CN103116811A - Patent value evaluation method based on value delivery - Google Patents

Abstract

The invention discloses a patent value evaluation method based on value delivery. The patent value evaluation method includes a patent value static evaluation process and a dynamic renewal process. In the patent value evaluation process, firstly, a patent quotation network is built according to similarities between patents, and patent value in each node is initialized; secondly, contribution degree is calculated by value from a 'quoted' patent to a target patent according to node value of the 'quoted' patent, and patent quantity quoted by the 'quoted' node and patent published time, and then target patent value is acquired by accumulation; finally, each patent value is sequentially calculated. When a new patent is available, the similarities between the patent and an existing patent are firstly calculated, and then a 'quoted' relationship is built up, and finally values of patents are sequentially updated. The patent value evaluation method can calculate value of each patent from a patent database based on a delivery idea of PageRank algorithm, and has high effectiveness.

Description

A kind of patent valve estimating method based on value delivery

Technical field

The present invention relates to a kind of patent valve estimating method, especially relate to a kind of patent valve estimating method based on value delivery.

Background technology

Development along with economic globalization process, knowledge and Useful Information present explosive growth speed, knowledge-based industry at home in economy shared proportion improve constantly, intellecture property has become the focus of competing between country and national, enterprise and enterprise.Intellecture property has caused that various countries pay attention to widely at present, has just proposed the national tactful of foundation " intellecture property state " as far back as Japan in 2002.

Patent is maximum in the world technical information source, is the carrier of scientific and technical information and knowledge, is the innovation and creation that are subjected to the legal norm protection.It refers to that innovation and creation propose patented claim to national approving authority, the patent rights of these innovation and creation being enjoyed in setting time through examining in accordance with the law that qualified backward patent applicant authorizes.The patentee enjoys its patent right that has and monopolizing or right to exclude all others, and without its license or the special circumstances of method patent right rule regulation occur, anyone must not use, otherwise namely consists of infringement.Non-patentee wants to use other people patented technology, and the agreement that must in accordance with the law obtain the patentee perhaps can.Obtain other people patent right and generally buy to realize by patent, therefore non-patentee's patent buy and patentee's patent transfer process in, it is especially important that the assessment of patent value seems.

In the U.S. CHI of the intellecture property consulting firm twentieth century early stage indicator evaluation system that the proposes seventies, quote index as proof, be influence power and the value that number of times that patent is cited is used to the assessment objective patent, the citation of patent is higher, represent that this technology belongs to basic or leading technology, be in core technology or be positioned at the technology point of crossing.But this method has only considered that quantity that the target patent is cited is on the impact of patent value, do not consider the value of referenced patents itself and novelty of patent to the impact of patent value, the disappearance of Chinese patent citation relation has brought difficulty also for this method.Although have document to propose the value of the factor analysis patents such as owner of technical field, the patent of country origin, patent from the inventor, do not provide the scheme of perfect analysis patent value.Especially, Chinese technology business in 2011 has proposed " patent value analysis indexes system ", weigh the value of patent by legal value degree, technological value degree and economic worth degree three aspects:, but index wherein relies on and artificial assessment fully, need to expend a large amount of human resources and time resource.

Summary of the invention

The present invention solves the existing technical matters of prior art; Provide a kind of and considered that potential adduction relationship is on the impact of patent value; Solved a kind of patent valve estimating method based on value delivery of the impact that in Chinese patent, part citation information disappearance causes.

It is to solve the existing technical matters of prior art that the present invention also has a purpose; A kind of impact of patent novelty on patent value of having considered is provided; A kind of patent valve estimating method based on value delivery with good value assessment effect.

Above-mentioned technical matters of the present invention is mainly solved by following technical proposals:

A kind of patent valve estimating method based on value delivery is characterized in that, based on definition: the set D={d of patent file ₁, d ₂..., d _n, the time set T={T that patent is delivered ₁, T ₂..., T _n; Wherein, d _iThe document content of expression patent i; t _iThe application time of expression patent i; Comprise the patent value that is undertaken by the static evaluation module of patent value static evaluation step and by dynamically updating the step that dynamically updates that module carries out, specifically comprise:

The step of the static evaluation of the patent value that is undertaken by the static evaluation module of patent value:

Step 1.1 is calculated in the patent set similarity between patent file in twos;

Step 1.2 is according to the similarity between patent and the time structure similarity network of delivering;

Step 1.3, the value of all patents of initialization is calculated the value of each patent successively according to the thought of value delivery;

The step that dynamically updates that dynamically updates the patent value that module carries out by patent value:

Step 2.1 when new patent arrives, is upgraded the current value of each patent file according to the value decay factor;

Step 2.2 adds new patent file in patent similarity network;

Step 2.3 is upgraded the value of each patent successively according to new patent similarity network.

