Emerging technologies can have major economic impacts and affect strategic
stability. Yet, early identification of emerging technologies remains
challenging. In order to identify emerging technologies in a timely and
reliable manner, a comprehensive examination of relevant scientific and
technological (S&T) trends and their related references is required. This
examination is generally done by domain experts and requires significant
amounts of time and effort to gain insights. The use of domain experts to
identify emerging technologies from S&T trends may limit the capacity to
analyse large volumes of information and introduce subjectivity in the
assessments. Decision support systems are required to provide accurate and
reliable evidence-based indicators through constant and continuous monitoring
of the environment and help identify signals of emerging technologies that
could alter security and economic prosperity. For example, the research field
of hypersonics has recently witnessed several advancements having profound
technological, commercial, and national security implications. In this work, we
present a multi-layer quantitative approach able to identify future signs from
scientific publications on hypersonics by leveraging deep learning and weak
signal analysis. The proposed framework can help strategic planners and domain
experts better identify and monitor emerging technology trends.
Detecting Emerging Technologies and their Evolution using Deep Learning and Weak Signal Analysis Ashkan Ebadi1,*, Alain Auger2, and Yvan Gauthier3 1. National Research Council Canada, Montreal, QC H3T 1J4, Canada 2. Science and Technology Foresight and Risk Assessment Unit, Defence Research and Development Canada, Ottawa, Ontario, Canada 3. National Research Council Canada, Ottawa, ON K1K 2E1, Canada * Corresponding author: Ashkan Ebadi (ashkan.ebadi@nrc-cn rc.gc.ca) Abstract Emerging technologies can have major economic impacts and affect strategic stability.
Detecting Emerging Technologies and their Evolution using Deep Learning and Weak Signal Analysis Ashkan Ebadi1,*, Alain Auger2, and Yvan Gauthier3 1. National Research Council Canada, Montreal, QC H3T 1J4, Canada 2. Science and Technology Foresight and Risk Assessment Unit, Defence Research and Development Canada, Ottawa, Ontario, Canada 3. National Research Council Canada, Ottawa, ON K1K 2E1, Canada * Corresponding author: Ashkan Ebadi (ashkan.ebadi@nrc-cn rc.gc.ca) Abstract Emerging technologies can have major economic impacts and affect strategic stability. 訳抜け防止モード: 深層学習と弱信号解析ashkan ebadi1を用いた新興技術とその進化 ※alain auger2及びyvan gauthier3 1. national research council canada モントリオール, qc h3 t 1j4, canada 2. science and technology foresight カナダ国防研究開発部 リスクアセスメント部門 オタワ カナダ オンタリオ カナダ カナダ カナダ オタワ k2e1 カナダ * 対応する著者 : ashkan ebadi (ashkan.ebadi@nrc-cn rc.gc.ca) 抽象的な新興技術は経済的に大きな影響を与え、戦略的安定性に影響を与える可能性がある。
0.84
Yet, early identification of emerging technologies remains challenging.
しかし、新興技術の早期発見は依然として困難である。
0.56
In order to identify emerging technologies in a timely and reliable manner, a comprehensive examination of relevant scientific and technological (S&T) trends and their related references is required.
This examination is generally done by domain experts and requires significant amounts of time and effort to gain insights.
この試験は一般的にドメインの専門家によって行われ、洞察を得るためにかなりの時間と労力を要する。
0.62
The use of domain experts to identify emerging technologies from S&T trends may limit the capacity to analyse large volumes of information and introduce subjectivity in the assessments.
Decision support systems are required to provide accurate and reliable evidence-based indicators through constant and continuous monitoring of the environment and help identify signals of emerging technologies that could alter security and economic prosperity.
For example, the research field of hypersonics has recently witnessed several advancements having profound technological, commercial, and national security implications.
In this work, we present a multi-layer quantitative approach able to identify future signs from scientific publications on hypersonics by leveraging deep learning and weak signal analysis.
Keywords emerging terms, future sign, weak signal, natural language processing, deep learning, hypersonics
キーワードの創発語、未来の記号、弱い信号、自然言語処理、ディープラーニング、ハイパーソニックス
0.73
1. Introduction The complex and rapidly evolving nature of modern science has given rise to emerging technologies with transformative and disruptive characteristics, as recently exemplified by deep learning [1], cryptocurrencies [2], mRNA vaccines [3], and solar cells [4].
Such technologies have changed the landscape of existing industries while creating new economic opportunities and affecting societies and the lives of people [5].
Given the highly competitive and evolving environment, planners always seek insights into emerging technologies that could affect long-term strategic stability [6], economic development, and national security.
Early detection of new technology trends is of critical importance for governments and businesses, as it enables them to identify opportunities and risks quickly and react to them accordingly by formulating appropriate research, development, and innovation strategies [7].
Several Scientometrics studies have considered various data sources, such as scientific publications and patents, and employed different techniques, such as bibliometrics and keyword network analysis, to identify the emergence and evolution of new technologies [8, 9, 10].
Using co-word analysis and by means of scientometric indicators, Lee [11] identified emerging research themes in the field of information security by analyzing patterns and trends.
In another study, Kim and Zhu [13] applied dynamic topic modeling to identify and investigate the thematic patterns and emerging trends of the scientific publications in scientometrics.
More recently, Park and Yoon [14] presented a quantitative approach to discover potential future technological opportunities from the patent-citation network.
On the other hand, the fast-evolving nature of modern science and technology has made strategic planning more complex [16].
一方で、近代科学技術の急速な進化によって、戦略的計画がより複雑になった [16]。
0.80
The incomplete and asynchronous information, in some cases, adds to this complexity [17].
不完全で非同期な情報は、場合によってはこの複雑さを増す[17]。
0.75
These attributes along with the availability of advanced computer science algorithms provide new opportunities to gain better insights from S&T databases.
For example, in a recent study, Xu et al [18] combined several machine learning models and proposed a framework to identify and foresight emerging research topics at the thematic level.
例えば、最近の研究では、Xu et al [18]はいくつかの機械学習モデルを組み合わせて、テーマレベルで新しい研究トピックを特定し、予測するためのフレームワークを提案しました。
0.63
Computerized systems can help human experts keep pace with increasing and evolving data and extract knowledge about a specific technology domain [19].
For instance, the research field of hypersonics is rapidly progressing and has various commercial and military applications [20].
例えば、超音速の研究分野は急速に進歩しており、様々な商業・軍事用途がある[20]。
0.71
The term hypersonic means “pertaining to or moving at a speed greatly in excess of the speed of sound, usually meaning greater than Mach 5. All speeds in excess of the speed of sound are supersonic, but to be hypersonic requires even higher speed” [21].
Research in hypersonics is now central to many civilian and military aerospace programs.
ハイパーソニックスの研究は現在、多くの民間および軍事航空計画の中心となっている。
0.56
In this research, we propose a multi-layer approach to extract signals and trace their evolution, combining various techniques such as natural language processing (NLP), deep learning, and weak signal analysis.
Our contribution is threefold: first, we improve upon keyword extraction by going beyond traditional statistical approaches and leverage a pre-trained transformer-based deep learning technique, namely BERT [22], and get domain experts to validate the extracted keywords.
As a test case to measure the ability of the approach to identify trends and extract early signals, we focus on scientific publications in the domain of hypersonics for the period from 1985 to 2020.
