Scientific and technological achievement supply and demand matching method and system based on index fingerprinting and dynamic bitmap indexing

Abstract

The invention discloses a scientific and technological achievement supply and demand matching method and system based on index fingerprinting and dynamic bitmap indexing, and belongs to the technical field of information retrieval and text data processing. In order to solve the problems of low matching efficiency, insufficient accuracy and insufficiency in interpretability and the like caused by fuzzy expression of enterprise demands and fragmentation of achievement index description, the specific scheme of the invention comprises the following steps of: firstly, performing normalization processing on index terms of achievements and demand texts, generating binary index fingerprints and constructing an achievement dynamic bitmap index; secondly, analyzing the demand to generate a demand fingerprint and quantitative constraint information; performing dynamic gating pruning based on the coverage degree through multi-strategy mixed recall preliminary screening; and finally, fusing the semantic similarity and the index satisfaction degree to carry out score sorting, and outputting an interpretable matching result. According to the method, rapid preliminary screening is achieved through fingerprinting bitmap indexing, the docking efficiency and the result reliability are improved in combination with multi-strategy fusion, and the defects of an existing method in the aspects of matching efficiency, accuracy and interpretability are effectively overcome.

Description

Technical Field

[0001] This invention relates to the field of information retrieval and text data processing technology, specifically to a method and system for matching the supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing. Background Technology

[0002] In the current strategic context of promoting the deep integration of technological and industrial innovation and boosting the development of new productive forces, building an application-oriented research ecosystem has become a core task for strengthening the leading role of enterprises in innovation. Precise matching of technology supply and demand, as a key link connecting high-quality technology supply with industrial applications, is a prerequisite for leveraging the supporting and leading role of technological innovation and achieving exponential industrial growth. If the matching mechanism fails, it will not only lead to the inefficient accumulation of a large amount of research resources but also exacerbate the disconnect between science and industry, severely restricting the improvement of regional core competitiveness.

[0003] Existing matching methods have evolved from manual screening to keyword retrieval platforms and semantic retrieval solutions, but efficiency bottlenecks remain. While AI solutions have been introduced for matching, they are often simplified to text similarity calculations, failing to overcome the semantic differences between colloquial enterprise needs and professional descriptions of results. Noise in expression leads to non-standardized training data, a lack of unified dimensions for technical indicators making quantitative comparison difficult, and a lack of interpretability and robustness in matching models, introducing uncertainty and risk into enterprise decision-making.

[0004] The paper "Research on a Two-Stage Technology Supply and Demand Matching Method Based on Text Matching-Configuration Optimization" (Journal: Journal of Information Science) proposes a two-stage matching framework that first performs technology text matching and then combines it with configuration optimization. This framework has been validated in a real technology market scenario, demonstrating its ability to improve the feasibility of supply and demand matching to some extent. However, this framework primarily addresses the macro-level screening and decision-making problems caused by excessive matching results and complex factors. It does not focus on the unified definition, unit conversion, hard threshold verification, or bit-level rapid filtering mechanism for massive databases of quantifiable indicators in enterprise technology needs. To address this issue, this invention uses a chain of "indicator normalization → fingerprinting → bitmap indexing → gating pruning → interpretable scoring" to move the searchability, verifiability, and interpretability of hard indicators to the recall and gating stages, thus meeting the needs of enterprises for aligned and quantifiable comparison of technology indicators.

[0005] The patent "Intelligent Recommendation Method for Scientific and Technological Achievements Based on Feature Clustering and Similarity Calculation Matching" (Publication No.: CN119760221A) proposes to extract key data and textual semantic information from scientific and technological achievements, perform domain identification and cluster feature word construction, and combine user browsing and demand analysis to achieve recommendations, which can improve the coverage and efficiency of achievement recommendations. However, the core of this solution is still driven by "clustering + similarity / preference", lacking a unified structured expression and verifiable satisfaction determination for the hard indicator constraints in enterprise needs. To address this issue, this invention maps achievements and needs to a standardized indicator space, generates K-bit binary fingerprints, and uses compressed bitmap indexing + bitwise operations + coverage gating to first eliminate candidates that are "not covered by the indicator dimension" in the recall stage, and then integrates semantic similarity and indicator satisfaction to output ranking and explanation reports, thereby balancing the reliability of hard constraints, retrieval efficiency and interpretability.

