Emerging technology identification method and device based on large-scale corpus

Abstract

The present disclosure relates to the field of data analysis technology, and specifically to a method and device for identifying emerging technologies based on large-scale corpus, the method comprising: determining a research field to construct a candidate document set, and extracting keywords from the candidate document set to obtain a candidate keyword data set; filtering the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set; calculating the emerging score value of each keyword in the candidate keyword filter set; screening the candidate keyword filter set based on the emerging score value of each keyword and a set emerging score threshold to obtain a candidate emerging technology keyword data set; processing the candidate emerging technology keyword data set using a dynamic backtracking method to obtain a target emerging technology keyword data set. The method disclosed herein improves the accuracy of identifying emerging technologies.

Description

Technical Field

[0001] The present disclosure relates to the field of data analysis technology, and in particular to a method and device for identifying emerging technologies based on large-scale corpus. Background Art

[0002] Identifying the current state of technology and perceiving its dynamics have always been an important part of theoretical and practical research in the field of technology forecasting. With the acceleration of technological evolution and the increase in the complexity of the knowledge system, it is becoming increasingly difficult to quickly identify emerging technologies with great development potential from emerging technology hotspots. Emerging technologies are innovative technologies based on science, with characteristics such as novelty, rapid growth, and uncertainty. They may have the potential to create a new industry or change an old industry. Identifying and evaluating emerging technologies is critical to the formulation of development strategies and technological iteration and updating.

[0003] At present, the complex and ever-changing technological environment has put forward higher requirements for the prediction and research of potential emerging technologies. Faced with the current international environment of technological monopoly and technological borderization, whether domestic knowledge-intensive enterprises and government departments can effectively identify technological opportunities and break through technological barriers will directly affect the technological innovation competitiveness of enterprises, industries and the entire country. Faced with the continued slowdown in world economic growth, innovation is an important strategy in the new round of scientific and technological revolution and an important engine of economic growth. Grasping the global technological development trends and dynamically evaluating global emerging technologies are the key to building a world science and technology power and seizing the breakthrough of scientific and technological innovation. In the process of planning the national science and technology strategic layout, following the forefront of science and technology, exploring the defense line of emerging research and development, formulating development policies and leading the direction of scientific research are important scientific bases. From the national level, strengthening the identification of emerging technologies will help optimize the overall scientific research layout and make long-term plans; from the enterprise level, strengthening the identification of emerging technologies will help determine R&D priorities, investment directions and reduce investment risks.

[0004] Many scholars at home and abroad have explored methods for identifying and predicting emerging technologies. Initial research and exploration focused on the concept and definition of emerging technologies. For example, George Day and Paul Shoemaker Walton. On the Management of Emerging Technologies [M]. (Translated by Shi Ying. Beijing: Huaxia Publishing House, 2002.) Emerging technologies have attracted widespread attention. Relevant scholars define emerging technologies as a kind of scientific innovation that has the potential to establish new industries or transform existing industries. Among existing technologies, there are four dimensions that define the characteristic attributes of emerging technologies: novelty, persistence, community, and growth; there are also five characteristics of emerging technologies: radical novelty, relatively fast growth, coherence, prominent impact, and uncertainty ambiguity. Emerging technologies are defined as innovative technologies that can gather a large amount of resources in a short period of time and grow rapidly in the process of scientific research, and have the potential to influence future economic and social development. Theoretical research on emerging technologies analyzes from the perspective of technological evolution, laying the foundation for further technology identification and prediction.

[0005] With the development of bibliometrics and artificial intelligence, building scientific and effective technology evaluation and prediction models based on big data will help improve the management capabilities of technological innovation. In recent years, emerging technology mining results that combine classical theories of scientific evaluation research with new generation information technology have continued to emerge. The results of emerging technology mining in existing technologies include, for example, using online media Wikipedia to predict emerging trends and studying technology trends and life cycles related to the emerging Hadoop market; conducting text mining on multiple online community data in the field of renewable energy, and predicting future development trends in the field by analyzing the discussion content of community members; proposing key attributes for determining the emergence of technology, namely novelty, persistence, growth, and attention, which provides a reference for the conceptual analysis and prediction of emerging technologies. However, this method treats key terms "equally" and does not take into account the influence of publication authority (top journals and top conferences should have higher weights). At the same time, the research lacks analysis of patent data; Huang Lu, Zhu Yihe, and Zhang Yi. "Research on Emerging Technology Topic Identification Based on Weighted Network Link Prediction" ([J]. Journal of Information Science, 2019, 38(4): 335-341) proposed an emerging topic identification method based on weighted network link prediction, and identified emerging technologies from the two perspectives of technology novelty and technology influence. This method only evaluates patent data and only considers the novelty and influence of technology in terms of analysis dimensions, without measuring its growth and persistence. At the same time, in the field of emerging technology assessment, researchers currently focus more on a single data source, and research on the deep integration of multi-source data is still relatively rare.

[0006] In addition, in the work related to science and technology assessment, there is a phenomenon: scientific researchers inevitably need to cite classic technologies when writing papers and patents, which will lead to a high frequency of citations for classic technologies. This will also lead to classic technologies usually being in the top 10 among candidate terms, and there is a high probability that they will occupy the top of the emerging technology list for a long time. On the other hand, the current common field division method is more inclined to macro classification, and does not obtain enough fine-grained research directions in large fields (for example, the field of computer science can be subdivided into AI at most, but AI can currently be subdivided into two major categories: natural language processing (NLP) and computer vision (CV). Usually, there will be many similar candidate terms in popular fields, which leads to the phenomenon that the traditional method based on word frequency will result in a cluster of candidate technologies in the initial selection, and the diversity of candidate technologies is insufficient, and there is a lot of redundancy and repetition. In summary, the accuracy of the technology identification results of the existing emerging technology identification methods needs to be improved. Summary of the invention

[0007] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0008] To this end, the first objective of the present disclosure is to propose an emerging technology identification method based on large-scale corpus to improve the accuracy of emerging technology identification.

[0009] The second objective of the present disclosure is to propose an emerging technology identification device based on large-scale corpus.

[0010] A third objective of the present disclosure is to provide an electronic device.

[0011] To achieve the above objectives, the first embodiment of the present disclosure proposes a method for identifying emerging technologies based on a large-scale corpus, comprising:

[0012] Determine a research field, construct a candidate document set in the research field, and perform keyword extraction on the candidate document set to obtain a candidate keyword data set;

[0013] Filtering the candidate keyword data set according to the number of candidate documents in the candidate document set and relevant information of the keyword to obtain a candidate keyword filtered set;

[0014] Calculating the emerging score value of each keyword in the candidate keyword filtering set;

[0015] The candidate keyword filter set is screened based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set;

[0016] The candidate emerging technology keyword data set is processed by a dynamic backtracking method to obtain a target emerging technology keyword data set.