We present our findings in Section 4 and discuss them in Section 5.
本報告は第4節で報告し,第5節で考察する。
0.53
Section 6 covers the limitations of this work and draws some directions for future research.
第6節では、この研究の限界をカバーし、今後の研究の方向性を定めている。
0.46
2. Related Work Our proposed approach is based on automatic keyword extraction using state-of-the-art techniques and weak signal analysis.
2. 関連作業 提案手法は,最先端技術と弱信号解析を用いた自動キーワード抽出に基づく。
0.91
In this section, these topics are briefly introduced and previous work is discussed.
本節では、これらのトピックを簡潔に紹介し、先行研究について論じる。
0.52
2.1. Automatic Keyword Extraction Automatic keyword extraction (AKE) is the process of automatically extracting representative keywords (and/or phrases) from a document (or a collection of documents) [23].
This is crucial especially if the user deals with large data sets, which is usually the case in real-world
特にユーザが大規模なデータセットを扱う場合、これは非常に重要です。
0.67
2
2
0.42
英語(論文から抽出)
日本語訳
スコア
applications. AKE has been widely studied.
アプリケーション。 明は広く研究されている。
0.68
At a high-level, keyword extraction approaches can be classified into four categories: 1) statistical approaches, 2) linguistics approaches, 3) machine/deep learning approaches, and
We briefly introduce these approaches in the following sections.
以下にこれらのアプローチを簡単に紹介する。
0.64
2.1.1. Statistical Approaches Statistical methods for keyword extraction are mostly based on simple statistics calculated from nonlinguistic features of a (set of) document(s) [25].
Several papers have reported results for the extraction of a set of keywords from a corpus using simple statistical measures such as term frequency (TF) [26], word co-occurrences [27], and term frequency-inverse document frequency (TF-IDF) [28].
Since these approaches are based on the frequency of terms/occurrences, their results could be noisy and not very precise [25].
これらの手法は項/事象の頻度に基づいており、その結果はノイズがあり、非常に正確ではない[25]。
0.68
2.1.2. Linguistic Approaches Linguistic methods use natural language processing techniques and linguistic features of words in a (set of) document(s) to detect and extract representative keywords [25].
For example, [29] used lexical analysis to summarize documents, [30] employed syntactic analysis to improve automatic keyword extraction, and [31] performed discourse analysis to automatically structure and summarize documents.
The learning phase can be done in a supervised setting where a model is trained using (large) annotated text corpora or in an unsupervised setting without using any annotated dataset.
There are several works in the literature that performed supervised learning approaches for keyword extraction using various machine learning techniques such as support vector machine (SVM) [32] and Naive Bayes [33].
Rapid Automatic Keyword Extraction (RAKE) [34] is an example of an unsupervised, domain- and language-independent method for automatic keyword extraction from documents.
YAKE! [35] is another example of an unsupervised automatic keyword extraction approach using statistical text features.
YAKE! [35]は統計テキストの特徴を用いた教師なしの自動キーワード抽出手法の一例である。
0.76
With recent advancements in language representation such as the Bidirectional Encoder Representations from Transformers (BERT) method [22], new techniques are introduced that use BERT embeddings for keyword and keyphrase retrieval [36, 37].
2.1.4. Hybrid Approaches Hybrid methods combine the above-mentioned approaches or use heuristics to calculate the best features from the target text corpora and use them to extract keywords and/or keyphrases [38, 39].
2.2. Weak Signal Analysis Detection of weak signals, trends, and issues in the evolving technology landscape as early as possible has been emphasized in the literature for strategic planning [40, 41].
Different types of changes could be of interest from an unexpected discontinuity in a technology trend to a new emerging (mega-)trend with a high impact on society and the technology landscape itself [42].
4) can be regarded as a threat or opportunity to a specific group, 5) can be downgraded by those who perceive and know this signal, 6) needs time to become mature and mainstream, and
weak signals are “early signs of possible but not confirmed changes that may later become more significant indicators of critical forces”, hence, they can be regarded as signals about future trends of a given technology.
In other words, these signals are potential indicators of the discontinuity or emergence of a target technology even with no current significant impact [45].
More recently in 2021, van Veer and Ortt [46] reviewed 68 definitions of weak signals and proposed a common unified definition that crosses foresight domains: “A perception of strategic phenomena detected in the environment or created during interpretation that are distant to the perceiver’s frame of reference”.
Hiltunen [47] suggested three dimensions for future signs:
Hiltunen [47] は将来の兆候の3次元を示唆している。
0.57
1) signal, i.e., an indicator of visibility,
1)信号、すなわち可視性の指標。
0.67
2) issue, i.e., an indicator of diffusion of a future sign, and
2)課題,すなわち,将来の記号の拡散の指標,及び
0.69
3) interpretation, i.e., the meaning of the future sign to the receiver.
3) 解釈,すなわち,受信者に対する将来の符号の意味。
0.70
Signs can be categorized into weak to strong signals in this three dimensional space.
符号はこの3次元空間において弱い信号から強い信号に分類できる。
0.82
Figure 1 shows Hiltunen’s three dimensions of future signs and the characteristics of weak and strong signals [47].
図1は、ヒルトゥネンの将来の記号の3次元と弱い信号と強い信号の特性を示しています [47]。
0.77
In this future sign space, a weak signal has low levels of signal, issue, and interpretation and can turn into a strong signal exhibiting higher levels of the mentioned dimensions.
Figure 1. Weak and strong signals in Hiltunen’s three dimensions of future signs.
図1に示す。 ヒルトゥネンの3次元の将来の記号における弱く強い信号。
0.60
Several qualitative studies have been conducted on the concept of the weak signals to analyze business environments, assist corporate decision-making, and strategy-creation process [48, 49], support strategic planning and foresight services in corporations [50], and utilize an image-based medium to trigger employees’ future thinking in analyzing organizations [51], among others.
Although Hiltunen’s qualitative framework could assist experts in identifying signals, it may suffer from subjectivity and may require lots of resources to be implemented [52].
Yoon [45] proposed a quantitative method for Hiltunen’s framework [47] based on text mining to mitigate analysts’ subjectivity and overcome the limitations of the qualitative approach.
Despite its limitations, term frequency is often considered as a measure of the importance of a term [53].
その制限にもかかわらず、用語の頻度は、しばしば[53]の重要性の尺度と見なされる。
0.80
Document frequency represents the number of documents in which a specific term has appeared [54] and is used as a measure of dissemination of a term in a collection of documents [53].
Yoon [45] also described a strong signal as a term with high term and document frequencies and a high growth rate.
ユーン[45]はまた、強い信号は、高い周期と文書の頻度と高い成長率を持つ用語であると記述した。
0.69
4
4
0.43
英語(論文から抽出)
日本語訳
スコア
3. Data and Methodology 3.1.
3.データと方法論 3.1
0.86
Data Researchers mostly convey their scientific discoveries and findings to the scientific community and the general public via scientific publications [55], hence it is considered the major output of scientific research [56].
Therefore, in this study, we focused on scientific publications as the primary data source to identify emerging technologies in hypersonics using machine learning and weak signal analysis.
Data collection and preparation involved several steps.