In a kind of above-mentioned patent valve estimating method based on value delivery, described step 1.1 comprises following substep:

Step 1.11, the space vector that calculates patent file represent, concrete operation method is as follows:

For patent file set D={d ₁, d ₂..., d _nIn any patent d _i, utilize the space vector of one group of keyword to represent; Its process is at first all patent files to be carried out Chinese word segmentation, then according to the stop words in self-defined or public stop words dictionary removal document; For removing lexical item after stop words, calculate the weight of each lexical item in document, its computing formula is:

$w(t_j,d_i)=\frac{\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j}+0.01)}{\sqrt{{\mathrm{\Σ}}_{t_j\∈d_i}{[\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j})+0.01]}^2}}$ Formula one;

Wherein, w (t _j, d _i) be lexical item t _jAt text d _iIn weight, and tf (t _j, d _i) be word t _jAt text d _iIn word frequency, N is the sum of patent set D Patent,

Lexical item t appears in D for the patent set _jThe patent file number, denominator is normalized factor;

At last, represent each piece patent file with space vector corresponding to each lexical item, be expressed as

W wherein _ijBe lexical item t _jAt patent file d _iIn weight;

Step 1.12 represents according to the space vector of document, and calculate the similarity between patent file in twos, concrete operation method is:

To any two patent file d _iAnd d _j, coming the similarity of measurement with the included angle cosine between the vector of its correspondence, its formula is:

$\mathrm{sim}(d_i,d_j)=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k\left(d_i\right)\×w_k\left(d_j\right)}{\sqrt{\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_i\right)\right)\×\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_j\right)\right)}}$ Formula two;

W wherein _k(d _i) be that k lexical item is at document d _iIn weight, w _k(d _j) be that k lexical item is at document d _jIn weight.

A kind of patent valve estimating method based on value delivery above-mentioned is characterized in that, described step 1.2 comprises following substep:

Step 1.21, in the similarity network, concrete operation method is as follows with the patent compound mapping:

Set up digraph G=(V, E), wherein V is corresponding to the set D of patent file, and E is corresponding to the contact between patent file, and the contact here is that similarity and the patent between patent file delivered temporal association;

Step 3.2, the limit of structure patent network, concrete operation method is as follows:

Set similarity threshold sim_thrd, for any two patent file d _iAnd d _jIf, both similarity sim (d _i, d _j)=sim_thrd, set up a limit between patent i and j; If the time of delivering of patent i is more Zao than patent j, set up so patent j to the directed edge of patent i; If otherwise set up patent i to the directed edge of j; Its formula is:

e(d _i,d _j)=1 if sim(d _i,d _j)>=sim_thrd and T _i>T _j;

e(d _i, d _jThe 1 if sim (d of)=- _i, d _j)=sim_thrd and T _i＜T _j; Formula three;

e(d _i,d _j)＝0 else

T wherein _iAnd T _jRepresent respectively the time that patent i and patent j deliver, e (d _i, d _j)=1 expression is set up node i to the directed edge of node j, and-1 expression is set up node j to the directed edge of node i, does not have the limit between 0 expression node i and node j.

In a kind of above-mentioned patent valve estimating method based on value delivery, described step 1.3 comprises following substep:

Step 4.1, the value of all patents of initialization, concrete operation method is as follows:

The value of patent can decay over time, and the setting-up time decay factor is e ^-β, 0＜β＜1 wherein, its value is specified by the user.Deliver time and the decay factor value of each patent of initialization successively according to patent.For any patent i, its initialized patent value R ₀(i) computing formula is:

R ₀(i)=1 * e ^{-β t}Formula four;

Wherein t is that patent i delivers the time till now.

The computing formula of t is:

t=T _i-T ₀Formula five;

T wherein _iBe the time that patent i delivers, T ₀Be the current time.

