An intellectual property risk intelligent assessment method in a technology transfer process

Abstract

This application relates to the field of intellectual property risk assessment technology and discloses an intelligent method for assessing intellectual property risks during technology transfer. The method includes the following steps: S1. Data collection and preprocessing: Collecting textual data, examination history data, and related case data of the target patent; extracting technical features, examination challenge information, and invalidation criteria using text parsing technology; and constructing a standardized dataset for risk assessment. This standardized dataset includes at least the basic data required for technical feature difference degree D and technical effect improvement degree E. Through a dual-algorithm collaborative mechanism, the first assessment algorithm quickly filters out medium- and high-risk patents, while the second assessment algorithm supplements the analysis of potential risk points for low-risk patents. This solves the problem of "one-sided risk identification" in traditional assessments, improving the accuracy of intellectual property risk assessment in technology transfer and effectively reducing the technology transfer dispute rate caused by patent inventiveness defects.

Description

Technical Field

[0001] This invention relates to the field of intellectual property risk assessment technology, specifically to an intelligent method for assessing intellectual property risks during the technology transfer process. Background Technology

[0002] In the fields of intellectual property management and technology transactions, intellectual property risk assessment during the technology transfer process is a crucial step in ensuring transaction security.

[0003] In existing technologies, intellectual property risk assessment for technology transfer mainly employs three common technical methods: First, manual due diligence, where an assessment team composed of patent attorneys, technical experts, and legal advisors determines the risk level by studying patent documents, searching prior art, analyzing examination history, and combining industry experience; second, semi-automated analysis based on databases, utilizing the search and statistical functions of patent databases (such as CNKI Patent Database and PatSnap) to output indicators such as patent citation frequency and number of patent families to assist in assessing the stability of rights; and third, preliminary intelligent tools, which use techniques such as keyword matching and simple text comparison to identify explicit differences between patent documents and prior art and generate preliminary risk warnings.

[0004] However, manual due diligence relies on the experience of professionals, and the judgment of inventiveness defects is highly subjective. It is also difficult to efficiently identify implicit questions (such as implied inventiveness defects not directly stated in the examination opinion) and hidden problems such as the risk of multiple prior art combinations, resulting in incomplete risk identification. Semi-automated tools and preliminary intelligent tools lack in-depth verification mechanisms for evaluation results and have weak screening capabilities for potential inventiveness defects in low-risk patents, which can easily lead to misjudgment of risks. This increases the probability of disputes arising from patent invalidation after technology transfer, thus restricting the safety and efficiency of technology transfer. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent method for assessing intellectual property risks during the technology transfer process, thus solving the problems of low security and efficiency in technology transfer.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent assessment method for intellectual property risks during technology transfer, comprising the following steps:

[0007] S1. Data Acquisition and Preprocessing

[0008] Collect textual data, examination history data, and related case data of the target patent; extract technical features, examination questioning information, and invalidation determination elements through text parsing technology; and construct a standardized dataset for risk assessment.

[0009] The standardized dataset includes at least the basic data required for the technical feature difference degree D and the technical effect improvement degree E, as well as the basic data required for the existing technology combination correlation degree R1, the implicit questioning intensity R2, and the similar patent inefficiency R3.

[0010] S2, Preliminary Risk Assessment

[0011] Based on the standardized dataset of S1, the first evaluation algorithm is used to make a preliminary assessment of the inventiveness risk of the target patent and output the preliminary risk level.

[0012] The preliminary risk levels include low risk, medium risk, and high risk;

[0013] S3, Risk Correction Assessment

[0014] Given that patents initially identified as medium or high risk in S2 already have clear risk basis, in order to optimize the evaluation efficiency, for patents initially identified as low risk in S2, a second evaluation algorithm is used based on the standardized dataset of S1 to supplement the analysis of potential risk points not covered by the first evaluation algorithm. If significant risks are identified, the risk level is corrected and the corrected results are output.

[0015] S4, Comprehensive Evaluation Output

[0016] The preliminary risk level of S2 is integrated with the revised result of S3 to form the final intellectual property risk assessment report, which clarifies the risk level and the corresponding key risk basis; the key risk basis includes at least the core calculation parameters that lead to the risk level and their values.

[0017] Preferably, in S1, "text data" includes the claims, specification and drawings of the target patent, "examination history data" includes examination opinion notices and applicant response documents, and "related case data in the same field" includes patent judgments and invalidation requests that have been declared invalid in the same technical field due to inventiveness issues in the past five years.

[0018] Preferably, the "text parsing technology" in S1 includes natural language processing technology, specifically: using word segmentation tools to extract high-frequency technical terms as technical features, identifying review and questioning information through keyword matching, and extracting invalidation judgment elements through case text annotation.