データ収集と準備にはいくつかのステップがあった。
0.52
First, using the string “hypersonic*” as the search query, we collected the bibliographic data of hypersonics-related publications that were published within the period from 1985 to 2020 from the Elsevier’s Scopus database.
A total of (n = 21, 669) articles was retrieved in August 2021.
2021年(2021年)8月に全(n = 21, 669)が回収された。
0.57
This included many meta-data about each of the retrieved articles such as date of publication, title, abstract, and list of authors and affiliations.
これには、出版日、題名、要約、著者と所属者のリストなど、検索された各記事に関する多くのメタデータが含まれていた。 訳抜け防止モード: これには、出版日など、回収された各記事に関する多くのメタデータが含まれていた。 title , abstract , and list of author and affiliations
0.76
We checked and ensured that for all of the collected articles either the title or abstract was available.
収集された記事のタイトルや要約がすべて利用可能であることを確認し、確認しました。
0.58
We then created a new field combining title and abstract and removed duplicated records (n = 79).
We experimented with various time intervals for dividing the examined period and found that a 3year window would produce the best results in terms of a representative sampling of the keywords across the entire dataset.
These approaches include but are not limited to information filtering [57], graph-based methods [58], centrality measures [59], and domain knowledge [60].
These techniques are mostly based on term frequencies in the corpus.
これらの技術は主にコーパスの項周波数に基づいている。
0.72
Over the recent years, the natural language processing (NLP) community has developed and used large language models (LLMs) such as Bidirectional Encoder Representations from Transformers (BERT) [22].
Word embeddings were extracted for n-grams and cosine similarity was used to find the most representative keywords (and phrases) of each document in the dataset.
It is assumed that highly frequent keywords with more occurrences in recent years are important indicators of emergence.
近年の出現頻度が高いキーワードは,出現の指標として重要であると考えられる。
0.81
Term frequencies were used to calculate keywords’ degree of visibility.
用語周波数はキーワードの可視性を計算するのに使われた。
0.67
Following Yoon [45], we used a time-weighted formula to calculate the degree of visibility (DoV) of the extracted keywords (n = 4,366) in each of the 12 aforementioned time intervals, i.e., t1 to t12:
where TFij is the total frequency of keyword i in time interval j, Nj is the total number of documents in time interval j, n is the number of time intervals (n = 12, in our experiment), and w is a constant time-weight.
tfij が時間区間 j におけるキーワード i の総頻度である場合、nj は時間区間 j における文書の総数、n は時間間隔の数(実験では n = 12)、w は一定の時間重み付けである。
0.77
We set w = 0.05 as optimal after experimenting with different values and carefully reviewing the results.
我々は、w = 0.05 を、異なる値を実験し、結果を慎重にレビューした後、最適な値に設定した。 訳抜け防止モード: w = 0.05 を最適とする。 異なる値で実験し、結果を慎重にレビューします。
0.68
The Keyword Emergence Map (KEM) can next be created from the calculated DoV values.
次に、計算されたDoV値からキーワードEmergence Map(KEM)を作成することができる。
0.67
In KEM, the xaxis represents the average term frequencies of the keywords, and the y-axis indicates the DoV’s growth rate, calculated as the geometric mean.
For this purpose, we first used the following time-weighted formula to calculate the degree of diffusion (DoD) of the extracted keywords in each of the 12 aforementioned time intervals, i.e., t1 to t12:
In equation (2), DFij is the document frequency of keyword i in time interval j, Nj is the total number of publications in time interval j, n is the number of time intervals (n = 12, in our experiment), and w is a constant time-weight, w = 0.05.
Next, the Keyword Issue Map (KIM) was generated by plotting the average time-weighted increasing rate of keywords’ document frequencies on the y-axis, and keywords’ average document frequencies on the x-axis.
According to Yoon [45], weak signals correspond to those terms with relatively low document frequency but a high growth rate.
yoon [45] によると、弱い信号は文書の頻度が比較的低いが成長率が高い用語に対応している。
0.73
Therefore, similar to the approach explained in the previous section, we divided the KIM into 4 quadrants based on the medians of the values on each axis, as follows:
3.2.5. Signals Extraction Figure 4 shows the process of extracting signals from KEMs and KIMs by analyzing their intersection.
3.2.5. 信号抽出図4は、ケムスとキムの交点を分析して信号を抽出する過程を示す。
0.49
Signals appearing in the same quadrant in both KEM and KIM indicated similar degrees of visibility and dissemination and were extracted.
KEMとKIMで同一の四角形に現れる信号は、類似の可視性と拡散の度合いを示し、抽出した。
0.66
We did this process for KEMs and KIMs in each of the examined periods, i.e., P1 to P3, and extracted the final weak, strong, and well-known but not strong signals.
However, “Term-C” is not listed as a weak signal because it does not appear as a weak signal in both maps.
しかし、"term-c" は両方の地図に弱い信号として表示されないため、弱い信号としてリストされていない。 訳抜け防止モード: しかし、“ Term - C ” は弱い信号としてリストされない。 どちらの地図にも 弱い信号とは見えません
0.79
The extracted signals were then verified and validated by domain experts.
抽出されたシグナルはドメインの専門家によって検証され検証された。
0.49
Figure 4. The process of extracting signals from KEM and KIM.
図4。 KEMとKIMから信号を抽出するプロセス。
0.70
4. Results 4.1. Extracted Terms Categories Following the methodology that was explained in Section 3.2, senior scientists with domain expertise in hypersonics carefully verified and validated the extracted signals (n = 442) and classified them into 5 highlevel categories.
Letters in brackets, listed under the Abbreviation column of Table 1, are used in the rest of this paper to refer to term categories.
表1の省略欄に記載された括弧の文字は、この論文の残りの部分でカテゴリを参照するために使用される。
0.73
As illustrated, the Materials and structures and the Guidance, navigation, and control categories contain the highest and the lowest number of terms, respectively.
acetylene, aegis, afterburn, ammonia, apertures, air turbo rocket,
アセチレン、エージス、アフターバーン、アンモニア、アパーチャ、エアターボロケット
0.66
autoignition, ballistic, bursts, capsule, ...
自動発火 弾道 バースト カプセル...
0.48
Figure 5 shows the coverage percentage of defined categories in each examined period.
図5は、各調査期間における定義されたカテゴリのカバレッジ比率を示しています。
0.60
As shown in the figure, the “Materials and structures” category has a higher coverage in all periods, except for the strong signals in P1 where it is dominated by the “Vehicles, propulsion, and fuels” category.
The proportion of categories calculated based on all the extracted terms is almost similar in P1, P2, and P3, however, more changes in terms of categories coverage is observed for strong signals.
Moreover, although the proportion of the “Modelling, simulation, and analysis” category is not high compared to other categories, its weak signals coverage has increased from P1 to P3.
Figure 5. Percentage of related terms per category in each period for all terms, weak signals, and strong signals.
図5。 すべての項、弱い信号、強い信号について、各期間におけるカテゴリ毎の関連項の割合。
0.73
4.2. Signals Evolution over Time Before analyzing signals’ temporal changes, we first investigate the existence of temporal evolution in the documents during the examined period.
For this purpose, using the extracted keywords for each examined time interval (n = 500 ∗ 12, as explained in Section 3.2.2), a bipartite graph G(V,E) of keywords and time intervals is created such that each time interval is linked to its representative keywords.
