-related patent documents determined to be irrelevant as a result of screening in the patent search;
Obtaining one or more terms from each of the one or more related patent documents, obtaining a related patent feature vector, which is a vector of each related patent document using the one or more terms, for each related patent document; and Features that acquire one or more terms from each of one or more unrelated patent documents, and acquire, for each unrelated patent document, an unrelated patent feature vector that is a vector of each unrelated patent document using the one or more terms A vector acquisition unit;
Using the one or more related patent feature vectors and the one or more unrelated patent feature vectors to determine a non-related patent document corresponding to an unrelated patent feature vector that should belong to a class of related patents;
A patent search support apparatus comprising: an output unit that outputs information for specifying an unrelated patent document determined by the determination unit. The determination unit
An unrelated patent document corresponding to a judgment object unrelated patent feature vector, which is a vector of unrelated patent documents to be judged as to whether or not to belong to a related patent class, includes the one or more related patent feature vectors and the judgment target The patent search support device according to claim 1, wherein the determination is made using one or more unrelated patent feature vectors excluding unrelated patent feature vectors. A new patent document storage unit capable of storing one or more new patent documents that are patent documents excluding the one or more related patent documents and the one or more unrelated patent documents;
The feature vector acquisition unit
Obtaining one or more terms from each of the one or more new patent documents, obtaining a new patent feature vector, which is a vector of each new patent document, using the one or more terms for each new patent document;
The determination unit
A new patent document corresponding to a new patent feature vector that should belong to a class of related patents using the one or more related patent feature vectors or the one or more related patent feature vectors and the one or more unrelated patent feature vectors Decide
The output unit is
The patent search support device according to claim 1 or 2, wherein information specifying a new patent document that should belong to a class of related patents determined by the determination unit is output. The determination unit
Associations corresponding to related patent feature vectors that should not belong to a class of related patents using the one or more related patent feature vectors or the one or more related patent feature vectors and the one or more unrelated patent feature vectors Determine patent documents,
The output unit is
The patent search support device according to any one of claims 1 to 3, which outputs information specifying a related patent document that should not belong to a related patent class determined by the determination unit. As a result of screening work in patent search, a related patent document storage unit that can store one or more related patent documents that are determined to be related, and as a result of screening work in patent search, it was determined to be irrelevant Patent search support method processed using an unrelated patent document storage unit capable of storing one or more unrelated patent documents that are unrelated patent documents, a feature vector acquisition unit, a determination unit, and an output unit There,
The feature vector acquisition unit acquires one or more terms from the one or more related patent documents, and uses the one or more terms to determine a related patent feature vector that is a vector of each related patent document for each related patent document. And acquiring one or more terms from each of the one or more unrelated patent documents, and using the one or more terms, a non-related patent feature vector, which is a vector of each unrelated patent document, A feature vector acquisition step for each patent document;
The determination unit determines an unrelated patent document corresponding to an unrelated patent feature vector to belong to a related patent class using the one or more related patent feature vectors and the one or more unrelated patent feature vectors. A decision step ;
A patent search support method, comprising: an output step in which the output unit outputs information specifying the related patent document determined in the determination step. As a result of screening work in patent search, a related patent document storage unit that can store one or more related patent documents that are determined to be related, and as a result of screening work in patent search, it was determined to be irrelevant A computer accessible to an unrelated patent document storage section capable of storing one or more unrelated patent documents that are unrelated patent documents;
Obtaining one or more terms from each of the one or more related patent documents, obtaining a related patent feature vector, which is a vector of each related patent document using the one or more terms, for each related patent document; and Features that acquire one or more terms from each of one or more unrelated patent documents, and acquire, for each unrelated patent document, an unrelated patent feature vector that is a vector of each unrelated patent document using the one or more terms Vector acquisition part,
A determination unit that determines an unrelated patent document corresponding to an unrelated patent feature vector that should belong to a class of related patents using the one or more related patent feature vectors and the one or more unrelated patent feature vectors;
The program for functioning as an output part which outputs the information which specifies the unrelated patent document determined by the said judgment part.

Images