[0019] Preferably, the first assessment algorithm in S2 is implemented by calculating a preliminary risk index, the formula of which is:

[0020] P = D×0.6 + E×0.4

[0021] P is the preliminary risk index (range [0,1]). The larger D and E are, the larger P is and the lower the risk.

[0022] D represents the technical feature difference degree, which is calculated as "the number of unique technical features of the target patent / the total number of technical features of the target patent" (value range [0,1]). The data comes from the standardized dataset of S1.

[0023] E represents the improvement in technical effect, and the calculation formula is "(target patent technical effect parameter - existing technology benchmark parameter) / existing technology benchmark parameter" (the value range is [0,1], and it is counted as 1 when it exceeds 1). The technical effect parameter is a quantifiable indicator extracted from the S1 dataset.

[0024] The basis for "D weight 0.6, E weight 0.4" in the first algorithm is: based on the statistical analysis of 500 authorized patents, the influence weight of technical feature differences on inventiveness (0.6) is significantly higher than that of technical effect (0.4). The optimality of this weight combination is verified by regression analysis (the accuracy is improved by 8%-12% compared with other weight combinations).

[0025] The uniqueness of the "R1 weight 0.5" in the second algorithm: In the context of technology transfer, multiple combinations of existing technologies are the primary factor leading to patent invalidity (accounting for 63% of invalidation cases), so it is given the highest weight, which is different from the conventional practice of distributing weights equally in existing technologies.

[0026] When P ≥ 0.7, it is initially judged as low risk; when 0.3 ≤ P < 0.7, it is initially judged as medium risk; when P < 0.3, it is initially judged as high risk.

[0027] Preferably, the second assessment algorithm is implemented by calculating a modified risk index, the formula of which is:

[0028] C = R1×0.5 + R2×0.3 + R3×0.2

[0029] in:

[0030] C is the modified risk index (range [0,1]), and the larger the value, the higher the potential risk;

[0031] R1 is the prior art combination correlation degree (value range [0,1]), which is calculated by the average text similarity of multiple prior art and target patent in the S1 dataset;

[0032] R2 represents the intensity of implicit challenges (range [0,1]), determined based on the number and / or type of implicit challenges identified in the S1 dataset;

[0033] R3 represents the inefficiency of similar patents (range [0,1]), which is the proportion of similar patents in the same field in the S1 dataset that are invalidated due to inventiveness.

[0034] When C≥0.6, the low risk is revised to medium risk; when C≥0.8, it is revised to high risk; when C<0.6, it remains low risk.

[0035] Preferably, the "technical effect parameters" in S2 specifically include the following extracted from the dataset in S1: processing speed (unit: times / second), material utilization rate (unit: %), and energy consumption reduction rate (unit: %). The weighted average of the improvement of each technical effect parameter is used in the calculation of E.

[0036] Preferably, the "text similarity" in S3 is calculated using the cosine similarity algorithm, with the formula: sim(A,B)= (A·B) / (||A||×||B||) where A is the feature vector of the target patent technology, B is the feature vector of a single prior art, A·B is the vector inner product, and ||A|| and ||B|| are the vector magnitudes.

[0037] Preferably, the "fusion method" in S4 is as follows: if S3 outputs a correction result (medium risk or high risk), the final risk level is the correction result; if no correction result is output, the final risk level is the preliminary risk level of S2, and the report must separately mark the P value (first algorithm) and the C value (second algorithm) and the corresponding parameters.

[0038] This invention provides an intelligent method for assessing intellectual property risks during the technology transfer process. It has the following beneficial effects:

[0039] 1. This invention employs a dual-algorithm collaborative mechanism, utilizing a first evaluation algorithm to quickly screen out medium- and high-risk patents, while a second evaluation algorithm supplements the analysis of potential risk points for low-risk patents. This solves the problem of "one-sided risk identification" in traditional evaluations, improves the accuracy of intellectual property risk assessment in technology transfer, and effectively reduces the rate of technology transfer disputes caused by patent inventiveness defects.