[0006] Therefore, there is an urgent need for a supply and demand matching method and system that can unify the indicator criteria and enable rapid screening of massive results, so as to obtain more accurate and interpretable matching results under complex semantic differences and indicator constraints. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a method and system for matching the supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing. This solves the problems in existing technologies, such as the difficulty in standardizing the expression of enterprise needs, the inconsistency in supply and demand terminology leading to poor matching, and the low efficiency of searching massive amounts of achievements while failing to guarantee the matching of key indicators. By normalizing and fingerprinting indicators, using compressed bitmap gating for screening, and integrating semantic similarity and indicator satisfaction to output interpretable results, this invention is suitable for the accurate matching of scientific and technological achievement databases and technology demand databases.

[0008] In a first aspect, the present invention provides a method for matching the supply and demand of scientific and technological achievements based on index fingerprinting and dynamic bitmap indexing, which adopts the following technical solution: A method for matching the supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing includes: S1. Normalize the terminology of scientific and technological achievements texts, and generate a K-bit binary achievement fingerprint based on the standard indicator identifier set. Maintain the version mapping relationship between indicator identifiers and fingerprint bits, and build a compressed bitmap index library based on the result fingerprints.

[0009] Preferably, the terminology normalization in step S1 includes: pre-classifying the scientific and technological achievement text in the technical field before terminology mapping; calling the terminology-indicator association knowledge base corresponding to the field based on the pre-classification results to perform polysemous word disambiguation; calculating the mapping confidence for candidate terminology-indicator mapping and setting the mapping confidence threshold T; when the mapping confidence of a candidate terminology is lower than the threshold T, direct positioning mapping is not performed, but the candidate term is marked as a candidate match to enter keyword retrieval, omission filling, and proofreading processing; The K-bit achievement fingerprint includes core technology bits and extension bits, which are dynamically expanded through a versioned bit mapping mechanism. When a new standard indicator identifier is added, a new fingerprint bit is added to the extension bits and the version mapping relationship from the standard indicator identifier to the fingerprint bit is updated. At the same time, only the corresponding compressed bitmap is updated incrementally for the newly added fingerprint bit. The compressed bitmap index library establishes corresponding compressed bitmaps for each of the K bits of the achievement fingerprint. Each compressed bitmap records the set of scientific and technological achievement identifiers with the corresponding bit value of 1. The compressed bitmap adopts a Roaring Bitmap structure.

[0010] S2. Parse the enterprise requirement text to extract requirement indicator identifiers and quantitative constraint information, and generate a K-bit binary requirement fingerprint based on the standard indicator identifier set. ).

[0011] Preferably, the quantization constraint information in step S2 adopts a generalized constraint description specification, and the operators include at least " " " "and interval constraints" Furthermore, it supports using "Minimize / Maximize" class constraints as ranking preferences.

[0012] S3. Perform semantic retrieval and keyword retrieval in parallel to obtain the first candidate set. Based on the demand fingerprint, perform bitwise operations in the compressed bitmap index library to obtain the second candidate set. Merge and remove duplicates to form the initial candidate set.

[0013] Preferably, the bitwise operation filtering in step S3 includes: dividing the requirement indicator identifiers into a set of mandatory indicators and a set of preferred indicators; performing an intersection operation on the compressed bitmap corresponding to the set of mandatory indicators to obtain a set of mandatory candidates; performing a union operation on the compressed bitmap corresponding to the set of preferred indicators to obtain a set of bonus candidates; and obtaining a second candidate set based on the set of mandatory candidates and the set of bonus candidates. The set of indicators that must be met refers to the core technologies that the enterprise explicitly requires and that are indispensable in its needs. The set of indicators that should be given priority refers to the non-rigid indicators that the enterprise believes are better and can be considered as a bonus. The merging and deduplication process includes retaining records that meet the mandatory candidate set constraints when the first candidate set and the second candidate set contain the same scientific and technological achievement.