Figure 6-b shows the degree distribution of keyword nodes.
図6-bはキーワードノードの度数分布を示しています。
0.62
As shown in the figure, most of the nodes are of degree 1 indicating that they have only appeared in 1 time interval, therefore, the extracted keywords are highly specific.
Figure 7 depicts the temporal evolution of terms that are weak signals in either P1 or P2 and become strong signals in the subsequent period.
図7は、P1 または P2 の弱い信号であり、その後に強い信号となる用語の時間的進化を描いている。
0.74
Terms in black are the stems of the signals and the red terms in brackets are their corresponding terms.
黒の用語は信号の語幹であり、括弧の赤い語は対応する語である。
0.63
In the figure, “odw” stands for Oblique Detonation Wave, “pse” for Parabolized Stability Equations, and “swbli” for Shock-Wave/Boundary- Layer Interaction.
However, some terms are related to significant advancements in hypersonics over the last 30 years, e g , “aeroshell”, “magnetohydrodynamics ”, and “microstructures”.
Figure 7. Weak signals in P1 or P2 that converted into a strong signal at P2 or P3.
図7。 P1またはP2の弱信号はP2またはP3の強い信号に変換される。
0.74
Terms in black are stems of the signals and the red terms in brackets are their corresponding terms.
黒の用語は信号の語幹であり、括弧の赤い語は対応する語である。
0.66
Stems are sorted based on their category.
ステムはそのカテゴリに基づいてソートされる。
0.60
Out of 131 weak signals in the final period (P3), 58 of them have no appearance in the prior periods, suggesting they could be emerging weak signals of new research activities.
Figure 8 lists these terms along with their representative category.
図8は、これらの用語とその代表的カテゴリを列挙している。
0.56
The majority of the terms belong to the “Materials and structures” category, followed by “Aerothermodynamics” and “Modelling, simulation, and analysis”.
用語の大部分が"Materials and Structure"カテゴリに属し、続いて"Aerothermodynamics&q uot;と"Modelling, Simulation, and Analysis"が続く。
0.73
Figure 8. Weak signals in P3 that did not appear as a signal in P1 and P2.
図8。 P1とP2のシグナルとして現れなかったP3の弱信号。
0.77
12
12
0.42
英語(論文から抽出)
日本語訳
スコア
4.3. Terms with Specific Signal Patterns The proposed framework allows strategic planners to extract terms with specific signal patterns of interest.
This implicitly verifies the validity of the signals as sinusoidal patterns are not expected to be frequent in the evolving landscape of the case technology.
However, some of the extracted signals follow a constant signal strength over time.
しかし、抽出された信号のいくつかは時間とともに一定の信号強度に従う。
0.65
“Controllability”, “coolant”, “fibers”, and “decomposition” are strong signals in all the examined periods (n = 4), that may indicate enduring challenges mobilizing the research community.
5. Discussion and Conclusion Given that the research field of hypersonics encompasses a wide range of technologies, thoroughly assessing the performance of weak signal detection is a challenging task.
Although they are all relevant to the field of hypersonics, many of them do not represent a technology per se but are rather indicators of scientific and technical research areas.
Another set of terms seems more specific to the measurement and observation of physical phenomena during experimentation than the technology itself, such as “blast”, “burst”, “explos”, “shockwave”, “swbli” (for shock-wave-boundary- layer interaction), and “throughput”.
Although these terms are not directly representative of a technology, they help understand how the field has been developing (e g , engineers paying more attention to ice formation on wings) and still have value, especially when analyzed in the context of their category (e g , propulsion).
Furthermore, even for a broad field such as hypersonics, it is not too onerous for a subject-matter expert engaged in technology foresight to review the list and retain terms with signal patterns of interest.
In the present case, among the terms that remain, many represent actual technologies that had a major and demonstrable impact on the broader field of hypersonics1, including:
• Aeroshel - Aeroshells are the rigid heat-shielded shells that help decelerate and protect spacecraft from pressure, with shapes now computationally designed to achieve specified lift-to-drag ratios [65].
エアロシェル - エアロシェルは、宇宙船を圧力から減速し、保護する硬い熱シールドシェルで、現在は特定のリフト・アンド・ドラッグ比[65]を達成するように設計されている。 訳抜け防止モード: エアロシェル(Aerosheel)は、宇宙船の圧力を抑えるためのシェル。 形状を計算的に設計し to achieve certain lift―contained lift ---ドラッグ比 [65 ]。
0.61
• Magnetohydrodynam - Magnetohydrodynamic (MHD) bypass has become a common way to boost airbreather engine performance in hypersonic propulsion systems [63, 66].
The technology was originally used on Russian scramjet demonstrators.
この技術はもともとロシアのスクラムジェットのデモに使われた。
0.60
• Microstructur - New thermal protection structure designs rely heavily on the conduct of complex
・ミクロ構造体-錯体の導電性に大きく依存する新しい耐熱構造体
0.69
aero-thermo-elastic simulations and thermoelastic materials [63].
エアロ・サーモ弾性シミュレーションと熱弾性材料 [63]
0.75
Accordingly, we can conclude that for the particular field and dataset analyzed, the multi-layer approach presented in this paper has successfully identified weak signals associated with emerging technologies in the field of hypersonics, without presenting experts with an overwhelming number of technical terms to review and validate.
This approach could thus have important benefits for organizations engaged in technology roadmapping as a means to inform and prioritize research and development activities.
1 If we consider terms that went from weak to well-known, many also represent significant advances in the field, including scramjet engines that, although still under development, have been propelling many test aircraft already [63].
We also find endothermic reactions now leveraged to cool aircraft [63, 64], especially through “reformed” fuels such as steam-reforming hydrocarbon fuel that create endothermic reactions [63].
Limitations As illustrated above, the approach often returns terms that, although relevant, do not seem directly associated with technologies (e g , “hazard”, “kw”), and as such the signal-to-noise ratio (i.e., precision) of the output could still be improved.
Another limitation is that although the approach successfully identified early many technologies of importance in hypersonics, it also identified many of them very late.
Yet, some types of hypersonic platforms, such as boost-glide missiles, have been conceived over 80 years ago and in development since the early 2000s [67].
They are now becoming of concern due to claims of them being fielded in military operations [68].
軍事作戦に投入されたとの主張から、現在では懸念の対象となっている[68]。
0.67
This omission could be due to the use of a single source of data or to the fact that such platforms were linked to multiple stems (e g , “boost”, “glide”) of different strengths.
As such, this limitation might be more a limitation in the data or post-processing than a limitation of the approach itself.
このように、この制限は、アプローチ自体の制限よりも、データや後処理の制限であるかもしれない。
0.74
Nevertheless, it shows that the recall performance of the method could be improved to reduce the risk of missing important terms, but this would have to be done without compromising precision performance.
6.2. Future Work A follow-on project will validate the proposed multi-layer approach by applying it to a different scientific and technical research area and re-assess the performance of the approach in detecting weak signals of emergence.
7. Acknowledgements We would like to thank Inbal Marcovitch and Karla Cisneros Rosado from Defence Research and Development Canada (DRDC) for their useful comments and suggestions.
We also want to acknowledge the contribution of domain experts who reviewed and validated lists of signals.