[0017] The method of the disclosed embodiment determines the research field to construct a candidate document set, and extracts keywords from the candidate document set to obtain a candidate keyword data set; filters the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set; calculates the emerging score value of each keyword in the candidate keyword filter set; filters the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set; processes the candidate emerging technology keyword data set using the dynamic backtracking method to obtain a target emerging technology keyword data set. In this case, the keywords are filtered using the number of candidate documents in the candidate document set, the relevant information of the keywords, the emerging score value of each keyword, and the set emerging score threshold, and the filtered keywords are processed using the dynamic backtracking method, which reduces the dominance of traditional technology keywords in the keyword set, thereby improving the accuracy of emerging technology identification.

[0018] In an emerging technology identification method based on a large-scale corpus in an embodiment of the first aspect of the present disclosure, the candidate keyword data set is filtered according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set, including: determining a research time range, and determining a basic period, a recent period and an active period based on the research time range; calculating the first number of occurrences of each keyword in the candidate keyword data set in the active period; calculating the second number of occurrences of each keyword in the candidate keyword data set in the basic period, and determining a first evaluation ratio of each keyword based on the first number of occurrences and the second number of occurrences of each keyword; calculating the number of institutions that use each keyword in the candidate keyword data set; filtering the candidate keyword data set based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set.

[0019] In an emerging technology identification method based on a large-scale corpus of an embodiment of the first aspect of the present disclosure, the candidate keyword data set is filtered based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set, including: determining a first active threshold and a second active threshold based on the number of candidate documents, and calculating a third number of occurrences of all keywords during the active period and a fourth number of occurrences of all keywords during the base period, obtaining a second evaluation ratio based on the third number of occurrences and the fourth number of occurrences, and obtaining an evaluation threshold based on the second evaluation ratio and the second active threshold; filtering out keywords whose first number of occurrences is less than the first active threshold, keywords whose first evaluation ratio is less than the evaluation threshold, and keywords whose number of institutions is less than a preset number threshold, thereby obtaining a candidate keyword filter set.

[0020] In an emerging technology identification method based on a large-scale corpus in an embodiment of the first aspect of the present disclosure, the calculation of the emerging score value of each keyword in the candidate keyword filter set includes: calculating the first change trend of each keyword in the candidate keyword filter set in the recent period; calculating the second change trend of each keyword in the candidate keyword filter set in the active period; calculating the third change trend of each keyword in the candidate keyword filter set in the research time range; setting quality parameters, and obtaining the emerging score value of each keyword based on the quality parameters, the first change trend, the second change trend and the third change trend.

[0021] In an emerging technology identification method based on a large-scale corpus in an embodiment of the first aspect of the present disclosure, setting quality parameters includes: determining the publishing journals and publishing organizations corresponding to the candidate documents containing each keyword in the candidate keyword filtering set; and obtaining the quality parameters based on the average journal influence of the publishing journals and the average research institution ranking of the publishing organization.

[0022] In an emerging technology identification method based on a large-scale corpus according to an embodiment of the first aspect of the present disclosure, the candidate emerging technology keyword dataset is processed using a dynamic backtracking method to obtain a target emerging technology keyword dataset, including: setting corresponding year backtracking parameters and order down adjustment parameters based on different research fields; selecting year backtracking parameters and order down adjustment parameters corresponding to the research field to which the candidate document set belongs, and using the dynamic backtracking method to perform a descending operation on the candidate emerging technology keyword dataset to obtain a target emerging technology keyword dataset.

[0023] In a large-scale corpus-based emerging technology identification method according to an embodiment of the first aspect of the present disclosure, a maximum marginal correlation algorithm is used to optimize the target emerging technology keyword data set.

[0024] In a method for identifying emerging technologies based on large-scale corpus in an embodiment of the first aspect of the present disclosure, keyword extraction is performed on the candidate document set to obtain a candidate keyword data set, including: using an automatic phrase mining algorithm and a word frequency-inverse document frequency algorithm to extract keywords from the candidate document set to obtain a candidate keyword data set.

[0025] To achieve the above-mentioned purpose, the second aspect of the present disclosure proposes an emerging technology identification device based on large-scale corpus, including:

[0026] An acquisition module is used to determine a research field, construct a candidate document set in the research field, and perform keyword extraction on the candidate document set to obtain a candidate keyword data set;

[0027] A filtering module, configured to filter the candidate keyword data set according to the number of candidate documents in the candidate document set and relevant information of the keyword to obtain a candidate keyword filtered set;

[0028] A calculation module, used to calculate the emerging score value of each keyword in the candidate keyword filter set;

[0029] A screening module, configured to screen the candidate keyword filter set based on the emerging score value of each keyword and a set emerging score threshold to obtain a candidate emerging technology keyword data set;

[0030] The processing module is used to process the candidate emerging technology keyword data set by adopting a dynamic backtracking method to obtain a target emerging technology keyword data set.

[0031] In the device of the disclosed embodiment, the acquisition module determines the research field to construct a candidate document set, and extracts keywords from the candidate document set to obtain a candidate keyword data set; the filtering module filters the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set; the calculation module calculates the emerging score value of each keyword in the candidate keyword filter set; the screening module screens the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set; the processing module processes the candidate emerging technology keyword data set using a dynamic backtracking method to obtain a target emerging technology keyword data set. In this case, the keywords are screened using the number of candidate documents in the candidate document set, the relevant information of the keywords, the emerging score value of each keyword, and the set emerging score threshold, and the screened keywords are processed using a dynamic backtracking method, which reduces the problem of traditional technology keywords dominating the keyword set, thereby improving the accuracy of emerging technology identification.

[0032] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the emerging technology identification method based on large-scale corpus of the first aspect embodiment of the present disclosure.

[0033] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0035] Figure 1 A flowchart of a method for identifying emerging technologies based on large-scale corpus provided in an embodiment of the present disclosure;

[0036] Figure 2 A schematic diagram of the division of the basic period, recent period and active period provided in an embodiment of the present disclosure;

[0037] Figure 3 A schematic diagram of the structure of an emerging technology identification device based on large-scale corpus provided in an embodiment of the present disclosure;

[0038] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the embodiments of the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the embodiments of the present disclosure as detailed in the appended claims.

[0040] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0041] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. It should also be understood that the term "and / or" used in the present disclosure refers to and includes any or all possible combinations of one or more associated listed items.

[0042] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0043] Figure 1 The following is a flow chart of a method for identifying emerging technologies based on large-scale corpus provided by an embodiment of the present disclosure. Figure 1 As shown, the emerging technology identification method based on large-scale corpus includes the following steps:

[0044] Step S101, determine the research field, construct a candidate document set in the research field, and extract keywords from the candidate document set to obtain a candidate keyword data set.