[0040] 2. This invention is based on standardized datasets to construct and quantify algorithm models, transforming abstract risk factors such as technical feature differences and the intensity of implicit doubts into calculable numerical parameters. It also clarifies the risk basis through structured reports, which not only eliminates the reliance on human experience, but also provides both parties in technology transfer with traceable and verifiable risk assessment basis, helping decision-makers to accurately control patent stability risks. Attached Figure Description

[0041] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example:

[0044] Please see the appendix Figure 1 This invention provides an intelligent assessment method for intellectual property risks during the technology transfer process, comprising:

[0045] S1: Data acquisition and preprocessing, used to construct the standardized dataset required for the evaluation. The specific process is as follows:

[0046] S1.1, Data collection range:

[0047] Collect textual data of the target patent (including claims, specification and drawings, where the drawings are used to help understand the structural relationship of the technical features), examination history data (including examination opinion notices and applicant's response documents, which must cover all questions raised during the examination process regarding inventiveness), and related case data in the same field (select patent judgments and invalidation requests that have been declared invalid due to inventiveness issues in the same technical field in the past five years, including at least 10 valid cases).

[0048] S1.2 Specific applications of text parsing technology:

[0049] The above data was processed using natural language processing techniques:

[0050] Technical feature extraction: Use word segmentation tools (such as jieba word segmentation) to segment the claims, and filter out technical terms that appear ≥3 times (such as "data encryption module", "risk assessment model", etc.). Combine the functional description of the technical terms in the specification to form the technical feature set of the target patent (such as feature A1, A2, A3, etc.) and the technical feature set of the prior art (such as feature B1, B2, B3, etc.).

[0051] Extraction of Examination Questions: By matching keywords (such as "insufficient inventiveness", "obvious", "inspiration from prior art", etc.), explicit questions (such as "the technical solution of claim 1 lacks inventiveness compared to prior art document 1") and implicit questions (such as "the non-obviousness of the improvement in technical effect is not clearly stated") are identified from the examination opinion notice and response documents, and the number and content of questions are recorded.

[0052] Explicit challenges: Challenges that directly include keywords such as "inventiveness" or "non-obviousness" in the examination opinion (e.g., "Claim 1 lacks inventiveness").

[0053] Implicit challenges: Semantic analysis of review comments using the BERT model to identify expressions that implicitly negate creativity, such as "common combination," "easy to think of," and "not beyond expectations" (model training data source: 5000 annotated review comments).

[0054] Invalidation determination element extraction: Text annotation of judgment documents of related cases in the same field is performed to extract the inventiveness defect elements that lead to the invalidation of patents (such as "multiple prior art combinations can directly obtain the target technical solution" and "the technical effect does not exceed the expectations of those skilled in the art"). The inefficiency of similar patents is calculated (number of invalid cases / total number of cases).

[0055] The technical feature sets of the target patent and the prior art patent are transformed into TF-IDF feature vectors:

[0056] A dictionary containing all technical terms;

[0057] Count the frequency of each term in the text and generate vectors A (target patent) and B (comparative patent).

[0058] S1.3 Construction of standardized datasets:

[0059] The extracted technical features, examination and questioning information, and invalidation determination elements are organized into a structured table with fields including "Technical Feature ID", "Feature Description", "Question Type (Explicit / Implicit)", "Invalidation Element Label", and "Data Source (as claimed in the claim paragraph, examination opinion number)" to ensure that the data can be directly accessed in subsequent evaluation steps.

[0060] S2: Preliminary risk assessment, which uses the first assessment algorithm to make a preliminary determination of creative risk. The specific process is as follows:

[0061] S2.1 Parameter calculation of the first evaluation algorithm:

[0062] Technical Feature Difference (D): The calculation formula is "number of unique technical features of the target patent / total number of technical features of the target patent". For example, if the target patent has a total of 5 technical features (A1-A5), among which A3 and A5 are unique compared with the existing technical feature set (B1-B4), then D=2 / 5=0.4;

[0063] Technical effect improvement (E): Quantifiable technical effect parameters (such as "data processing efficiency improved by 30%" or "risk identification accuracy improved by 25%) are extracted from the "Detailed Implementation" section of the specification. The calculation formula is "(target patent technical effect parameter - prior art benchmark parameter) / prior art benchmark parameter". For example, if the risk identification accuracy rate of the prior art is 60% and that of the target patent is 85%, then E = (85% - 60%) / 60% ≈ 0.42;

[0064] Specifically, the range of technical performance parameters includes, but is not limited to, processing speed, accuracy, cost, energy consumption, and material utilization rate. The specific selection is based on the characteristics of the technical field (e.g., in the field of data processing, accuracy and speed are prioritized, while in the field of manufacturing, cost and material utilization rate are prioritized).

[0065] Preliminary risk index (P): Calculated according to the formula "P=D×0.6 + E×0.4". In the example above, P=0.4×0.6 +0.42×0.4=0.24 + 0.168=0.408.