Step 4.2 is calculated the value of each patent successively, and concrete operation method is as follows:

At first all patents in the patent set were sorted from back to front according to its time of delivering.Then, calculate successively the value of each patent.For any patent i, its value is determined by the size of all patents of quoting it to its Value Contribution degree.For the patent j of referenced patents i, the big or small computing formula of patent value that it passes to patent i is as follows:

$\mathrm{Contribution}(j,i)\frac{\mathrm{Value}\left(j\right)}{\mathrm{De}\left(j\right)}\×\mathrm{sim}(d_i,d_j)\×e^{-\mathrm{\β}(T_j-T_i)}$ Formula six;

Wherein Value (j) is the current value of patent j, the number of the patent that De (j) quotes for patent j, the i.e. out-degree of node j in the similarity network G.

The patent value of patent i is by the Value Contribution degree sum decision of all patents of quoting i to it, and computing formula is:

Formula seven;

Wherein δ is ratio of damping, and 0＜δ＜1 is arranged by system or user.

In a kind of above-mentioned patent valve estimating method based on value delivery, described step 2.1 comprises following substep:

Step 2.11 is upgraded the current time

Concrete operation method is as follows:

The time that defines new patent arrival is (t+1), and the previous time is t, upgrades the current time of the time of (t+1) Its formula is:

$T_0^{t+1}=T_0^t+1$ Formula eight;

Wherein

Be the current time of (t+1) time,

Be the current time of t time;

Step 2.12 calculates according to being worth decay factor the value that has each patent under patent condition, and concrete method of operating is as follows: before new patent added network G, the value of patent decayed in time, and its formula is:

Value ^t+1(i)=e ^-β* Value ^t(i) formula nine;

E wherein ^-βBe decay factor, β is specified and 0＜β＜1 by the user; Value ^t+1(i) be the initial patent value of (t+1) time patent i, Value ^t(i) be the patent value of t time patent i.

In a kind of above-mentioned patent valve estimating method based on value delivery, described step 2.2 comprises following substep:

Step 2.21, calculate new patent file and the similarity of existing patent file, concrete method of operating is: according to the step in step 1.1, with new document carry out participle, stop words removes processing, and calculate the weight of lexical item in each new patent file according to formula one, utilize space vector to represent the patent file that each is new; Application of formula two afterwards, calculate the similarity of new patent file between existing document;

Step 2.22, new patent file is added patent similarity network, concrete method of operating is as follows: application of formula three judges the relation of new patent file between existing patent successively, sets up the related of limit between new patent and existing patent in the similarity network G.

In a kind of above-mentioned patent valve estimating method based on value delivery, the concrete grammar of described step 2.2 namely upgrades the value of each patent successively, concrete method of operating is: at first existing patent was sorted from back to front by the time, then for there is the patent file in path with new patent in network, upgrade iteratively successively in order the patent value of each patent, its computing formula is:

Formula ten;

Wherein, De ^t+1(j) be the out-degree of (t+1) time patent j in the G of networking.

Therefore, the present invention has following advantage: considered that potential adduction relationship is on the impact of patent value 1.; 2. solved the impact that in Chinese patent, part citation information disappearance causes; 3. considered the impact of patent novelty on patent value; 4. has good value assessment effect.

Description of drawings

Fig. 1 builds the schematic diagram of similarity network in the present invention.

Fig. 2 is the schematic diagram of the similarity network of foundation.

Fig. 3 is the schematic diagram that patent new in the present invention adds network.

Fig. 4 is the flow process of Patent Valuation Method of the present invention.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment:

A kind of patent valve estimating method based on value delivery comprises the static evaluation step of patent value and dynamically updates step, carries out respectively the static evaluation of patent value and dynamically updates.It is based on definition: the set D={d of patent file ₁, d ₂, d _n, the time set T={T that patent is delivered ₁, T ₂, T _n.Wherein, d _iThe document content of expression patent i; T _iThe application time of expression patent i.The method concrete operations are as follows:

The static evaluation step of described patent value:

Step 1.1 is calculated in the patent set similarity between patent file in twos.