[0014] S4. Calculate the index coverage of the candidate scientific and technological achievements in the initial candidate set, comparing their achievement fingerprints with their demand fingerprints. The candidate set after gating is obtained based on dynamic gating threshold filtering.

[0015] Preferably, the indicator coverage in step S4 is calculated using the following formula: ; in, For the fingerprint results, For fingerprint requirements, This indicates a bitwise AND operation. This represents the count of the number of bits that are 1 in a binary vector; when At that time, coverage gating skips or Set to 1; The dynamic gating threshold is adaptively determined based on the candidate set coverage distribution, and at least satisfies one of the following: the dynamic gating threshold is taken as a preset quantile of the coverage distribution; the dynamic gating threshold satisfies... , These are preset parameters; When the size of the initial candidate set exceeds a preset threshold, sampling is used to estimate the coverage distribution to determine the dynamic gating threshold.

[0016] S5. Obtain the achievement values ​​of candidate scientific and technological achievements indicators from the gated candidate set, and calculate the comprehensive score by combining semantic similarity and indicator satisfaction. The matching results are sorted and output along with interpretable matching information.

[0017] Preferably, the indicator achieved in step S5 is ( ), obtained from the value-associative storage structure; The value-associated storage structure stores the indicator realization values ​​and unit information of candidate scientific and technological achievements using scientific and technological achievement identifiers and indicator identifiers as keys. The satisfaction level of the aforementioned indicators is expressed in a sigmoid form containing a monotonic gain function to reduce the saturation effect for any demand indicator. Single indicator satisfaction: ; in, For candidate scientific and technological achievements in terms of indicators The realized value, Demand for indicators The target value, For the Sigmoid function, Indicates the polarity of the indicator. It is a monotonic gain function and is either a logarithmic gain function or a piecewise gain function. It is a scale parameter and is determined by at least one of the target value scale, the baseline value scale, or the historical distribution scale; The overall score is calculated using a benefit-risk separation model, and is based on the following formula: ; in, This is the normalized value of semantic similarity. The normalized value of indicator satisfaction. Baseline confidence level, These represent the achieved values ​​of the corresponding indicators for the candidate scientific and technological achievements. The baseline value corresponding to the indicator. To prevent zero constants, the penalty term is only applied when... Triggered at time; The baseline value Based on statistical inferences derived from industry, company size, and technology sector, when the sample size is... At times, the statistical scope is expanded to implement downgraded inferences; The baseline confidence It decreases as the bucket dispersion increases and increases as the sample size increases; The baseline confidence Calculated using the following formula: ; in, For the variance of the buckets, The average value of the buckets. These are preset parameters. For sample size; The interpretable matching information is a structured output, including coverage, missing indicator set, indicator difference information, baseline confidence, and evidence fragment information corresponding to the indicator realization value.

[0018] Secondly, a technology achievement supply and demand matching system based on indicator fingerprinting and dynamic bitmap indexing includes: a fingerprint and index construction module, used to normalize the indicator terms of the technology achievement text to generate a K-bit binary achievement fingerprint, and to construct a compressed bitmap index library based on the achievement fingerprint; The requirement processing module is used to parse enterprise requirement text to generate K-bit binary requirement fingerprints and quantitative constraint information; The hybrid recall and pruning module is used to perform semantic retrieval and keyword retrieval in parallel to obtain a first candidate set, and to perform bitwise operations on the compressed bitmap index library based on the requirement fingerprint to obtain a second candidate set. The candidate sets are merged and deduplicated to form an initial candidate set, and gating pruning is performed based on coverage and dynamic thresholds to obtain a gated candidate set. The fusion output module is used to obtain the indicator achievement values ​​of candidate scientific and technological achievements in the gating candidate set, calculate the comprehensive score by fusing semantic similarity and indicator satisfaction, and sort them, and output the matching results and structured interpretable information.