[0045] In step S101, there is at least one research field. For each research field, candidate documents of the corresponding research field are obtained from a large-scale corpus to construct a corresponding candidate document set. The candidate document set includes multiple candidate documents, and the multiple candidate documents may include different types of technical documents such as paper documents and patent data. Keywords can also be called terms, and the candidate keyword data set includes terms of paper documents and terms of patent data.

[0046] In step S101, keywords are extracted from the candidate document set to obtain a candidate keyword data set, including: using an automatic phrase mining algorithm and a term frequency-inverse document frequency algorithm to extract keywords from the candidate document set to obtain a candidate keyword data set. Therefore, a keyword preprocessing method combining an improved AutoPhrase (automatic phrase mining algorithm) and TF-IDF (TermFrequency-inverse Document Frequency) is used to process the keywords of patent data and paper documents to construct a candidate keyword data set, which provides an important foundation and basis for subsequent emerging technology evaluation. In addition, compared with the keywords in the existing methods, the candidate keyword data set obtained by combining paper documents and patent data has a more comprehensive data source, making the evaluation results of the subsequent discovery and identification of emerging technologies more credible (i.e., more accurate).

[0047] Specifically, the keyword extraction statistics of the candidate document set include term extraction of patent data and term extraction of paper documents. For the term extraction of patent data: since the patent data itself does not provide keywords, the widely used AutoPhrase in the industry is used in step S101 to extract keywords from patent data. AutoPhrase is a keyword extraction algorithm published by the team of knowledge mining expert Han Jiawei in the top international journal TKDE (IEEE Transactions on Knowledge and Data Engineering) in 2018; the extracted keywords also need to be initially filtered using TF-IDF. In addition, taking English terms as an example, considering that the part of speech of a term is usually a combination of nouns and adjectives, and the term length is usually 1-3 words, the results after the initial filtering can also be first screened out to find keywords with a phrase length of 1-3, and then the results obtained after part of speech matching filtering are used as candidate terms for patent data. Therefore, the keywords are initially filtered by preprocessing the part of speech matching and phrase length, which can ensure the reliability of the extraction results.

[0048] Term extraction for paper documents: In order to effectively avoid the influence of common words on keywords, the correlation between keywords and articles is improved. In order to avoid the influence of common words on the index calculation, TF-IDF can be used in step S101 to extract the terms of paper documents, wherein TF-IDF is a statistical analysis method for keywords, which is used to evaluate the importance of a word to a file set or a corpus. The importance of a word is proportional to the number of times it appears in the article and inversely proportional to the number of times it appears in the corpus. When TF-IDF is used for extraction, the IDF value of the term is mainly calculated, and the terms with IDF values ​​lower than the preset value (the preset value is, for example, 1) are filtered out, and the remaining terms are regarded as the terms of the paper document. In addition, since the paper document also has its own keywords, these keywords can comprehensively and accurately reflect the subject of the document, and the built-in keywords and the keywords filtered by TF-IDF can be used as the keywords of the paper document.

[0049] In some embodiments, the candidate document sets corresponding to each research field in step S101 can be obtained from Web of Scinence (Science Web) and USPTO (United States Patent and Trademark Office). Specifically, the research field and research time range are determined, and patent data and paper documents corresponding to the research field and research time range are collected from Web of Scinence and USPTO to form a candidate document set D. After preprocessing from the candidate document set D, AutoPhrase is used to extract the terms of the patent data, and the TF-IDF screening results are combined with keywords to obtain the terms of the paper documents, thereby obtaining a candidate keyword data set.

[0050] In this embodiment, the research field can involve various fields, and the research time range can be dynamically adjusted as needed. Therefore, the method disclosed in the present invention has the adaptability of field and time scale, and can extract, evaluate and screen emerging terms in all fields and any time period.

[0051] In some embodiments, the research field may be, for example, computer science, and the classification number corresponding to the research field is G06. In other embodiments, based on the research conclusion, the research time range may be ten years. For example, 2011-2020 may be selected, and then the papers and documents from 2011-2020 may be downloaded from Web of Science first, and then the patent data from 2011-2020 with the classification number of G06 may be retrieved and downloaded from USPTO, and the candidate document set D may be formed using the downloaded patent data and papers and documents.

[0052] In some embodiments, the terms in the candidate keyword data set may be arranged by year, and the number of times each term appears may be counted.

[0053] In some embodiments, if the term is in English, in order to reduce the problem of inconsistent word formats, the words can be unified into lowercase to reduce the word repetition problem (ie, reduce the semantic repetition problem).

[0054] Step S102, filtering the candidate keyword data set according to the number of candidate documents in the candidate document set and relevant information of the keywords to obtain a candidate keyword filtered set.

[0055] In step S102, the candidate keyword data set is filtered according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set, including: determining the research time range, and determining the basic period, recent period and active period based on the research time range; calculating the first number of occurrences of each keyword in the candidate keyword data set in the active period; calculating the second number of occurrences of each keyword in the candidate keyword data set in the basic period, and determining the first evaluation ratio of each keyword based on the first number of occurrences and the second number of occurrences of each keyword; calculating the number of institutions that use each keyword in the candidate keyword data set; filtering the candidate keyword data set based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set.

[0056] In step S102, the research time range of the candidate keyword data set is the research time range of the candidate documents in the candidate document set. The research time range is a time period and can be represented by the symbol T. Figure 2 This is a schematic diagram of the division of the basic period, recent period and active period provided by the embodiment of the present disclosure. Figure 2 As shown in the figure, the time period T can be divided into the basic period A, active period B and recent period C, where the basic period A is the first 30% of the time in the time period T; the active period B is the most recent 70% of the time in the time period T; and the recent period C is the most recent 40% of the time in the time period T. The active period B includes the recent period C, and the basic period A and the active period B constitute the entire time period T. If expressed in time t, the time t corresponding to the basic period A (i.e., the tth year) ranges from 0.1T to 0.3T, the time t corresponding to the active period B ranges from 0.4TT, and the time t corresponding to the recent period C ranges from 0.7TT. The time period T and time t are generally in years. If the research time range T is 10 years, the basic period A is 3 years (i.e., the first to the third year), the active period B is 7 years (i.e., the fourth to the tenth year), and the recent period C is 4 years (i.e., the seventh to the tenth year). If the research time range is 2011-2020, the base period is 2011-2013, the active period is 2014-2020, and the recent period is 2017-2020. In this case, since a shorter period of time will produce noisier results (i.e., too sensitive to specific treatments), and a longer period is contrary to the goal of seeking current emerging technologies, a research time range of 10 years is moderate and is conducive to improving the accuracy of subsequent identification of emerging technology keywords.