また、信号のリストをレビューし検証したドメインの専門家の貢献を認めたいと思っています。
0.55
References [1] Y. LeCun, Y. Bengio, G. Hinton, Deep learning, nature 521 (7553) (2015) 436–444.
参照:[1] Y. LeCun, Y. Bengio, G. Hinton, Deep Learning, nature 521 (7553) (2015) 436–444。
0.94
[2] S. Nakamoto, Bitcoin: A peer-to-peer electronic cash system, Decentralized Business Review (2008)
[2] S. Nakamoto, Bitcoin: a peer-to-peer Electronic cash system, Decentralized Business Review (2008)
0.47
21260. [3] U. Sahin, K. Karik´o, ¨ O. T¨ureci, mRNA-based therapeutics—developing a new class of drugs, Nature
21260. U. Sahin, K. Karik ́o, . O. T 'ureci, mRNAベースの治療薬 : 新しい種類の薬物、Natureを開発 訳抜け防止モード: 21260. [3 ]U. Sahin, K. Karik ́o, 新しい種類の薬物を開発するmRNAベースの治療薬、Nature
0.61
reviews Drug discovery 13 (10) (2014) 759–780.
薬品発見13 (10) (2014) 759-780を参照。
0.80
[4] X. Li, Q. Xie, L. Huang, Identifying the development trends of emerging technologies using patent analysis and web news data mining: the case of perovskite solar cell technology, IEEE Transactions on Engineering Management (2019).
4] x. li, q. xie, l. huang, 特許分析とwebニュースデータマイニングによる新興技術の発展トレンドの特定: ペロブスカイト太陽電池技術の事例, ieee transactions on engineering management (2019)
0.73
[5] J. Rifkin, The third industrial revolution: how lateral power is transforming energy, the economy, and
[6] T. S. Sechser, N. Narang, C. Talmadge, Emerging technologies and strategic stability in peacetime, crisis,
[6]T. S. Sechser, N. Narang, C. Talmadge, Emerging Technology and Strategy stability in peacetime, crisis, 訳抜け防止モード: [6 ]T. S. Sechser, N. Narang, C. Talmadge, 平和期における新興技術と戦略的安定
0.87
and war, Journal of strategic studies 42 (6) (2019) 727–735.
そして、journal of strategic studies 42 (6) (2019) 727–735。
0.78
[7] X. Li, Y. Zhou, L. Xue, L. Huang, Integrating bibliometrics and roadmapping methods: A case of dyesensitized solar cell technology based industry in China, Technological Forecasting and Social Change 97 (2015) 205–222.
[7] X. Li, Y. Zhou, L. Xue, L. Huang, Integrating bibliometrics and roadmapping methods: a case of dyesensitized solar cell technology based industry in China, Technological Forecasting and Social Change 97 (2015) 205–222。 訳抜け防止モード: [7 ]X. Li, Y. Zhou, L. Xue, L. Huang, 書誌学とロードマップ作成の融合 中国における色素増感太陽電池産業の事例 Technological Forecasting and Social Change 97 (2015 ) 205–222。
0.77
14
14
0.42
英語(論文から抽出)
日本語訳
スコア
[8] T. U. Daim, G. Rueda, H. Martin, P. Gerdsri, Forecasting emerging technologies: Use of bibliometrics
T. U. Daim, G. Rueda, H. Martin, P. Gerdsri, Forecasting emerging technology: using bibliometrics 訳抜け防止モード: [8 ]T. U. Daim, G. Rueda, H. Martin P. Gerdsri, 新興技術の予測 書誌学の活用
0.80
and patent analysis, Technological forecasting and social change 73 (8) (2006) 981–1012.
そして特許分析、技術予測、社会変化 73 (8) (2006) 981–1012。
0.71
[9] F. Dotsika, A. Watkins, Identifying potentially disruptive trends by means of keyword network analysis,
9] f. dotsika, a. watkins, キーワードネットワーク分析による破壊的傾向の同定 訳抜け防止モード: [9]F.Dotsika,A.Watkins, キーワードネットワーク分析による潜在的破壊的傾向の同定
0.81
Technological forecasting and social change 119 (2017) 114–127.
技術予測と社会変化 119 (2017) 114-127。
0.83
[10] H. Noh, Y. K. Song, S. Lee, Identifying emerging core technologies for the future: Case study of patents published by leading telecommunication organizations, Telecommunications Policy 40 ((10–11)) (2016) 956–970.
[10]H. Noh, Y. K. Song, S. Lee, Identifying emerging core technology for the future: Case study of patents published by leading telecommunication organization, Telecommunications Policy 40 ((10–11)) (2016) 956–970。 訳抜け防止モード: 〔10〕h.能、y.k.歌、s.リー 次世代のコア技術を見極める : 主要な通信機関である電気通信政策40 (10–11 ) (2016 ) 956–970 による特許の事例研究
0.72
[11] W. Lee, How to identify emerging research fields using scientometrics: An example in the field of
11]W. Lee, サイエントメトリックスを用いた新しい研究分野の特定方法--分野の例
0.53
information security, Scientometrics 76 (3) (2008) 503–525.
情報セキュリティ、scientometrics 76 (3) (2008) 503–525。
0.84
[12] R. K. Abercrombie, A. W. Udoeyop, B. G. Schlicher, A study of scientometric methods to identify
12] r. k. abercrombie, a. w. udoeyop, b. g. schlicher, a study of scientometric methods to identify 訳抜け防止モード: [12 ] R. K. Abercrombie, A. W. Udoeyop, B. G. Schlicher 科学的手法の同定に関する研究
0.80
emerging technologies via modeling of milestones, Scientometrics 91 (2) (2012) 327–342.
[13] M. C. Kim, Y. Zhu, Scientometrics of scientometrics: mapping historical footprint and emerging
13]M.C. Kim, Y. Zhu, Scientometrics of scientometrics: mapping historical footprint and emerging
0.45
technologies in scientometrics, in: Scientometrics, 2018, pp. 9–27.
Scientometrics, 2018, pp. 9-27。
0.44
[14] I. Park, B. Yoon, Technological opportunity discovery for technological convergence based on the prediction of technology knowledge flow in a citation network, Journal of Informetrics 12 (4) (2018) 1199–1222.
I. Park, B. Yoon, Technological opportunity discovery for technology convergence based on the prediction of technology knowledge flow in a citation network, Journal of Informetrics 12 (4) (2018) 1199–1222。 訳抜け防止モード: [14]I. Park, B. Yoon, 引用ネットワークにおける技術知識フローの予測に基づく技術的収束の技術的機会発見 Journal of Informetrics 12 ( 4 ) ( 2018 ) 1199–1222 。
0.82
[15] M. Bengisu, R. Nekhili, Forecasting emerging technologies with the aid of science and technology
[15]M. Bengisu, R. Nekhili, 科学技術の力を借りて新興技術の予測
0.87
databases, Technological Forecasting and Social Change 73 (7) (2006) 835–844.
データベース、Technological Forecasting and Social Change 73 (7) (2006) 835–844。
0.86
[16] C. Muhlroth, M. Grottke, A systematic literature review of mining weak signals and trends for
C. Muhlroth, M. Grottke, A systematic literature review of mining weak signal and trend for
0.36
corporate foresight, Journal of Business Economics 88 (5) (2018) 643–687.
corporate foresight, journal of business economics 88 (5) (2018) 643–687。
0.43
[17] R. Rohrbeck, M. Bade, Environmental scanning, futures research, strategic foresight and organizational future orientation: a review, integration, and future research directions, paper presented at the ISPIM Annual Conference, June 2012 (2012).