[0019] In summary, the present invention has the following beneficial technical effects: Compared with existing technologies, the present invention provides a method and system for matching supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing, which has the following beneficial effects: it unifies the fragmented natural language in enterprise needs and scientific and technological achievement texts into a standardized indicator space, enabling quantifiable comparison and efficient retrieval of needs and achievements; in the scenario of massive achievement databases, fingerprinting representation and compressed indexing can achieve low-overhead storage and fast filtering, and in the recall stage, candidates that do not meet the key indicator coverage are eliminated in advance to reduce computational costs and improve response efficiency; in the sorting stage, semantic relevance and indicator satisfaction are comprehensively considered and uncertainty is constrained to improve the accuracy and reliability of matching results; and it outputs structured and interpretable information and evidence, which facilitates verification and decision-making, thereby improving the efficiency and usability of supply and demand matching. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a technology supply and demand matching method based on index fingerprinting and dynamic bitmap indexing according to the present invention.

[0021] Figure 2 This is a structural block diagram of a technology achievement supply and demand matching system based on index fingerprinting and dynamic bitmap indexing, according to the present invention. Detailed Implementation

[0022] Example 1: A method for matching the supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing, such as... Figure 1 As shown, it includes the following steps: S1. Normalize the terminology of scientific and technological achievements texts, and generate a K-bit binary achievement fingerprint based on the standard indicator identifier set. Maintain the version mapping relationship between indicator identifiers and fingerprint bits, and build a compressed bitmap index library based on the result fingerprints; Furthermore, the specific steps of S1 include: S1.1: The text of scientific and technological achievements comes from the scientific and technological achievements database. It is obtained and entered into the database through interface calls to external data sources, data synchronization, and file import. It should be noted that each scientific and technological achievement text includes an achievement identifier ID, title, abstract, main text, source information, and extractable technical indicator values ​​and unit information; S1.2: Preprocess the text of scientific and technological achievements, including text cleaning, sentence and segmentation, word segmentation, homograph normalization, numerical format normalization and entity recognition; input the preprocessed text into the natural language processing model to extract indicator terms, numerical values ​​and unit information; It should be noted that the terminology used includes performance indicators, process parameters, material indicators, environmental adaptability indicators, etc. S1.3: Before term mapping, the resulting text is input into a text classification model for pre-classification in the technical field to obtain domain labels; Load the corresponding domain's "Terminology - Standard Indicator Identifier" related knowledge base based on the domain tag, and disambiguate synonyms, abbreviations and polysemous words; Calculate the mapping confidence level between each candidate term and the standard indicator identifier. : ; in, This is the normalized value of semantic similarity. This is the normalized value of literal similarity. This is the domain relevance normalized value. , , As preset weights, and ; When the mapping confidence is lower than the threshold T, direct mapping is not performed. Instead, the technical term is marked as a candidate match and enters the keyword retrieval stage to fill in the gaps. It then enters the manual verification process to update the associated knowledge base. S1.4: Generate a binary result fingerprint of length K based on the standard indicator identifier set. , where the i-th bit being 1 indicates that the result contains the standard indicator identifier corresponding to the i-th bit; Maintain "Standard Indicator Labels" The version mapping relationship of "fingerprint bits" is defined, and the fingerprint bits are divided into core technology bits and extended bits. When a new standard indicator is added, a new fingerprint bit is added to the extension bit and the mapping relationship is updated. The fingerprints of existing results remain unchanged, and only the newly added indicator bit is incrementally set. S1.5: Create compressed bitmaps for each of the K bits of the result fingerprint, with each bit map recording the set of result IDs where the value of that bit is 1; The compressed bitmap uses a Roaring Bitmap structure. When a new index bit is added, a corresponding bitmap is created for the new bit and updated incrementally. S2. Parse the enterprise requirement text to extract requirement indicator identifiers and quantitative constraint information, and generate a K-bit binary requirement fingerprint based on the standard indicator identifier set. ; Furthermore, the specific steps of S2 include: S2.1: Enterprise requirement texts are obtained from the requirement library and imported into the library through API