[0057] After dividing into three periods, it is also necessary to count the number of candidate documents in each period. For example, the number of candidate documents in the active period can be represented by the symbol N Active The number of recent candidate documents can be represented by the symbol N Recent express.

[0058] In step S102, the candidate keyword data set is filtered based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set, including: determining a first active threshold and a second active threshold based on the number of candidate documents, and calculating a third number of occurrences of all keywords during the active period and a fourth number of occurrences of all keywords during the base period, obtaining a second evaluation ratio based on the third number of occurrences and the fourth number of occurrences, and obtaining an evaluation threshold based on the second evaluation ratio and the second active threshold; filtering out keywords whose first number of occurrences is less than the first active threshold, keywords whose first evaluation ratio is less than the evaluation threshold, and keywords whose number of institutions is less than a preset number threshold, thereby obtaining a candidate keyword filter set.

[0059] In step S102, the candidate keyword data set is filtered based on the first occurrence number, the first evaluation ratio, the number of institutions and the number of candidate documents, and is mainly evaluated from three dimensions: technical persistence, growth and attention, so as to obtain a candidate keyword filter set.

[0060] Specifically, for technical persistence, the first occurrence number and the first active threshold of each keyword in the candidate keyword dataset are used for evaluation. The basic idea of ​​persistence is that the more times a term appears in a time period T, the more active the term is in time period T. If each term is represented by symbol i, the number of occurrences of the term at time t is represented by symbol n. it Indicates that the number of occurrences of all terms at time t is represented by N t Represents. For each term i, calculate the number of times the term appears each year, then calculate the first number of occurrences of the keyword (i.e., the term) during the active period, and combine it with the first active threshold to determine whether the persistence of the term meets the requirements, where the first active threshold is determined based on the number of candidate documents during the active period. If the first number of occurrences of the keyword during the active period is greater than or equal to the first active threshold, the keyword meets the persistence index, and the keyword is retained and the growth index is judged. Otherwise, the keyword does not meet the persistence index and is deleted. That is, the evaluation of technical persistence needs to meet the following requirements:

[0061]

[0062] where the first occurrence of term i is The first active threshold can be 7×N Active÷100000. Since the number of candidate documents published in different fields varies, the first active threshold in this embodiment can be dynamically adjusted according to the total number of candidate documents in the active period of the selected research field. In this way, the threshold of the evaluation index for terms in different research fields can be automatically matched and adapted. For example, in the field of computer science, the research time range is 10 years. Since the number of papers and patent data in this research field is large every year, the first active threshold is set to 14, that is,

[0063] As for growth, it is mainly evaluated through the first evaluation value and the evaluation threshold. Because growth generally considers two types of growth. One is absolute growth, which is measured by the average number of records containing emerging technology terms; the other is relative growth, which is measured by the proportion of records containing emerging terms in the total data set. Terms with higher relative growth account for a larger proportion than other terms; in addition, it is more promising and valuable to regard terms with higher relative growth as emerging technologies of real interest. For each term i (i.e. keyword i), determine whether the growth rate of the term in the base period and the active period is greater than or equal to a dynamic multiple value of the growth rate of all terms in the corresponding period. If it is greater than or equal to, the keyword meets the growth index, retains the keyword and judges the attention index, otherwise the keyword does not meet the growth index and is deleted. Among them, the growth rate in the base period and the active period refers to the ratio of the number of occurrences in the active period to the number of occurrences in the base period. That is, the evaluation of growth needs to meet:

[0064]

[0065] Among them, the second occurrence number of keyword i is The third occurrence of all keywords is The fourth occurrence of all keywords is The evaluation ratio is obtained based on the third occurrence number and the fourth occurrence number. β represents the second active threshold, and the first evaluation value of keyword i is given by The evaluation threshold is obtained by get.

[0066] In some embodiments, based on statistical experience, the basic value of the second activity threshold β can be set to 1.5. Specifically, the second activity threshold satisfies β=1.5×N Active÷100000. The second active threshold can reflect the activity level (i.e., field prosperity) of different research fields at different times. Since the number of candidate documents published in different fields is different, the second active threshold in this embodiment can be dynamically adjusted according to the total number of candidate documents in the active period of the selected research field. In this way, the threshold of the evaluation index for terms in different research fields can be automatically matched and adapted. For example, in the field of computer science, when the first active threshold is 14, the corresponding second active threshold is 3. For attention, it is mainly to judge whether the term is very popular and whether it is recognized by multiple institutions. Considering the use of the breadth of the community network as a standard for treating emerging technology topics, as the network covers more users, the network becomes more valuable. Network effect is a feature of emerging technologies, which means that the value of emerging technologies increases with the number of scholars participating in them, and the community can be regarded as a network. The expansion of the network related to a specific technology indicates future activities and further development. The basic idea is that the identified emerging research is more attractive than other research topics in the field of technology. Therefore, the attention can be evaluated by the number of institutions and the preset number threshold. Among them, the preset number threshold is a dynamic threshold, and the value of the threshold is related to the research field. To evaluate the attention, we need to determine how many organizations (i.e., institutions) have used term i. First, we need to define whether term i is used by organization m. The usage result is expressed as y mi Indicates that:

[0067]

[0068] Then determine whether the attention of term i meets the requirements, that is:

[0069]

[0070] Where M is the set of organizations m. Considering that a term of an emerging technology should be used by multiple organizations even in different fields, the preset number threshold can be set to 2. If the number of occurrences of a keyword in a certain year exceeds 20,000, the preset number threshold can be appropriately increased, for example, the preset number threshold can be set to 5.

[0071] The term set filtered by the above three dimensions (i.e., the candidate keyword filter set) can be preliminarily identified as a term set with an emerging degree (i.e., meeting the basic conditions of emerging terms). The terms in the candidate keyword filter set are then scored to screen out candidate emerging technology keyword data sets that represent the most emerging hot technologies in the research field.

[0072] Step S103, calculating the emerging score value of each keyword in the candidate keyword filtering set.

[0073] In step S103, the emerging score value of each keyword in the candidate keyword filter set is calculated, including: calculating the first change trend of each keyword in the candidate keyword filter set in the recent period; calculating the second change trend of each keyword in the candidate keyword filter set in the active period; calculating the third change trend of each keyword in the candidate keyword filter set in the research time range; setting quality parameters, based on the quality parameters, the first change trend, the second change trend and the third change trend, obtaining the emerging score value of each keyword. Thus, the emerging degree of the keyword can be quantified by the emerging score value, so as to achieve the purpose of efficiently and accurately calculating the emerging technology in the corresponding field.