[0066] Furthermore, the weighted average of the improvement in each technical effect parameter is also included in the calculation of E:

[0067] Parameter 1: Risk identification accuracy (current 60% → target 85% → improvement 0.4167);

[0068] Parameter 2: Data processing speed (currently 100 times / second → target 120 times / second → improvement 0.2);

[0069] Weighting: Accuracy weighted at 0.7, speed weighted at 0.3 (based on the importance of the technology);

[0070] The weighted average E value = (0.4167 × 0.7) + (0.2 × 0.3) = 0.3517;

[0071] Weight determination criteria: The Analytic Hierarchy Process (AHP) was used, and five experts in the field were invited to score the importance of the parameters. The weights were determined through a consistency test (e.g., in the data processing field: accuracy 0.6, speed 0.4; in the manufacturing field: cost 0.5, material utilization 0.3, energy consumption 0.2).

[0072] S2.2 Preliminary Risk Level Assessment:

[0073] The risk level is determined by the P-value: P ≥ 0.7 indicates low risk, 0.3 ≤ P < 0.7 indicates medium risk, and P < 0.3 indicates high risk. In the example above, P = 0.408, so it is initially classified as medium risk.

[0074] S3: Risk Correction Assessment, used to supplement the analysis of potential risks for patents initially determined to be low-risk, through a second assessment algorithm. The specific process is as follows:

[0075] S3.1 Extraction of potential risk points:

[0076] Based on the standardized dataset from step S1, extract the risk points not covered by the first evaluation algorithm:

[0077] Implicit challenges: such as inventiveness defects not directly mentioned in the examination comments (e.g., "failure to explain the necessity of combining technical features A3 and A5"), count the number of such challenges;

[0078] Implicit challenges include: defects in creativity (T1), defects in specification support (T2), etc.

[0079] Relevance of existing technology combination: The text matching degree between multiple existing technologies and the target patent is calculated by the cosine similarity algorithm. The formula is "sim (A,B)=(A・B) / (||A||×||B||)" (A is the feature vector of the target patent, and B is the feature vector of the existing technology combination). The average value is taken as the relevance degree (R1).

[0080] Similar patent inefficiency (R3): the proportion of similar patents in the same field that are invalidated due to inventiveness, as statistically analyzed in step S1.

[0081] S3.2 Correction calculation of the second evaluation algorithm:

[0082] The modified risk index (C) is calculated using the formula “C=R1×0.5 + R2×0.3 + R3×0.2”, where R2 is the intensity of implicit skepticism.

[0083] For example:

[0084] One inventiveness implied defect (type T1) was identified and assigned a value of 0.4;

[0085] One instruction manual support defect (type T2) was identified and assigned a value of 0.3;

[0086] R2 = Σ(Implicit challenge assignments for each type), and takes the value 1.0 when it exceeds 1.0.

[0087] Specifically, based on the actual impact weight of various implicit challenges in invalid cases in the same field, T1 (creativity defect) is assigned a base value of 0.4 per case, and T2 (inventory support defect) is assigned a base value of 0.3 per case. When the cumulative value exceeds 1.0, it is counted as 1.0.

[0088] Risk level adjustment: If C ≥ 0.6, it is adjusted to medium risk; if C ≥ 0.8, it is adjusted to high risk; otherwise, it remains low risk. In the example above, C = 0.43, so the original low risk level is maintained.

[0089] S4: Comprehensive evaluation output, used to integrate the preliminary evaluation and revised evaluation results to form the final report:

[0090] S4.1 Result Fusion Rules:

[0091] If step S3 does not output a correction result (i.e., correction is not triggered), the final risk level is the preliminary level of step S2; if correction is triggered, the correction result shall prevail.

[0092] The risk level thresholds are set based on historical case statistical analysis. Based on the statistics of related cases (≥10) in the same field in S1, when P≥0.7, the probability of the patent being invalidated is ≤5%; when 0.3≤P<0.7, the probability of invalidation is 10%-30%; when P<0.3, the probability of invalidation is ≥70%.

[0093] S4.2, Contents of the Assessment Report:

[0094] Clearly define the final risk level (low / medium / high) and indicate the key evidence, including:

[0095] The calculation process of the first evaluation algorithm (such as D, E, P values ​​and corresponding parameters);

[0096] The second evaluation algorithm is modified based on factors such as implicit questions and the correlation value of existing technology combinations.

[0097] Risk improvement suggestions (such as "supplementing the inventive step data for technical feature A3" or "limiting the scope of features disclosed in the prior art in the claims").