calls to external data sources, data synchronization, and file import. It should be noted that each enterprise requirement document includes the requirement document, industry, enterprise size, technical field information, and constraints. S2.2: Parse the enterprise requirement text, including text preprocessing, extraction of requirement indicator identifiers and quantitative constraint information; It should be noted that the preprocessed requirement text is input into the natural language processing model to identify the indicator terms involved in the requirement and their corresponding standard indicator labels. The model output includes indicator terms, corresponding candidate standard indicator labels, extracted values ​​and units, quantization constraint operators, extraction confidence scores and their position index in the text. Quantization constraints are described using generalized constraint specifications, including the operator " " " "and interval constraints" It also supports Minimize / Maximize as sorting preferences; S2.3: Perform unit identification and standardization conversion on the extracted values ​​and units. When the unit cannot be converted or belongs to a non-standard dimension, it is recorded as "unquantifiable constraint". In the subsequent scoring stage, it will only be processed by semantic similarity or evidence matching, and will not participate in hard numerical comparison. S2.4: Generate a binary requirement fingerprint of length K based on a standard indicator identifier set. Where the i-th bit is 1, it indicates that the requirement includes the standard indicator identifier corresponding to the i-th bit. If no settable indicator is extracted from the requirement, then... ; S3. Parallel semantic retrieval and keyword retrieval are performed to obtain the first candidate set. Based on the demand fingerprint, bitwise operations are performed in the compressed bitmap index library to obtain the second candidate set. The two candidate sets are then merged and deduplicated to form the initial candidate set. Furthermore, the specific steps of S3 include: S3.1: Parallel execution of semantic retrieval and keyword retrieval to obtain the first candidate set. ; It should be noted that semantic retrieval encodes the demand text into a vector representation, performs an approximate nearest neighbor search in the result vector library to obtain Top-N candidates, and records the similarity score and the hit sentence segment; keyword retrieval obtains results containing key terms based on the inverted index, and retains the hit criteria for each candidate record, including similarity score, hit keywords, and hit field position, for interpreting the output; S3.2: Divide the demand indicators into a set of mandatory indicators and a set of preferred indicators. Perform an intersection operation on the compressed bitmap corresponding to the mandatory indicator set to obtain the mandatory candidate set. Perform a union operation on the compressed bitmap corresponding to the preferred indicator set to obtain the bonus candidate set. Combine the mandatory candidate set and the bonus candidate set to obtain the second candidate set. ; It should be noted that the set of indicators must be the core technologies that the company explicitly requires and that are indispensable in its needs, while the set of indicators to be prioritized refers to non-rigid indicators that the company believes are better and can be considered as a bonus. S3.3: Merge and Deduplication yields the initial candidate set. ; It should be noted that when the same result exists simultaneously and When necessary, retain records that satisfy the required candidate set constraint; S4. Calculate the index coverage of the candidate scientific and technological achievements in the initial candidate set, comparing their achievement fingerprints with their demand fingerprints. The candidate set after gating is obtained based on dynamic gating threshold filtering; Furthermore, the specific steps of S4 include: S4.1: For the initial candidate set Coverage of candidate results in calculation : ; in, For the fingerprint results, For fingerprint requirements, This indicates a bitwise AND operation. This represents the count of the number of bits that are 1 in a binary vector; when At that time, coverage gating skips or Set to 1.

[0023] S4.2: The dynamic gating threshold is adaptively determined based on the candidate set coverage distribution, and at least one of the following must be satisfied: the dynamic gating threshold is taken as a preset quantile of the coverage distribution; the dynamic gating threshold satisfies... , These are preset parameters; S4.3: When the initial candidate set size exceeds the threshold, the coverage distribution can be determined by sampling estimation. ,reserve The candidate sets constitute the gating post-candidate set ; S5. Gated candidate set Obtain the achievement values ​​of candidate scientific and technological achievements indicators, and calculate a comprehensive score by combining semantic similarity and indicator satisfaction. Sort the results and output the matching results along with interpretable matching information; Furthermore, the specific steps of S5 include: S5.1: Establish a value-associative storage structure, using the result ID and indicator identifier as keys to store the indicator realization value and unit information, and perform gating on the candidate set. On-demand batch query to retrieve the indicator implementation values ​​of candidate results .