[17]R. Rohrbeck, M. Bade, Environmental scanning, futures research, Strategy foresight and Organization future orientation: a review, integration and future research direction, presented at the ISPIM Annual Conference, June 2012 (2012)。 訳抜け防止モード: [17 ]R. Rohrbeck, M. Bade, 環境スキャン, 未来研究・戦略的展望・組織的未来志向 : レビュー・統合・未来研究の方向性 2012年6月(2012年)、ISPIM年次大会にて発表。
0.65
[18] S. Xu, L. Hao, X. An, G. Yang, F. Wang, Emerging research topics detection with multiple machine
[18]S.Xu, L. Hao, X. An, G. Yang, F. Wang, Emerging research topics detection with multiple machine 訳抜け防止モード: 18 ] s. xu, l. hao, x. an. g. yang氏, f. wang氏, マルチマシンによる新たな研究トピックの検出
0.66
learning models, Journal of Informetrics 13 (4) (2019) p.100983.
Journal of Informetrics 13 (4) (2019) p.100983。
0.63
[19] J. Keller, A. Heiko, The influence of information and communication technology (ict) on future foresight processes—results from a delphi survey, Technological Forecasting and Social Change 85 (2014) 91–92.
19] j. keller, a. heiko, the influence of information and communication technology (ict) on future foresight process—results from a delphi survey, technical forecasting and social change 85 (2014) 91-92
0.43
Deloitte, Hypersonics,
Deloitte ハイパーソニックス
0.30
Breaking from: [20]
破断 から [20]
0.47
New Barriers: The
新しいもの 障壁: その...
0.41
Rise of https://www2.deloitt e.com/content/dam/De loitte/us/Documents/ energyresources/us-b reakingnew-barriers. pdf, 2020.
[21] definitions.net, https://www.definiti ons.net/definition/h ypersonic, Accessed 10 Mar 2022.
関連スポンサーコンテンツ [21] definitions.net, https://www.definiti ons.net/definition/h ypersonic, access 10 mar 2022
0.56
[22] J. Devlin, M. W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for
J. Devlin, M. W. Chang, K. Lee, K. Toutanova, Bert: ディープ双方向トランスフォーマーの事前学習
0.77
language understanding, ArXiv preprint arXiv:1810.04805 (2018).
言語理解, ArXiv preprint arXiv:1810.04805 (2018)。
0.41
[23] Z. Nasar, S. W. Jaffry, M. K. Malik, Textual keyword extraction and summarization: State-of-the-art,
Z. Nasar, S. W. Jaffry, M. K. Malik, Textual keyword extract and summarization: State-of-the-art, 訳抜け防止モード: [23 ]Z. Nasar, S. W. Jaffry, M. K. Malik テキストキーワード抽出と要約 : 現状 - アート-
0.79
Information Processing & Management 56 (6) (2019) p.102088.
Information Processing & Management 56 (6) (2019) p.102088。
0.91
[24] C. Zhang, Automatic keyword extraction from documents using conditional random fields, Journal of
[24] c. zhang, 条件付き乱数を用いた文書の自動キーワード抽出, journal of
0.84
Computational Information Systems 4 (3) (2008) 1169–1180.
計算情報システム 4 (3) (2008) 1169–1180。
0.84
[25] S. K. Bharti, K. S. Babu, Automatic keyword extraction for text summarization: A survey, arXiv preprint
[25] S. K. Bharti, K. S. Babu, Automatic keyword extract for text summarization: A survey, arXiv preprint
0.49
arXiv:1704.03242 (2017).
arXiv:1704.03242 (2017)。
0.34
[26] H. P. Luhn, A statistical approach to mechanized encoding and searching of literary information, IBM
26] h. p. luhn, ibm, 機械化符号化と文芸情報の検索への統計的アプローチ
0.69
Journal of research and development 1 (4) (1957) 309–317.
journal of research and development 1 (4) (1957) 309–317。
0.46
[27] Y. Matsuo, M. Ishizuka, Keyword extraction from a single document using word co-occurrence
【27]松尾英,石塚氏,単語共起を用いた単一文書からのキーワード抽出
0.71
statistical information, International Journal on Artificial Intelligence Tools 13 (01) (2004) 157–169.
International Journal on Artificial Intelligence Tools 13 (01) (2004) 157–169。
0.34
[28] J. Ramos, Using tf-idf to determine word relevance in document queries, in Proceedings of the first
[28] j. ramos, tf-idf を用いて文書クエリにおける単語の関連性を第一審の手続で決定する
0.68
instructional conference on machine learning (Vol. 242, No. 1, pp. 29-48), December 2003 (2003).
機械学習に関する教育会議 (Vol. 242, No. 1, pp. 29-48) 2003年12月(2003年)。
0.89
[29] R. Barzilay, E. M., Using lexical chains for text summarization, Advances in automatic text
[29] r. barzilay, e. m., 辞書連鎖を用いたテキスト要約, 自動テキストの進歩
0.80
summarization (1999) 111–121.
1999年) 111-121頁。
0.40
15
15
0.43
英語(論文から抽出)
日本語訳
スコア
[30] A. Hulth, Improved automatic keyword extraction given more linguistic knowledge, in Proceedings of the 2003 conference on Empirical methods in natural language processing, ACL, 2003, pp. 216-223 (2003).
[30] a. hulth, improved automatic keyword extraction given more language knowledge, in proceedings of the 2003 conference on empirical methods in natural language processing, acl, 2003, pp. 216-223 (2003) 訳抜け防止モード: [30 ]A.Hulth, より言語的な知識を与えられた自動キーワード抽出の改善。 In Proceedings of the 2003 Conference on Empirical Method in natural language processing, ACL, 2003, pp . 216 - 223 ( 2003 ) .
0.89
[31] G. Salton, A. Singhal, M. Mitra, C. Buckley, Automatic text structuring and summarization, Information
[31]G. Salton, A. Singhal, M. Mitra, C. Buckley, Automatic Text structuring and summarization, Information
0.49
Processing & Management 33 (2) (1997) 193–207.
Processing & Management 33 (2) (1997) 193–207。
0.46
[32] K. Zhang, H. Xu, J. Tang, J. Li, Keyword extraction using support vector machine, in international conference on web-age information management (pp. 85-96).
[32] k. zhang, h. xu, j. tang, j. li, keyword extraction using support vector machine, in international conference on web-age information management (pp. 85-96) 訳抜け防止モード: 32 ] k. zhang, h. xu, j. tang, j. li, support vector machine, in international conference on web - age information management (pp 85 - 96) によるキーワード抽出
0.79
Springer, Berlin, Heidelberg, June 2006 (2006).
スプリンガー、ベルリン、ハイデルベルク、2006年6月(2006年)。
0.66
[33] E. Frank, G. W. Paynter, I. H. Witten, C. Gutwin, C. G. Nevill-Manning, Domain-specific key-phrase extraction, in 16th International Joint Conference on Artificial Intelligence, vol.