[0074] In step S103, the first change trend of keyword i specifically compares the changes of keyword i in the recent two years (i.e., the corresponding time t range is 0.9TT) and the previous two years (i.e., the corresponding time t range is 0.7T-0.8T). The first change trend of keyword i can also be called the recent trend score of keyword i. i ), the first change trend of keyword i satisfies:

[0075] In step S103, the second change trend of keyword i specifically compares the change of keyword i from the last three years of the active period (i.e., the corresponding time t range is 0.8TT) to the first three years of the active period (i.e., the corresponding time t range is 0.4T-0.6T). The second change trend of keyword i can also be called the active period trend score of keyword i (Active Trend Score). i ), the second change trend of keyword i satisfies:

[0076] In step S103, the third change trend of keyword i is specifically the change in the entire time period T. The third change trend is an average change trend, which can be reflected by the slope. The third change trend of keyword i can also be called the average change trend score of keyword i (Slope i )The third change trend of keyword i satisfies: Among them, n iT Indicates the number of occurrences of keyword i when the value of t is T, n i0.7T represents the number of occurrences of keyword i when the value of t is 0.7T, N T Indicates the number of occurrences of all keywords when the value of t is T, N 0.7T It represents the number of occurrences of all keywords when the value of t is 0.7T. For example, for the computer science field from 2011 to 2020, the third trend satisfies:

[0077] In step S103, when calculating the emerging score value, it is taken into account that the terms in high-quality journal papers should have a higher confidence level. Therefore, a quality parameter is also set. This quality parameter can reflect the average impact factor of the publishing journal of the document where the keyword is located and the average research institution ranking of the publishing organization of the document. Through the dynamic change of this quality parameter, it can fully reflect that the technology in high-quality journals has a higher representativeness of emerging technologies. At the same time, it is also in line with people's cognition that the research conducted by high-quality institutions is generally groundbreaking research. Specifically, setting the quality parameter includes: determining the publishing journal and publishing organization corresponding to the candidate document where each keyword is located in the candidate keyword filtering set; obtaining the quality parameter based on the average journal influence of the publishing journal and the average research institution ranking of the publishing organization. Among them, the quality parameter of keyword i can be expressed using α i The quality parameter α of keyword i is i satisfy:

[0078] α i =2*average Journal Impact of the publishing journals*average Research Institution Ranking of the publishing institutions

[0079] The average ranking of the research institutions of the publishing institution may be, for example, the average ranking of the institutions by QS (Quacquarelli Symonds, World University Rankings).

[0080] In step S103, a quality parameter is set, and based on the quality parameter, the first change trend, the second change trend, and the third change trend, an emerging score value (Score i ),Right now

[0081] Score i =α i ×(ActiveTrend i +RecentTrend i +Slope i ).

[0082] In some other embodiments, considering that more attention should be paid to the trend score during the active period, the emerging score value (Score i ) can also satisfy:

[0083] Score i =α i ×ActiveTrend i +RecentTrend i +Slope i .

[0084] In step S103, after the emerging score value of each keyword is calculated, the candidate keyword filter set can also be arranged in descending order based on the emerging score value.

[0085] Step S104 , screening the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set.

[0086] In step S104, after observing the emerging score value of each keyword, it is found that the emerging score value of each keyword presents an approximately normal distribution and meets the triple screening in step S102. There are still many terms that meet the requirements, so further screening is required. The screening methods include threshold screening and ranking screening.

[0087] In some embodiments, for threshold screening, considering the score gap in different research fields, a corresponding emerging score threshold is set, and keywords with emerging score values ​​less than the set emerging score threshold are removed. The remaining keyword set is the candidate emerging technology keyword data set. For example, the emerging score threshold in the field of computer science can be set to 10, which means that the emerging score threshold meets the Score i Keywords with values ​​≥10 are retained to obtain the candidate emerging technology keyword dataset.

[0088] In other embodiments, for ranking screening, considering the differences in different research fields, the corresponding emerging score thresholds also vary greatly, and the candidate keyword filter set after descending order can be screened by using the arrangement order, for example, taking the top 20% of emerging score values ​​as the evaluation criteria, removing the keywords ranked after the top 20%, and taking the keywords ranked in the top 20% of emerging score values ​​as the candidate emerging technology keyword data set. In this case, the candidate emerging technology keyword data set can also be called a candidate emerging technology keyword table.

[0089] Step S105 , using a dynamic backtracking method to process the candidate emerging technology keyword data set to obtain a target emerging technology keyword data set.

[0090] In step S105, the candidate emerging technology keyword dataset is processed by a dynamic backtracking method to obtain a target emerging technology keyword dataset, including: setting corresponding year backtracking parameters and order down adjustment parameters based on different research fields; selecting year backtracking parameters and order down adjustment parameters corresponding to the research field to which the candidate document set belongs, and performing a descending operation on the candidate emerging technology keyword dataset by a dynamic backtracking method to obtain a target emerging technology keyword dataset. The year backtracking parameter is the research time range.

[0091] Among them, taking into account the actual differences in different fields (such as differences in the number of documents), the year backtracking parameter T and the order reduction parameter Z are set according to the actual situation of the field to perform a descending operation on the ranking of terms, thereby reducing the interference of classic technologies on emerging technologies.

[0092] In some embodiments, the candidate emerging technology keyword table is used as an object, X represents the current year, T represents the year backtracking parameter (i.e., the research time range, XT represents the backtracking year, l X represents the candidate emerging technology keyword list for the current year, l X-T represents the candidate emerging technology keyword table of the backtracking year, and Z represents the order reduction parameter. The specific implementation algorithm of the dynamic backtracking method is as follows:

[0093] Input: X, t, Z, current year candidate emerging technology keyword table l X Table 1 of candidate emerging technology keywords for retrospective years X-T ;

[0094] Output: The candidate emerging technology keyword table L after adjustment in the current year X ;

[0095] Step 1: Traverse l in descending order X The keyword with preset ranking (such as TOP10) in the list is judged whether the keyword appears in the list. X-T If yes, go to step 2, otherwise go to step 3;

[0096] Step 2: Lower the ranking of the keyword in the existing vocabulary by Z places and return the new l X , execute step 1 again to make a judgment;

[0097] Step 3: Keep the keyword in l X , execute step 1 again until l X All emerging technology keywords in l do not exist X-T In the example, let L X = l X , get the candidate emerging technology keyword table L after adjustment for the current year X .

[0098] In step S105, classic terms (i.e., traditional technical keywords) are removed through dynamic backtracking method to obtain a keyword list after the classic terms are removed, which solves the problem of traditional technical keywords dominating the list (i.e., the problem of traditional technical keywords being ranked before the test in the keyword list), and lays the foundation for further screening and analysis.