[0024] S5.2: For each demand indicator Calculate the satisfaction level of a single indicator: ; in, For candidate scientific and technological achievements in terms of indicators The realized value, Demand for indicators The target value, For the Sigmoid function, Indicates the polarity of the indicator. It is a monotonic gain function and is either a logarithmic gain function or a piecewise gain function. It is a scale parameter and is determined by at least one of the target value scale, the baseline value scale, or the historical distribution scale; S5.3: Calculate the overall score Using a profit-risk separation model: ; in, This is the normalized value of semantic similarity. The normalized value of indicator satisfaction. Baseline confidence level, These represent the achieved values ​​of the corresponding indicators for the candidate scientific and technological achievements. The baseline value corresponding to the indicator. To prevent zero constants, the penalty term is only applied when... Triggered at time; S5.4: Output the list of matching results by Sort by result, each result includes: Result ID, Title, Source Information, The explanatory fields include: coverage, set of coverage indicators, set of missing indicators, indicator difference information, baseline confidence, unit conversion information, and evidence fragment information. It should be noted that the evidence fragment information includes the original text fragments that support the achievement value of the indicator or the semantic content hit, as well as their position pointers, for verification and review.

[0025] Example 2: A technology supply and demand matching system based on indicator fingerprinting and dynamic bitmap indexing, such as... Figure 2 As shown, it includes: a fingerprint and index construction module, used to normalize the indicator terms of scientific and technological achievement text to generate a K-bit binary achievement fingerprint, and to construct a compressed bitmap index library based on the achievement fingerprint; The requirement processing module is used to parse enterprise requirement text to generate K-bit binary requirement fingerprints and quantitative constraint information; The hybrid recall and pruning module is used to perform semantic retrieval and keyword retrieval in parallel to obtain a first candidate set, and to perform bitwise operations on the compressed bitmap index library based on the requirement fingerprint to obtain a second candidate set. The candidate sets are merged and deduplicated to form an initial candidate set, and gating pruning is performed based on coverage and dynamic thresholds to obtain a gated candidate set. The fusion output module is used to obtain the indicator achievement values ​​of candidate scientific and technological achievements in the gating candidate set, calculate the comprehensive score by fusing semantic similarity and indicator satisfaction, and sort them, and output the matching results and structured interpretable information.

[0026] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for matching the supply and demand of scientific and technological achievements based on indicator fingerprinting and dynamic bitmap indexing, characterized in that, Includes the following steps: S1. Normalize the terminology of scientific and technological achievements texts, generate K-bit binary achievement fingerprints based on the standard indicator identifier set, maintain the version mapping relationship from indicator identifiers to fingerprint bits, and use achievement fingerprints to build a compressed bitmap index library; S2. Parse the enterprise requirement text to extract requirement indicator identifiers and quantitative constraint information, and generate a K-bit binary requirement fingerprint based on the standard indicator identifier set; S3. Parallel semantic retrieval and keyword retrieval are performed to obtain the first candidate set. Based on the demand fingerprint, bitwise operations are performed in the compressed bitmap index library to obtain the second candidate set. The two candidate sets are then merged and deduplicated to form the initial candidate set. S4. Calculate the index coverage of the achievement fingerprint and demand fingerprint of the candidate scientific and technological achievements in the initial candidate set, and obtain the gated candidate set based on the dynamic gating threshold. S5. Obtain the achievement values ​​of candidate scientific and technological achievements indicators from the gated candidate set, calculate the comprehensive score by combining semantic similarity and indicator satisfaction, sort them, and output the matching results and interpretable matching information.

2. The method according to claim 1, characterized in that: The terminology normalization in step S1 includes: pre-classifying the scientific and technological achievement text in the technical field before terminology mapping; calling the terminology-indicator association knowledge base of the corresponding field to disambiguate polysemous words based on the pre-classification results; calculating the mapping confidence for candidate terminology-indicator mappings and setting a confidence threshold T; when the confidence of a candidate mapping is lower than the threshold T, the corresponding fingerprint position is not executed, but the candidate term is marked as a candidate matching item for subsequent keyword retrieval and correction processing.