E. Frank, G. W. Paynter, I. H. Witten, C. Gutwin, C. G. Nevill-Manning, Domain-specific key-phrase extract, in 16th International Joint Conference on Artificial Intelligence, vol. 訳抜け防止モード: He 33 ] E. Frank, G. W. Paynter, I. H. Witten. C. Gutwin, C. G. Nevill - Manning, Domain - specific key - phrase extract, 第16回人工知能国際会議に参加して
0.75
2, 1999, pp.
1999年2月2日、p。
0.61
668-67 (1999).
668-67 (1999).
0.88
[34] S. Rose, D. Engel, N. Cramer, W. Cowley, Automatic keyword extraction from individual documents,
[34] S. Rose, D. Engel, N. Cramer, W. Cowley, 個々の文書から自動キーワード抽出。
0.84
Text mining: applications and theory 1 (2010) 1–20.
テキストマイニング:アプリケーションと理論 1 (2010) 1–20。
0.86
[35] R. Campos, V. Mangaravite, A. Pasquali, A. Jorge, C. Nunes, A. Jatowt, Yake! keyword extraction from
[35]R. Campos, V. Mangaravite, A. Pasquali, A. Jorge, C. Nunes, A. Jatowt, Yake!キーワード抽出 訳抜け防止モード: [35 ] R. Campos, V. Mangaravite, A. Pasquali A. Jorge , C. Nunes , A. Jatowt , Yake ! キーワード抽出
0.94
single documents using multiple local features, Information Sciences 509 (2020) 257–289.
[36] Y. Qian, C. Jia, Y. Liu, Bert-based text keyword extraction, in Journal of Physics: Conference Series (Vol.
Y. Qian, C. Jia, Y. Liu, Bert-based text keyword extract, in Journal of Physics: Conference Series (Vol。 訳抜け防止モード: [36 ]Y. Qian, C. Jia, Y. Liu, Bert-based text keyword extract, in Journal of Physics : Conference Series ( Vol )
0.48
1992, No. 4, p. 042077).
1992年4月4日、p.042077。
0.63
IOP Publishing (2021).
IOPパブリッシング(2021年)。
0.59
[37] P. Sharma, L. Y., Self-supervised contextual keyword and keyphrase retrieval with self-labelling
[37]P. Sharma, L. Y., Self-supervised contextual keyword and keyphrase search with self-labelling
0.43
(2019). [38] J. K. Humphreys, Phraserate: An html key-phrase extractor, dept. of Computer Science, University of
(2019). [38] j. k. humphreys, phraserate: an html key-phrase extractor, dept. of computer science, university of university 訳抜け防止モード: (2019). [38 ]J. K. Humphreys, Phraserate : An html key - phrase extractor, 大学コンピュータサイエンス科
0.60
California, Riverside, California, USA, Tech.
カリフォルニア、リバーサイド、カリフォルニア、アメリカ、テック。
0.75
Rep. (2002). [39] J. R. Thomas, S. K. Bharti, K. S. Babu, Automatic keyword extraction for text summarization in enewspapers, in Proceedings of the International Conference on Informatics and Analytics, ACM, 2016, pp. 86-93 (2016).
2002年)。 J.R. Thomas, S. K. Bharti, K. S. Babu, Automatic keyword extract for text summarization in enewspapers, in the Proceedings of the International Conference on Informatics and Analytics, ACM, 2016 pp. 86-93 (2016) 訳抜け防止モード: 2002年)。 [39 ]J. R. Thomas, S. K. Bharti, K. S. Babu 国際情報・分析学会紀要におけるenewspapersにおけるテキスト要約のためのキーワードの自動抽出 ACM, 2016, pp . 86 - 93 ( 2016 ) .
0.65
[40] H. I. Ansoff, Managing strategic surprise by response to weak signals, California management review
[40]H.I.アンソフ、弱信号への対応による戦略的サプライズ管理、カリフォルニア州経営見直し
0.70
18 (2) (1975) 21–33.
18 (2) (1975) 21–33.
0.49
[41] R. Rohrbeck, N. Thom, H. Arnold, It tools for foresight: The integrated insight and response system of deutsche telekom innovation laboratories, Technological Forecasting and Social Change 97 (2015) 115–126.
[41] r. rohrbeck, n. thom, h. arnold, it tools for for foresight: the integrated insight and response system of deutsche telekom innovation laboratories, technical forecasting and social change 97 (2015) 115–126。 訳抜け防止モード: [41 ] R. Rohrbeck, N. Thom, H. Arnold, 予見ツール : ドイツテレコム研究所の総合的な洞察と応答システム Technological Forecasting and Social Change 97 (2015 ) 115–126。
0.78
[42] H. I. Ansoff, Futures signals sense-making framework (fssf): A start-up tool to analyse and categorise weak signals, wild cards, drivers, trends and other types of information, Futures 42 (1) (2010) 42–48.
[43] B. Coffman, Weak signal research, part i: Introduction, Journal of Transition Management 2 (1) (1997).
[43] b. coffman, weak signal research, part i: introduction, journal of transition management 2 (1) (1997)
0.39
[44] O. Saritas, J. E. Smith, The big picture–trends, drivers, wild cards, discontinuities and weak signals,
44] O. Saritas, J. E. Smith, The big picture–trends, driver, wild card, discontinuity and weak signal。 訳抜け防止モード: 44 ] O. Saritas, J. E. Smith, The big picture – trend? ドライバー ワイルドカード 不連続と弱い信号
0.77
Futures 43 (3) (2011) 292–312.
43 (3) (2011) 292-312。
0.32
[45] J. Yoon, Detecting weak signals for long-term business opportunities using text mining of web news,
[50] O. Pirinen, Weak signal based foresight service, (Master’s thesis) (2010).
[50] O. Pirinen, Weak signal based foresight service, (Master'sthesis) (2010)
0.38
[51] E. Hiltunen, The futures window-a medium for presenting visual weak signals to trigger employees’
[51] hiltunen, the futures window--従業員の引き金となる視覚弱信号提示媒体
0.68
futures thinking in organizations, hSE Publications, working paper (2007).
futures thinking in organizations, hse publications, working paper (2007)を参照。
0.41
[52] C. Park, S. Cho, Future sign detection in smart grids through text mining, Energy Procedia 128 (2017)
52]C. Park, S. Cho, Future sign detection in Smart Grids through text mining, Energy Procedia 128 (2017) 訳抜け防止モード: 52]C. Park, S. Cho, Future sign detection in Smart Grids through text mining エネルギープロセディア128(2017年)
0.74
79–85. 16
79–85. 16
0.39
英語(論文から抽出)
日本語訳
スコア
[53] G. Salton, C. Buckley, Term-weighting approaches in automatic text retrieval, Information processing
53]G. Salton, C. Buckley, 自動テキスト検索における項重み付け手法, 情報処理
0.86
& management 24 (5) (1988) 513–523.
& management 24 (5) (1988) 513–523.
0.50
[54] H. Joho, M. Sanderson, Document frequency and term specificity, in Proceedings of the Recherche
H. Joho, M. Sanderson, Document frequency and term specificity, in Proceedings of the Recherche
0.37
d’Information Assist´ee par Ordinateur Conference (RIAO).
d'Information Assist ́ee par Ordinateur Conference (RIAO)
0.42
Sheffield (2007).
シェフィールド(2007年)。
0.75
[55] D. Nelkin, Publication and promotion.