[0099] In some embodiments, considering that there are a lot of semantic duplication and redundancy problems in current terms, these problems will reduce the user experience, and in severe cases will make the acquired target emerging technology keyword data set meaningless. Therefore, the emerging technology identification method based on large-scale corpus also includes: optimizing the target emerging technology keyword data set using the maximum boundary correlation algorithm. The optimization process specifically refers to: first, the candidate terms are vector-embedded using the SimCSE (Simple Contrastive Learning of Sentence Embeddings, sentence embedding simple contrast learning algorithm) pre-training model to obtain a vectorized representation, and then the MMR algorithm (Maximal Marginal Relevance, maximum boundary correlation algorithm) is used to filter the candidate emerging terms (i.e., target emerging technology keywords) in the target emerging technology keyword data set to improve the diversity of the target emerging technology keyword data set and improve the user experience. Therefore, the innovative use of SimCSE word embedding and maximum boundary correlation algorithm to deduplicate candidate terms solves the problem of semantic clustering of candidate terms (i.e., semantic deduplication problem), further refines emerging terms, and improves the diversity of emerging technology lists.

[0100] Specifically, the maximum margin correlation algorithm is used to calculate the similarity between the query text and the search document, and then sort the search document. The algorithm formula is:

[0101]

[0102] In the formula, Q represents the query text, C represents the search document set, K is an initial set based on relevance, C\K represents the total set of the search document set C and the initial set K, C i represents an element in the total set (i.e., a sentence), C j represents an element in the initial set K, λsim(C i ,Q) refers to the similarity between a sentence in the total set and the query text, It refers to the similarity between a sentence in the total set and a sentence in the initial set K (i.e., the extracted summary sentence). λ is a hyperparameter that controls diversity.

[0103] If the maximum boundary correlation algorithm is applied to the term diversity expression of emerging technologies, that is, the search document set C corresponds to the target emerging technology keyword data set obtained by the dynamic backtracking method, and the initial set K corresponds to the extracted emerging term set, then sentence C i Corresponding to term i, sentence C j Corresponding to term j, we can set λ=0 and simplify the formula to:

[0104]

[0105] Considering that the text embedding vector expression of Bert (Bidirectional Encoder Representations from Transformers) is inaccurate, SimCSE from Princeton University is used to embed all terms before the MMR algorithm is performed, and the last layer of pooler_output is taken as its vector expression. It refers to the similarity between term i and term j. Here, cosine similarity is used to indicate the similarity between term i and term j.

[0106] In some embodiments, the process of the MMR algorithm is as follows:

[0107] 1) Embed all candidate terms (including the target emerging technology keyword dataset obtained by the dynamic backtracking method and all terms in the extracted emerging term set), take the last layer of pooler_output as its vector expression, and save the result as [term, corresponding vector], as input, and send it to the MMR algorithm;

[0108] 2) Initially, C represents the target emerging technology keyword dataset obtained by the dynamic backtracking method, K represents the extracted term set, and the term set K is empty at the beginning;

[0109] 3) Set the similarity score of the first term to 1, send it to the term set K, and save it in the list form of [term 1, 1];

[0110] 4) Do the following for the second to last term;

[0111] 5) Perform MMR calculation on term i: perform cosine similarity calculation on term i and the terms in term set K, take the highest similarity as its similarity score, and save it in the form of [term i, similarity score].

[0112] 6) After all candidate terms are calculated, a threshold is set to filter out those with similarity scores below the threshold, thereby obtaining a list of emerging technology terms with the greatest diversity.

[0113] In some embodiments, considering the technical and commercial potential contained in patent data, the method disclosed in the present invention can be regarded as a rapid evaluation method for emerging technologies that integrates multi-source scientific and technological data and artificial intelligence algorithms. The method uses a recognition model to rapidly evaluate emerging technologies. Given a time range, the input of the recognition model is multi-source heterogeneous data in a specified field, including paper documents and patent data, and the output is the emerging technology terminology in the field. In addition, considering the different numbers of documents in different fields, the evaluation threshold in the recognition model changes with the dynamic changes in the number of documents in the field, thereby enhancing the adaptability of the recognition model.

[0114] The emerging technology identification method based on large-scale corpus proposed in the disclosed embodiment determines the research field to construct a candidate document set, and extracts keywords from the candidate document set to obtain a candidate keyword data set; filters the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set; calculates the emerging score value of each keyword in the candidate keyword filter set; filters the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set; and processes the candidate emerging technology keyword data set using a dynamic backtracking method to obtain a target emerging technology keyword data set. In this case, the keywords are screened using the number of candidate documents in the candidate document set, the relevant information of the keywords, the emerging score value of each keyword, and the set emerging score threshold, and the screened keywords are processed using a dynamic backtracking method, which reduces the problem of dominance of traditional technology keywords in the keyword set, thereby improving the accuracy of emerging technology identification. The method disclosed in the present invention combines natural language processing technology with science and technology evaluation theory, and takes science and technology big data as the basis. First, an emerging technology evaluation data set is constructed, and terminology is extracted from paper documents and patent data through AutoPhrase and TF-IDF; then, the terms are filtered through the "three-dimensional" standards (indicators) of persistence, growth, and attention, and then the emerging scores of the filtered candidate terms are quantitatively calculated, and the candidate emerging technology keyword data set is obtained by sorting; finally, the dynamic backtracking method is used to lower the ranking of classic technologies to solve the problem of traditional technologies dominating the emerging technologies, and the SimCSE word embedding and maximum boundary correlation algorithm are used to perform semantic deduplication and diversification processing on the emerging technology list, so as to further optimize the emerging technology list and improve the diversity of the emerging technology list.

[0115] In order to verify the above-mentioned emerging technology identification method based on large-scale corpus, an experiment was conducted using the computer science field as an example. The specific process is as follows:

[0116] (1) Construct candidate documents for articles in the field of computer science in WOS (Web of Science) and calculate emerging terms. The data distribution is shown in Table 1.

[0117] Table 1 Number of candidate documents in each year from 2011 to 2020

[0118] years field Data volume 2011 Computer Science 126698 2012 Computer Science 132868 2013 Computer Science 142691 2014 Computer Science 168212 2015 Computer Science 191357 2016 Computer Science 195932 2017 Computer Science 204326 2018 Computer Science 209079 2019 Computer Science 221203 2020 Computer Science 167651

[0119] The efficiency of this method is verified based on such a large-scale data in Table 1. From the results, it only takes 0.0001 seconds to obtain the required set of emerging technology terms.

[0120] (2) The candidate emerging technology keyword dataset obtained based on the candidate documents in Table 1 is shown in Table 2. Among them, the terms in the candidate emerging technology keyword dataset in Table 2 are English terms.