The present invention intends adopting the method for value delivery in the similarity network value of patent is assessed, and at first needs to set up the similarity network of patent file collection.Traditional Documents Similarity method comprises cosine similarity, Jaccard coefficient, based on the similarity calculating method of HowNet, the present invention intends carrying out as an example of the cosine similarity example the similarity of patent file and calculates.Concrete operations are as follows:

Step 1.1.1, the space vector that calculates patent file represents:

For patent file set D={d ₁, d ₂..., d _nIn any patent d _i, can represent with the space vector of one group of keyword.Its process is, at first adopt existing Chinese word segmentation software, as the participle software I CTCLAS of the Chinese Academy of Sciences, all patent files are carried out Chinese word segmentation, then according to the stop words in self-defined or public stop words dictionary removal document, wherein stop words is the function word that there is no what physical meaning, as auxiliary words of mood, adverbial word, preposition, conjunction etc.For removing lexical item after stop words, calculate the weight of each lexical item in document, its computing formula is:

$w(t_j,d_i)=\frac{\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j}+0.01)}{\sqrt{{\mathrm{\Σ}}_{t_j\∈d_i}{[\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j})+0.01]}^2}}$ Formula 1

Wherein, w (t _j, d _i) be lexical item t _jAt text d _iIn weight, also be denoted as w _ij, and tf (t _j, d _i) be word t _jAt text d _iIn word frequency, N is the sum of patent set D Patent,

Lexical item t appears in D for the patent set _jThe patent file number, denominator is normalized factor.

At last, represent each piece patent file with space vector corresponding to each lexical item, be expressed as

W wherein _ijBe lexical item t _jAt patent file d _iIn weight.

Step 1.1.2 represents according to the space vector of document, calculates the similarity between patent file in twos:

To any two patent file d _iAnd d _j, coming the similarity of measurement with the included angle cosine between the vector of its correspondence, its formula is:

$\mathrm{sim}(d_i,d_j)=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k\left(d_i\right)\×w_k\left(d_j\right)}{\sqrt{\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_i\right)\right)\×\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_j\right)\right)}}$ Formula 2

W wherein _k(d _i) be that k lexical item is at document d _iIn weight, w _k(d _j) be that k lexical item is at document d _jIn weight.

Step 1.2 according to the similarity between the patent file of step 1.1 gained, in conjunction with delivering the time of patent, builds the similarity network.

Owing to lacking citation information between Chinese patent, therefore the present invention intends adopting the related and association that deliver the time of similarity between patent to simulate potential adduction relationship between patent, think that namely the content of two patent files is similar, the time of delivering, more late patent had potential adduction relationship to the time of delivering patent early.The similarity network that the present invention intends building is the digraph that comprises all patents in the patent file set, and it determines whether there is between any two link according to the size of the similarity between patent, delivers the priority of time according to patent, determines the direction of link.Concrete method of operating is as follows:

Step 1.2.1, in the similarity network, concrete operation method is as follows with the patent compound mapping:

Set up digraph G=(V, E), wherein V is corresponding to the set D of patent file, and E is corresponding to the contact between patent file, and the contact here is that similarity and the patent between patent file delivered temporal association.

Step 1.2.2 builds the link in the patent network:

Set similarity threshold sim_thrd, for any two patent file d _iAnd d _jIf, both similarity sim (d _i, d _j)=sim_thrd, consider to set up a limit between patent i and j.If the time of delivering of patent i is more Zao than patent j, set up so patent j to the directed edge of patent i; If otherwise set up patent i to the directed edge of j.Its formula is:

e(d _i,d _j)=1 if sim(d _i,d _j)>=sim_thrd and T _i>T _j;

e(d _i, d _jThe 1 if sim (d of)=- _i, d _j)=sim_thrd and T _i＜T _j; Formula 3

e(d _i,d _j)=0 else

T wherein _iAnd T _jRepresent respectively the time that patent i and patent j deliver, e (d _i, d _j)=1 expression is set up node i to the directed edge of node j, and-1 expression is set up node j to the directed edge of node i, does not have the limit between 0 expression node i and node j.

Step 1.3, the value of all patents of initialization is calculated the value of patent according to the order of sequence successively according to the thought of value delivery.

The present invention intends adopting the importance in the PageRank algorithm to transmit thought the value of patent is assessed, if namely patent A has quoted patent B, A is to the certain patent value of B contribution.The number of times that B " is quoted " is more, and the value of B is relatively large.Here " quoting " refers to the linking relationship of setting up in the similarity network, comprises directly " quoting " and " quoting " indirectly.In addition, the present invention has also considered the impact of novelty of patent on patent value, and the value of patent can decay in time, namely patent time of delivering longer, the value of patent is relatively low.Concrete operations are as follows:

Step 1.3.1, the value of all patents of initialization:

The value of patent can decay over time, and the stated value decay factor is e ^-β, time and the decay factor value of each patent of initialization is successively delivered according to patent in 0＜β＜1 wherein.For any patent i, its initialized patent value R ₀(i) computing formula is:

R ₀(i)=1 * e ^{-β t}Formula 4

Wherein t is that patent i delivers the time till now.