3. The method according to claim 1, characterized in that: The K-bit binary achievement fingerprint includes core technology bits and extension bits, which are dynamically expanded through a versioned bit mapping mechanism. When a new standard indicator identifier is added, a new fingerprint bit is appended to the extension bit and the versioned mapping relationship is updated. At the same time, incremental updates are performed on the compressed bitmap index library only for the newly added fingerprint bit. The compressed bitmap index library establishes corresponding compressed bitmaps for each of the K bits of the achievement fingerprint. Each compressed bitmap records the set of scientific and technological achievement identifiers with the corresponding bit value of 1. The compressed bitmap adopts the RoaringBitmap structure.

4. The method according to claim 1, characterized in that: The quantization constraint information in step S2 adopts the generalized constraint description specification, and the operators include at least " "" "" "and interval constraints" Furthermore, it supports using "Minimize / Maximize" class constraints as ranking preferences.

5. The method according to claim 1, characterized in that: The bitwise operation filtering in step S3 includes: dividing the requirement indicator identifiers into a set of mandatory indicators and a set of preferred indicators; performing an intersection operation on the compressed bitmap corresponding to the set of mandatory indicators to obtain a set of mandatory candidates; performing a union operation on the compressed bitmap corresponding to the set of preferred indicators to obtain a set of bonus candidates; and obtaining a second set based on the set of mandatory candidates and the set of bonus candidates. During the merging and deduplication process, when the first set of candidates and the second set of candidates contain the same scientific and technological achievement, the record that satisfies the constraint of the set of mandatory candidates is retained.

6. The method according to claim 1, characterized in that: The indicator coverage in step S4 is calculated based on the bitwise AND selection result of the result fingerprint and the demand fingerprint, as well as the number of bits with a value of 1 in the demand fingerprint. The dynamic gating threshold is adaptively determined based on the candidate set coverage distribution, including: taking a preset quantile of the coverage distribution, setting it according to the relationship between the coverage mean and standard deviation, or using sampling to estimate the coverage distribution when the initial candidate set size exceeds a preset threshold.

7. The method according to claim 1, characterized in that: The storage structure for the indicator realization values ​​in step S5 uses the scientific and technological achievement identifier and the indicator identifier as keys to store the indicator realization values ​​and unit information of the candidate scientific and technological achievements; the indicator satisfaction is calculated using the Sigmoid form containing a monotonic gain function; the comprehensive score is calculated using the benefit-risk separation model, and structured interpretable matching information is output in the matching results. The interpretable matching information includes coverage, missing indicator set, indicator difference information, baseline confidence, and evidence fragment information corresponding to the indicator realization values.

8. The method according to claim 7, characterized in that: The baseline value is obtained based on statistical inference of industry, enterprise size and technology field. When the sample size of the bucket is less than the preset minimum value, the statistical scope is expanded and downgraded inference is performed. The baseline confidence decreases as the bucket dispersion increases and increases as the sample size increases, and is calculated based on the bucket variance, mean and sample size.

9. A technology achievement supply and demand matching system based on indicator fingerprinting and dynamic bitmap indexing, characterized in that, include: The fingerprint and index construction module is used to normalize the indicator terms of scientific and technological achievement texts to generate K-bit binary achievement fingerprints, and to construct a compressed bitmap index library based on the achievement fingerprints. The requirement processing module is used to parse enterprise requirement text to generate K-bit binary requirement fingerprints and quantitative constraint information; The hybrid recall and pruning module is used to perform semantic retrieval and keyword retrieval in parallel to obtain a first candidate set, and to perform bitwise operations on the compressed bitmap index library based on the requirement fingerprint to obtain a second candidate set. The candidate sets are merged and deduplicated to form an initial candidate set, and gating pruning is performed based on coverage and dynamic thresholds to obtain a gated candidate set. The fusion output module is used to obtain the indicator achievement values ​​of candidate scientific and technological achievements in the gating candidate set, calculate the comprehensive score by fusing semantic similarity and indicator satisfaction, and sort them, and output the matching results and structured interpretable information.

Images