[55] d. nelkin, publication and promotion.
0.43
the performance of science, Lancet 352 (1998) 893–894.
科学のパフォーマンス lancet 352 (1998) 893–894。
0.74
[56] D. Rennie, V. Yank, L. Emanuel, When authorship fails: a proposal to make contributors accountable,
D. Rennie, V. Yank, L. Emanuel: 著者が失敗したとき: コントリビュータに説明責任を与える提案。
0.72
Jama 278 (7) (1997) 579–585.
ジャマ278 (7) (1997) 579-585。
0.81
[57] Z. Boger, T. Kuflik, P. Shoval, B. Shapira, Automatic keyword identification by artificial neural networks compared to manual identification by users of filtering systems, Information Processing & Management 37 (2) (2001) 187–198.
57] Z. Boger, T. Kuflik, P. Shoval, B. Shapira, 人工知能によるキーワードの自動識別, フィルタリングシステム, Information Processing & Management 37 (2) (2001) 187–198。 訳抜け防止モード: [57 ]Z. Boger, T. Kuflik, P. Shoval, B. Shapira, ニューラルネットワークによるキーワードの自動識別 : フィルタリングシステムのユーザによる手動識別との比較 Information Processing & Management 37 (2 ) (2001 ) 187–198 。
0.87
[58] S. Anjali, N. M. Meera, M. G. Thushara, A graph based approach for keyword extraction from International Conference on Advanced Computational and
58] S. Anjali, N. M. Meera, M. G. Sohara, a graph based approach for keyword extract from International Conference on Advanced Computational and 訳抜け防止モード: [58 ]S. Anjali, N. M. Meera, M. G. Sohara International Conference on Advanced Computational and からのキーワード抽出のためのグラフベースアプローチ
0.87
in 2019 Second documents, Communication Paradigms (ICACCP) (pp. 1-4).
2019年 第2回 文書,通信パラダイム (ICACCP) (pp. 1-4)
0.76
IEEE (2019).
IEEE (2019)。
0.80
[59] G. K. Palshikar, Keyword extraction from a single document using centrality measures, in International conference on pattern recognition and machine intelligence (pp. 503-510).
[59]g. k. palshikar, international conference on pattern recognition and machine intelligence (pp. 503-510) において、集中度尺度を用いた単一の文書からキーワードを抽出する。
0.75
Springer, Berlin, Heidelberg (2007).
springer, berlin, heidelberg (2007)を参照。
0.70
[60] A. Hulth, J. Karlgren, A. Jonsson, H. Bostr¨om, L. Asker, Automatic keyword extraction using domain knowledge, in International Conference on Intelligent Text Processing and Computational Linguistics (pp. 472-482).
A. Hulth, J. Karlgren, A. Jonsson, H. Bostr 'om, L. Asker, Automatic keyword extract using domain knowledge, in International Conference on Intelligent Text Processing and Computational Linguistics (pp. 472-482) 訳抜け防止モード: 60 ] a. hulth, j. karlgren, a. jonsson. ドメイン知識を用いたキーワードの自動抽出 知的テキスト処理と計算言語学に関する国際会議(pp . 472 - 482)において。
0.75
Springer, Berlin, Heidelberg (2001).
スプリンガー、ベルリン、ハイデルベルク(2001年)。
0.68
[61] M. E. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, L. Zettlemoyer, Deep contextualized word representations, in Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2018, New Orleans, Louisiana, USA, June 1-6, 2018, Volume 1 (Long Papers), pages 2227–2237.
M. E. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, L. Zettlemoyer, Deep contextualized word representations, in Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2018, New Orleans, Louisiana, USA, June 1-6, 2018, Volume 1 (Long Papers), page 2227–2237. 訳抜け防止モード: 61 ] M. E. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, L. Zettlemoyer 深層文脈化語表現 : 2018年北米計算言語学会北米支部講演要旨 : ヒューマン・ランゲージ・テクノロジー NAACL - HLT 2018, New Orleans, Louisiana, USA 2018年6月1日~6日、第1巻(長編) 2227-2237頁。
0.80
Association for Computational Linguistics (2018).
Association for Computational Linguistics (2018) の略。
0.77
[62] V. D. Blondel, J. L. Guillaume, R. Lambiotte, E. Lefebvre, Fast unfolding of communities in large
V. D. Blondel, J. L. Guillaume, R. Lambiotte, E. Lefebvre, Fast Unfolding of community in large 訳抜け防止モード: [62 ] V. D. Blondel, J. L. Guillaume, R. Lambiotte, E. Lefebvre, 大規模コミュニティの高速展開
0.89
networks, Journal of statistical mechanics: theory and experiment 10 (2008) p.P10008.
ネットワーク, Journal of statistical mechanics: theory and experiment 10 (2008) p.P10008。
0.93
[63] D. Sziroczak, H. Smith, A review of design issues specific to hypersonic flight vehicles, Progress in
[63]d. sziroczak, h. smith, a review of design issues specifically to hypersonic flight vehicles, progress in
0.42
Aerospace Sciences 84 (2016) 1–28.
エアロスペース・サイエンス 84 (2016) 1-28。
0.57
[64] S. Dinda, K. Vuchuru, S. Konda, A. N. Uttaravalli, Heat management in supersonic/hypersoni c vehicles
[64]S. Dinda, K. Vuchuru, S. Konda, A. N. Uttaravalli, Heat Management in Supersonic/hypersoni c vehicle 訳抜け防止モード: [64 ]S.Dinda,K.Vuchuru,S. Konda, 超音速・超音速車両における熱管理
0.71
using endothermic fuel: Perspective and challenges, ACS omega 6 (40) (2021) 26741–26755.
ACSオメガ6 (40) (2021) 26741-26755。
0.24
[65] J. E. Theisinger, R. D. Braun, Hypersonic entry aeroshell shape optimization, MS Special Problems
[65] J. E. Theisinger, R. D. Braun, Hypersonic entry aeroshell shape Optimization, MS Special Problems
0.49
Report, Georgia Institute of Technology 12 (2007).
ジョージア工科大学第12研究所(2007年)。
0.60
[66] A. Kuranov, A. Korabelnikov, Hypersonic technologies of atmospheric cruise flight under ajax concept, in: 15th AIAA international space planes and hypersonic systems and technologies conference, 2008, p. 2524.
[66] a. kuranov, a. korabelnikov, hypersonic technologies of atmospheric cruise flight under ajax concept, in: 15th aiaa international space planes and hypersonic systems and technologies conference, 2008 p. 2524. (英語) 訳抜け防止モード: A. Kuranov, A. Korabelnikov, Hypersonic Technologies of atmosphere Cruise flight under ajax concept 第15回AIAA国際宇宙機・超音速システム・技術会議 2008年、p.2524。
0.73
[67] J. M. Acton, Hypersonic boost-glide weapons, Science & Global Security 23 (3) (2015) 191–219.
67] J. M. Acton, Hypersonic boost-glide weapons, Science & Global Security 23 (3) (2015) 191–219。 訳抜け防止モード: [67 ]J.M.アクトン,ハイパーソニックブースター-グライダー兵器 Science & Global Security 23 (3 ) (2015 ) 191–219 。
0.72
[68] J. Ismay, Russia claims to use a hypersonic missile in attack on arms depot in Ukraine, New York Times