The computing formula of t is:

t=T _i-T ₀Formula 5

T wherein _iBe the time that patent i delivers, T ₀Be the current time.

Step 1.3.2, calculate the value of each patent successively:

After the value of initialization patent, according to PageRank thought, the value of each patent is assessed.At first, all patents in the patent set were sorted from back to front according to its time of delivering.Then, calculate successively the value of each patent.For any patent i, its value is determined by the size of all patents of quoting it to its Value Contribution degree.For the patent j of referenced patents i, the big or small computing formula of patent value that it passes to patent i is as follows:

$\mathrm{Contribution}(j,i)\frac{\mathrm{Value}\left(j\right)}{\mathrm{De}\left(j\right)}\×\mathrm{sim}(d_i,d_j)\×e^{-\mathrm{\β}(T_j-T_i)}$ Formula 6

Wherein Value (j) is the current value of patent j, the number of the patent that De (j) quotes for patent j, the i.e. out-degree of node j in the similarity network G.

The patent value of patent i is by the Value Contribution degree sum decision of all patents of quoting i to it, and computing formula is:

Formula 7

Wherein δ is ratio of damping, and 0＜δ＜1 can by the default default value, also can be set up on their own by the user.

Described patent value dynamically update step:

Step 2.1 when new patent arrives, is upgraded the current value of each patent file according to decay factor.

The static evaluation step of patent value can be assessed the value of all patents in static network.Yet along with the evolution of time, the value of patent also can change.When new patent file arrives, must upgrade the value of all patents.At first will be with T ₀Be updated to the time that new patent arrives, and add the value of network G existing patent before to upgrade new patent according to being worth decay factor.Concrete operations are as follows:

Step 2.1.1 upgrades the current time

Suppose that the time that new patent arrives is (t+1), the previous time is t, upgrades the current time of the time of (t+1)

Its formula is:

$T_0^{t+1}=T_0^t+1$

Formula 8

Wherein

Be the current time of (t+1) time,

Be the current time of t time.

Step 2.1.2, according to being worth with the value that is worth decay factor calculating each patent under the condition of existing patent:

The value of patent is that in time evolution decays gradually.Before new patent added, the value of patent decayed in time, and its formula is:

Value ^t+1(i)=e ^-β* Value ^t(i) formula 9

Value wherein ^t+1(i) add the network G value of patent i before, Value for new patent of (t+1) time ^t(i) be the value of t time patent i.

Step 2.2 adds new patent file in patent similarity network.

Before the patent value that carries out next step upgrades, need first to have patent and join in the similarity network G.New patent file adds the method for network identical with step 1.1 and step 1.2, namely at first represent with vector space model the patent file that each is new, adopt afterwards the similarity between cosine law tolerance patent, then related according between new patent and existing patent, comprise similarity with deliver temporal related, set up corresponding limit, new patent is joined in network G.Concrete operations are as follows:

Step 2.2.1, calculate new patent file and the similarity of existing patent file:

According to the step in step 1.1, with new document carry out participle, stop words removes processing, and calculates the weight of lexical item in each new patent file according to formula 1, utilizes space vector to represent the patent file that each is new.Application of formula 2 afterwards, calculate the similarity of new patent file between existing document.

Step 2.2.2 adds patent similarity network with new patent file:

Application of formula 3 judges the relation of new patent file between existing patent successively, sets up the related of limit between new patent and existing patent in the similarity network G.

Step 2.3 is upgraded the value of each patent successively according to new patent similarity network.

When new patent arrived, certain variation can occur in the value of current patent.For by the patent file of new patent citation, its value can increase because of new quoting of patent.Concrete operations are as follows:

Step 2.3.1 upgrades the value of each patent successively, and concrete method of operating is as follows:

At first existing patent was sorted from back to front by the time, then for the patent file that is communicated with new patent, upgrade iteratively successively in order the patent value of each patent in network, its computing formula is:

Formula 10

De wherein ^t+1(j) be the out-degree of (t+1) time patent j in the G of networking.