[0098] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent assessment of intellectual property risks during technology transfer, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing Collect textual data, examination history data, and related case data of the target patent; extract technical features, examination questioning information, and invalidation determination elements through text parsing technology; and construct a standardized dataset for risk assessment. The standardized dataset includes at least the basic data required for the technical feature difference degree D and the technical effect improvement degree E, as well as the basic data required for the existing technology combination correlation degree R1, the implicit questioning intensity R2, and the similar patent inefficiency R3. S2, Preliminary Risk Assessment Based on the standardized dataset of S1, the first evaluation algorithm is used to make a preliminary assessment of the inventiveness risk of the target patent and output the preliminary risk level. The preliminary risk levels include low risk, medium risk, and high risk; S3, Risk Correction Assessment Given that patents initially identified as medium or high risk in S2 already have clear risk basis, in order to optimize the evaluation efficiency, for patents initially identified as low risk in S2, a second evaluation algorithm is used based on the standardized dataset of S1 to supplement the analysis of potential risk points not covered by the first evaluation algorithm. If significant risks are identified, the risk level is corrected and the corrected results are output. S4, Comprehensive Evaluation Output The preliminary risk level of S2 is combined with the revised result of S3 to form the final intellectual property risk assessment report, which clarifies the risk level and the corresponding key risk basis. The key risk assessment criteria include at least the core calculation parameters and their values ​​that lead to the risk level.

2. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The "text data" in S1 includes the claims, specification and drawings of the target patent; the "examination history data" includes examination opinion notices and applicant response documents; and the "related case data in the same field" includes patent judgments and invalidation requests that have been declared invalid in the same technical field due to inventiveness issues in the past five years.

3. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The "text parsing technology" in S1 includes natural language processing technology, specifically: using word segmentation tools to extract high-frequency technical terms as technical features, identifying review and questioning information through keyword matching, and extracting invalidation judgment elements through case text annotation.

4. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The first assessment algorithm in S2 is implemented by calculating a preliminary risk index, the formula of which is: P = D×0.6 + E×0.4 P is the preliminary risk index (range [0,1]). The larger D and E are, the larger P is and the lower the risk. D represents the technical feature difference degree, which is calculated as "the number of unique technical features of the target patent / the total number of technical features of the target patent" (value range [0,1]). The data comes from the standardized dataset of S1. E represents the improvement in technical effectiveness, calculated as "(target patent technical effectiveness parameter - existing technology benchmark parameter) / existing technology benchmark parameter" (value range [0,1], with values ​​exceeding 1 counted as 1), where the technical effectiveness parameter is a quantifiable indicator extracted from the S1 dataset. When P ≥ 0.7, it is initially determined to be low risk; When 0.3 ≤ P < 0.7, it is initially judged as medium risk; when P < 0.3, it is initially judged as high risk.

5. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The second assessment algorithm is implemented by calculating a modified risk index, the formula of which is: C = R1×0.5 + R2×0.3 + R3×0.2 in: C is the modified risk index (range [0,1]), and the larger the value, the higher the potential risk; R1 is the prior art combination correlation degree (value range [0,1]), which is calculated by the average text similarity of multiple prior art and target patent in the S1 dataset; R2 represents the intensity of implicit challenges (range [0,1]), determined according to the following rules: Implicit challenges are divided into defects in inventiveness (T1) and defects in specification support (T2). T1 is assigned a value of 0.4 per defect, and T2 is assigned a value of 0.3 per defect. When the cumulative value exceeds 1.0, it is counted as 1.

0. If there are differences in both quantity and type, the cumulative value is used for calculation. R3 represents the inefficiency of similar patents (range [0,1]), which is the proportion of similar patents in the same field in the S1 dataset that are invalidated due to inventiveness. When C≥0.6, the low risk is revised to medium risk; when C≥0.8, it is revised to high risk; when C<0.6, it remains low risk.

6. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 4, characterized in that, The "technical effect parameters" in S2 specifically include the following extracted from the S1 dataset: processing speed (unit: times / second), material utilization rate (unit: %), and energy consumption reduction rate (unit: %). The weighted average of the improvement of each technical effect parameter is used in the calculation of E.

7. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The "text similarity" in S3 is calculated using the cosine similarity algorithm, with the formula: sim(A,B) = (A·B) / (||A||×||B||), where A is the feature vector of the target patent technology, B is the feature vector of a single prior art, A·B is the vector inner product, and ||A|| and ||B|| are the vector magnitudes.

8. The intelligent assessment method for intellectual property risks in the technology transfer process according to claim 1, characterized in that, The "fusion method" in S4 is as follows: if S3 outputs a correction result (medium risk or high risk), the final risk level is the correction result; if no correction result is output, the final risk level is the preliminary risk level of S2, and the report must separately mark the P value (first algorithm) and the C value (second algorithm) and the corresponding parameters.

Images