[0121] Table 2 Candidate emerging technology keyword dataset

[0122]

[0123] Judging from the recent wave of deep learning and artificial intelligence, the method disclosed in this disclosure can well summarize the emerging technologies in recent years. In addition, combined with the technology maturity curve released by Gartner, the terms IoT (Internet of Things) and Blockchain in this disclosure are consistent with Gartner's report. In addition, since the release of Bert in 2018, large-scale pre-trained models have also become popular. The terms Training, Data models, Deep learning, Machine learning, Optimization, and Mathematical model in the above figure all conform to the characteristics of large-scale pre-trained models. Therefore, the experiment shows that the method disclosed in this disclosure can well summarize the emerging technologies in a field. In addition, the method disclosed in this disclosure can also further verify the reliability of the emerging terms obtained through reports on WeChat public accounts.

[0124] This disclosure evaluates emerging technologies in the literature in the field of computer science in 2020. At the same time, the keywords in the 500 most cited documents since 2020 are selected for word frequency statistical analysis, as shown in the keyword comparison table in Table 3. Based on Table 3, it can be found that the two have a high degree of overlap. Highly cited papers prove that the paper is a high-quality paper, and its keywords are also current emerging hot technologies, which indirectly proves the effectiveness of the emerging technology evaluation model proposed in this invention.

[0125] Table 3 Keyword comparison table

[0126]

[0127]

[0128] In summary, the present disclosure verifies the algorithm based on relevant devices and proves that it has achieved rapid and efficient evaluation of emerging technologies in, for example, 80 million papers and 50 million patent data. The method supports the diversity of technical semantics and supports dynamic adjustment of evaluation thresholds. It has the characteristics of domain adaptation and can efficiently process and identify (evaluate and analyze) papers and patents within a specified time range. At the same time, it can solve the classic technical keyword list in the field. At the same time, it also solves the diversity and redundancy of candidate technologies and improves the user experience.

[0129] In order to implement the above embodiments, the present disclosure also proposes an emerging technology identification device based on large-scale corpus. Figure 3 The present invention provides a schematic diagram of a device for identifying emerging technologies based on large-scale corpus. Figure 3 As shown, the emerging technology identification device 10 based on large-scale corpus includes an acquisition module 11, a filtering module 12, a calculation module 13, a screening module 14 and a processing module 15, wherein:

[0130] An acquisition module 11 is used to determine a research field, construct a candidate document set in the research field, and extract keywords from the candidate document set to obtain a candidate keyword data set;

[0131] A filtering module 12 is used to filter the candidate keyword data set according to the number of candidate documents in the candidate document set and the related information of the keywords to obtain a candidate keyword filtered set;

[0132] A calculation module 13, used to calculate the emerging score value of each keyword in the candidate keyword filtering set;

[0133] A screening module 14 is used to screen the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set;

[0134] The processing module 15 is used to process the candidate emerging technology keyword data set by using a dynamic backtracking method to obtain a target emerging technology keyword data set.

[0135] The emerging technology identification device based on large-scale corpus proposed in the embodiment of the present disclosure comprises an acquisition module that determines the research field to construct a candidate document set, and extracts keywords from the candidate document set to obtain a candidate keyword data set; a filtering module filters the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keywords to obtain a candidate keyword filter set; a calculation module calculates the emerging score value of each keyword in the candidate keyword filter set; a screening module screens the candidate keyword filter set based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set; a processing module processes the candidate emerging technology keyword data set using a dynamic backtracking method to obtain a target emerging technology keyword data set. In this case, the keywords are screened using the number of candidate documents in the candidate document set, the relevant information of the keywords, the emerging score value of each keyword, and the set emerging score threshold, and the screened keywords are processed using a dynamic backtracking method, which reduces the problem of dominance of traditional technology keywords in the keyword set, thereby improving the accuracy of emerging technology identification.

[0136] It should be noted that the aforementioned explanation of the embodiment of the method for identifying emerging technologies based on large-scale corpus is also applicable to the device for identifying emerging technologies based on large-scale corpus of this embodiment, and will not be repeated here.

[0137] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0138] Figure 4 A schematic diagram of the structure of an electronic device provided for an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable electronic devices, and other similar computing devices. The components, connections and relationships of the components, and functions of the components shown in the present disclosure are merely examples and are not intended to limit the implementation of the present disclosure described and / or required in the present disclosure.

[0139] like Figure 4As shown, the electronic device 20 includes a computing unit 21, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 22 or a computer program loaded from a storage unit 28 into a random access memory (RAM) 23. In the RAM 23, various programs and data required for the operation of the electronic device 20 can also be stored. The computing unit 21, the ROM 22, and the RAM 23 are connected to each other through a bus 24. An input / output (I / O) interface 25 is also connected to the bus 24. Multiple components in the electronic device 20 are connected to the I / O interface 25, including: an input unit 26, such as a keyboard, a mouse, etc.; an output unit 27, such as various types of displays, speakers, etc.; a storage unit 28, such as a disk, an optical disk, etc., which is communicatively connected to the computing unit 21; and a communication unit 29, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 29 allows the electronic device 20 to exchange information / data with other electronic devices through a computer network such as the Internet and / or various telecommunication networks.

[0140] The computing unit 21 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 21 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 21 performs the various methods and processes described above, such as executing an emerging technology identification method based on a large-scale corpus. For example, in some embodiments, the execution of an emerging technology identification method based on a large-scale corpus may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 28. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 20 via the ROM 22 and / or the communication unit 29. When the computer program is loaded into the RAM 23 and executed by the computing unit 21, one or more steps of executing the emerging technology identification method based on a large-scale corpus described above may be performed. Alternatively, in other embodiments, the computing unit 21 may be configured in any other appropriate manner (for example, by means of firmware) to execute the emerging technology identification method based on large-scale corpus.

[0141] Various embodiments of the systems and techniques described above in the present disclosure may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0142] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0143] In the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or electronic device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or electronic device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage electronic device, a magnetic storage electronic device, or any suitable combination of the foregoing.

[0144] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0145] The systems and techniques described herein may be implemented in a computing system including a backend component (e.g., as a data server), or a computing system including a middleware component (e.g., an application server), or a computing system including a frontend component (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network. A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in a cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server can also be a server of a distributed system, or a server combined with a blockchain.

[0146] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this disclosure is not limited here.