What should illustrate is: the present invention has following main beneficial effect:

The patent valve estimating method that proposes in the one, the present invention has good validity, is described as follows:

(1) the present invention is based between patent " adduction relationship ", transmit according to using the importance with the PageRank algorithm of the grade of presentation web page and importance the algorithm that thought has proposed the assessment patent value, and added time factor according to the needs of patent novelty, the characteristics that the value evolution in time of consideration patent reduces gradually, algorithm is improved, effectively combine the many factors that affects patent value, have very high reliability and validity.

This paper delivers temporal association according to the similarity between patent and patent and has simulated potential adduction relationship between patent, has solved that in Chinese patent, quoted passage lacks the impact that brings to patent valve estimating.

The algorithm that proposes in the 2nd, the present invention has good ageing, is described as follows:

In the static evaluation step of patent value with in dynamically updating step, the time complexity of algorithm is polynomial time, can assess the value of all patents quickly, has good ageing and extensibility.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or replenish or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.

Claims (7)

1. the patent valve estimating method based on value delivery, is characterized in that, based on definition: the set D={d of patent file ₁, d ₂..., d _n, the time set T={T that patent is delivered ₁, T ₂..., T _n; Wherein, d _iThe document content of expression patent i; t _iThe application time of expression patent i; Comprise the patent value that is undertaken by the static evaluation module of patent value static evaluation step and by dynamically updating the step that dynamically updates that module carries out, specifically comprise:

The step of the static evaluation of the patent value that is undertaken by the static evaluation module of patent value:

Step 1.1 is calculated in the patent set similarity between patent file in twos;

Step 1.2 is according to the similarity between patent and the time structure similarity network of delivering;

Step 1.3, the value of all patents of initialization is calculated the value of each patent successively according to the thought of value delivery;

The step that dynamically updates that dynamically updates the patent value that module carries out by patent value:

Step 2.1 when new patent arrives, is upgraded the current value of each patent file according to the value decay factor;

Step 2.2 adds new patent file in patent similarity network;

Step 2.3 is upgraded the value of each patent successively according to new patent similarity network.

2. a kind of patent valve estimating method based on value delivery according to claim 1, is characterized in that, described step 1.1 comprises following substep:

Step 1.11, the space vector that calculates patent file represent, concrete operation method is as follows:

For patent file set D={d ₁, d ₂..., d _nIn any patent d _i, utilize the space vector of one group of keyword to represent; Its process is, at first all patent files carried out Chinese word segmentation, then removes stop words in document according to self-defined or public stop words dictionary; For removing lexical item after stop words, calculate the weight of each lexical item in document, its computing formula is:

$w(t_j,d_i)=\frac{\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j}+0.01)}{\sqrt{{\mathrm{\Σ}}_{t_j\∈d_i}{[\mathrm{tf}(t_j,d_i)\×\log (N/n_{t_j})+0.01]}^2}}$ Formula one;

Wherein, w (t _j, d _i) be lexical item t _jAt text d _iIn weight, and tf (t _j, d _i) be word t _jAt text d _iIn word frequency, N is the sum of patent set D Patent,

For lexical item t occurring in patent set D _jThe patent file number, denominator is normalized factor;

At last, represent each piece patent file with space vector corresponding to each lexical item, be expressed as

W wherein _ijBe lexical item t _jAt patent file d _iIn weight;

Step 1.12 represents according to the space vector of document, and calculate the similarity between patent file in twos, concrete operation method is:

To any two patent file d _iAnd d _j, coming the similarity of measurement with the included angle cosine between the vector of its correspondence, its formula is:

$\mathrm{sim}(d_i,d_j)=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k\left(d_i\right)\×w_k\left(d_j\right)}{\sqrt{\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_i\right)\right)\×\left(\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{w}_k^2\left(d_j\right)\right)}}$ Formula two;

W wherein _k(d _i) be that k lexical item is at document d _iIn weight, w _k(d _j) be that k lexical item is at document d _jIn weight.

3. a kind of patent valve estimating method based on value delivery according to claim 1, is characterized in that, it is characterized in that, described step 1.2 comprises following substep:

Step 1.21, in the similarity network, concrete operation method is as follows with the patent compound mapping:

Set up digraph G=(V, E), wherein V is corresponding to the set D of patent file, and E is corresponding to the contact between patent file, and the contact here is that similarity and the patent between patent file delivered temporal association;

Step 3.2, the limit of structure patent network, concrete operation method is as follows:

Set similarity threshold sim_thrd, for any two patent file d _iAnd d _jIf, both similarity sim (d _i, d _j)=sim_thrd, set up a limit between patent i and j; If the time of delivering of patent i is more Zao than patent j, set up so patent j to the directed edge of patent i; If otherwise set up patent i to the directed edge of j; Its formula is:

e(d _i,d _j)=1 if sim(d _i,d _j)>=sim_thrd and T _i>T _j;

e(d _i, d _jThe 1 if sim (d of)=- _i, d _j)=sim_thrd and T _i＜T _j; Formula three;

e(d _i,d _j)＝0 else

T wherein _iAnd T _jRepresent respectively the time that patent i and patent j deliver, e (d _i, d _j)=1 expression is set up node i to the directed edge of node j, and-1 expression is set up node j to the directed edge of node i, does not have the limit between 0 expression node i and node j.

4. a kind of patent valve estimating method based on value delivery according to claim 1, is characterized in that, described step 1.3 comprises following substep:

Step 4.1, the value of all patents of initialization, concrete operation method is as follows:

The value of patent can decay over time, and the setting-up time decay factor is e ^-β, deliver time and the decay factor value of each patent of initialization successively according to patent, for any patent i, its initialized patent value R ₀(i) computing formula is:

R ₀(i)=1 * e ^{-β t}Formula four;

Wherein t is that patent i delivers the time till now;

The computing formula of t is:

t=T _i-T ₀Formula five;

T wherein _iBe the time that patent i delivers, T ₀Be the current time;

Step 4.2 is calculated the value of each patent successively, and concrete operation method is as follows:

At first all patents in the patent set were sorted from back to front according to its time of delivering, then, calculate successively the value of each patent, for any patent i, its value is determined by the size of all patents of quoting it to its Value Contribution degree, for the patent j of referenced patents i, the big or small computing formula of patent value that it passes to patent i is as follows:

$\mathrm{Contribution}(j,i)\frac{\mathrm{Value}\left(j\right)}{\mathrm{De}\left(j\right)}\×\mathrm{sim}(d_i,d_j)\×e^{-\mathrm{\β}(T_j-T_i)}$ Formula six;

Wherein Value (j) is the current value of patent j, the number of the patent that De (j) quotes for patent j, i.e. and the out-degree of node j in the similarity network G,

The patent value of patent i is by the Value Contribution degree sum decision of all patents of quoting i to it, and computing formula is:

Formula seven;

Wherein δ is ratio of damping, and 0＜δ＜1 is arranged by system or user.

5. a kind of patent valve estimating method based on value delivery according to claim 1, is characterized in that, described step 2.1 comprises following substep:

Step 2.11 is upgraded the current time

Concrete operation method is as follows:

The time that defines new patent arrival is (t+1), and the previous time is t, upgrades the current time of the time of (t+1)

Its formula is:

$T_0^{t+1}=T_0^t+1$ Formula eight;

Wherein

Be the current time of (t+1) time,

Be the current time of t time;

Step 2.12 calculates according to being worth decay factor the value that has each patent under patent condition, and concrete method of operating is as follows: before new patent added network G, the value of patent decayed in time, and its formula is:

Value ^t+1(i)=e ^-β* Value ^t(i) formula nine;

E wherein ^-βBe decay factor, β is specified and 0＜β＜1 by the user; Value ^t+1(i) be the initial patent value of (t+1) time patent i, Value ^t(i) be the patent value of t time patent i.

6. a kind of patent valve estimating method based on value delivery according to claim 1, is characterized in that, described step 2.2 comprises following substep:

Step 2.21, calculate new patent file and the similarity of existing patent file, concrete method of operating is: according to the step in step 1.1, with new document carry out participle, stop words removes processing, and calculate the weight of lexical item in each new patent file according to formula one, utilize space vector to represent the patent file that each is new; Application of formula two afterwards, calculate the similarity of new patent file between existing document;

Step 2.22, new patent file is added patent similarity network, concrete method of operating is as follows: application of formula three judges the relation of new patent file between existing patent successively, sets up the related of limit between new patent and existing patent in the similarity network G.

7. a kind of patent valve estimating method based on value delivery according to claim 1, it is characterized in that, the concrete grammar of described step 2.2 namely upgrades the value of each patent successively, concrete method of operating is: at first existing patent was sorted from back to front by the time, then for there is the patent file in path with new patent in network, upgrade iteratively successively in order the patent value of each patent, its computing formula is:

Formula ten;

Wherein, De ^t+1(j) be the out-degree of (t+1) time patent j in the G of networking.

Images