[0147] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for identifying emerging technologies based on large-scale corpus, characterized in that: include: Determine a research field, construct a candidate document set in the research field, and perform keyword extraction on the candidate document set to obtain a candidate keyword data set; Filtering the candidate keyword data set according to the number of candidate documents in the candidate document set and relevant information of the keyword to obtain a candidate keyword filtered set; Calculating the emerging score value of each keyword in the candidate keyword filtering set; The candidate keyword filter set is screened based on the emerging score value of each keyword and the set emerging score threshold to obtain a candidate emerging technology keyword data set; Processing the candidate emerging technology keyword dataset using a dynamic backtracking method to obtain a target emerging technology keyword dataset; Among them, the filtering of the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keyword to obtain the candidate keyword filter set includes: determining the research time range, and determining the basic period, recent period and active period based on the research time range; calculating the first occurrence number of each keyword in the candidate keyword data set in the active period; calculating the second occurrence number of each keyword in the candidate keyword data set in the basic period, and determining the first evaluation ratio of each keyword based on the first occurrence number and the second occurrence number of each keyword; calculating the number of institutions that use each keyword in the candidate keyword data set; filtering the candidate keyword data set based on the first occurrence number, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain the candidate keyword filter set; The method of filtering the candidate keyword data set based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set includes: determining a first active threshold and a second active threshold based on the number of candidate documents, and calculating a third number of occurrences of all keywords during the active period and a fourth number of occurrences of all keywords during the base period, obtaining a second evaluation ratio based on the third number of occurrences and the fourth number of occurrences, and obtaining an evaluation threshold based on the second evaluation ratio and the second active threshold; filtering out keywords whose first number of occurrences is less than the first active threshold, keywords whose first evaluation ratio is less than the evaluation threshold, and keywords whose number of institutions is less than a preset number threshold, thereby obtaining a candidate keyword filter set; The dynamic backtracking method includes: Step 1: Traverse in descending order l X The preset ranking keyword in the list is used to determine whether the keyword appears in l X-T If yes, go to step 2, otherwise go to step 3; Step 2: Lower the ranking of the keyword in the existing vocabulary by Z places and return to the new l X , execute step 1 again to make a judgment; Step 3: Keep the keyword in l X , execute step 1 again until l X All emerging technology keywords do not exist in l X-T In, order L X = l X , get the candidate emerging technology keyword dataset adjusted for the current year L X ,in l X It is the candidate emerging technology keyword dataset for the current year. l X-T This is a dataset of candidate emerging technology keywords with retrospective years. Z represents the order reduction parameter, X represents the current year, and T represents the year retrospective parameter.

2. The emerging technology identification method based on large-scale corpus according to claim 1 is characterized in that: The calculating of the emerging score value of each keyword in the candidate keyword filtering set includes: Calculate the first change trend of each keyword in the candidate keyword filter set in the recent period; Calculate the second change trend of each keyword in the candidate keyword filter set during the active period; Calculate the third change trend of each keyword in the candidate keyword filter set within the research time range; A quality parameter is set, and based on the quality parameter, the first change trend, the second change trend, and the third change trend, an emerging score value of each keyword is obtained.

3. The emerging technology identification method based on large-scale corpus according to claim 2 is characterized in that: The setting of quality parameters includes: Determine the publishing journals and publishing institutions corresponding to the candidate documents where each keyword is located in the candidate keyword filtering set; The quality parameter is obtained based on the average journal impact of the publishing journal and the average research institution ranking of the publishing institution.

4. The emerging technology identification method based on large-scale corpus according to claim 3 is characterized in that: The method of processing the candidate emerging technology keyword dataset using the dynamic backtracking method to obtain the target emerging technology keyword dataset includes: Set corresponding year-backward parameters and order-downward adjustment parameters based on different research fields; The year backtracking parameter and order down adjustment parameter corresponding to the research field to which the candidate document set belongs are selected, and the candidate emerging technology keyword data set is subjected to a descending operation using a dynamic backtracking method to obtain a target emerging technology keyword data set.

5. The emerging technology identification method based on large-scale corpus according to claim 1 or 4, characterized in that: Also includes: The target emerging technology keyword data set is optimized using a maximum margin correlation algorithm.

6. The emerging technology identification method based on large-scale corpus according to claim 5 is characterized in that: The step of extracting keywords from the candidate document set to obtain a candidate keyword data set includes: An automatic phrase mining algorithm and a word frequency-inverse document frequency algorithm are used to extract keywords from the candidate document set to obtain a candidate keyword data set.

7. An emerging technology identification device based on large-scale corpus, characterized in that: include: An acquisition module is used to determine a research field, construct a candidate document set in the research field, and perform keyword extraction on the candidate document set to obtain a candidate keyword data set; A filtering module, configured to filter the candidate keyword data set according to the number of candidate documents in the candidate document set and relevant information of the keyword to obtain a candidate keyword filtered set; A calculation module, used to calculate the emerging score value of each keyword in the candidate keyword filter set; A screening module, configured to screen the candidate keyword filter set based on the emerging score value of each keyword and a set emerging score threshold to obtain a candidate emerging technology keyword data set; A processing module, used to process the candidate emerging technology keyword data set by using a dynamic backtracking method to obtain a target emerging technology keyword data set; Among them, the filtering of the candidate keyword data set according to the number of candidate documents in the candidate document set and the relevant information of the keyword to obtain the candidate keyword filter set includes: determining the research time range, and determining the basic period, recent period and active period based on the research time range; calculating the first occurrence number of each keyword in the candidate keyword data set in the active period; calculating the second occurrence number of each keyword in the candidate keyword data set in the basic period, and determining the first evaluation ratio of each keyword based on the first occurrence number and the second occurrence number of each keyword; calculating the number of institutions that use each keyword in the candidate keyword data set; filtering the candidate keyword data set based on the first occurrence number, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain the candidate keyword filter set; The method of filtering the candidate keyword data set based on the first number of occurrences, the first evaluation ratio, the number of institutions and the number of candidate documents to obtain a candidate keyword filter set includes: determining a first active threshold and a second active threshold based on the number of candidate documents, and calculating a third number of occurrences of all keywords during the active period and a fourth number of occurrences of all keywords during the base period, obtaining a second evaluation ratio based on the third number of occurrences and the fourth number of occurrences, and obtaining an evaluation threshold based on the second evaluation ratio and the second active threshold; filtering out keywords whose first number of occurrences is less than the first active threshold, keywords whose first evaluation ratio is less than the evaluation threshold, and keywords whose number of institutions is less than a preset number threshold, thereby obtaining a candidate keyword filter set; The dynamic backtracking method includes: Step 1: Traverse in descending order l X The preset ranking keyword in the list is used to determine whether the keyword appears in l X-T If yes, go to step 2, otherwise go to step 3; Step 2: Lower the ranking of the keyword in the existing vocabulary by Z places and return to the new l X , execute step 1 again to make a judgment; Step 3: Keep the keyword in l X , execute step 1 again until l X All emerging technology keywords do not exist in l X-T In, order L X = l X , get the candidate emerging technology keyword dataset adjusted for the current year L X ,in l X It is the candidate emerging technology keyword dataset for the current year. l X-T This is a dataset of candidate emerging technology keywords with retrospective years. Z represents the order reduction parameter, X represents the current year, and T represents the year retrospective parameter.

8. An electronic device, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the emerging technology identification method based on large-scale corpus according to any one of claims 1 to 6.

Images