ARTICLE DE RECHERCHE - Recherche en IA spécialisée au MIT

ARTICLE DE RECHERCHE

Towards establishing a research lineage via
identification of significant citations

Tirthankar Ghosal1**, Piyush Tiwary2**, Robert Patton3, and Christopher Stahl3

1Institute of Formal and Applied Linguistics, Faculty of Mathematics and Physics, Charles University, Czech Republic
2Indian Institute of Science, India
3Oak Ridge National Laboratories, NOUS
**equal contribution.

Mots clés: academic influence, citation classification, citation graph, citation significance
detection, machine learning, research lineage

ABSTRAIT

Finding the lineage of a research topic is crucial for understanding the prior state of the art and
advancing scientific displacement. The deluge of scholarly articles makes it difficult to locate
the most relevant previous work. It causes researchers to spend a considerable amount of time
building up their literature list. Citations play a crucial role in discovering relevant literature.
Cependant, not all citations are created equal. The majority of the citations that a paper receives
provide contextual and background information to the citing papers. In those cases, the cited
paper is not central to the theme of citing papers. Cependant, some papers build upon a given
paper and further the research frontier. In those cases, the concerned cited paper plays a
pivotal role in the citing paper. Ainsi, the nature of the citation that the former receives from
the latter is significant. In this work, we discuss our investigations towards discovering
significant citations of a given paper. We further show how we can leverage significant
citations to build a research lineage via a significant citation graph. We demonstrate the
efficacy of our idea with two real-life case studies. Our experiments yield promising results
with respect to the current state of the art in classifying significant citations, outperforming the
earlier ones by a relative margin of 20 points in terms of precision. We hypothesize that such
an automated system can facilitate relevant literature discovery and help identify knowledge
flow for a particular category of papers.

INTRODUCTION

Literature searches are crucial to discover relevant publications. The knowledge discovery that
ensues forms the basis of understanding a research problem, finding the previously explored
frontiers and identifying research gaps, which eventually leads to the development of new
ideas. Cependant, with the exponential growth of scientific literature (including published
papers and preprints) (Ghosal, Sonam et al., 2019b), it is almost impossible for a researcher
to go through the entire body of the scholarly works, even in a very narrow domain. Citations
play an important role here in finding the relevant articles that further topical knowledge.
Cependant, not all citations are equally effective (Zhu, Turney et al., 2015) effective in finding
relevant research. A majority of papers cite a work contextually (Pride & Knoth, 2017un) pour
providing additional background context. Such background contextual citations help in the
broader understanding; cependant, they are not central to the citing paper’s theme. Some papers
use the ideas in a given paper, build upon those ideas, and displace the body of relevant

un accès ouvert

journal

Citation: Ghosal, T., Tiwary, P., Patton,
R., & Stahl, C. (2021). Towards
establishing a research lineage via
identification of significant citations.
Études scientifiques quantitatives, 2(4),
1511–1528. https://doi.org/10.1162/qss
_a_00170

EST CE QUE JE:
https://doi.org/10.1162/qss_a_00170

Auteur correspondant:
Tirthankar Ghosal
ghosal@ufal.mff.cuni.cz

droits d'auteur: © 2021 Tirthankar Ghosal,
Piyush Tiwary, Robert Patton, et
Christopher Stahl. Published under a
Creative Commons Attribution 4.0
International (CC PAR 4.0) Licence.

La presse du MIT

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

recherche. Such papers are expected to acknowledge the prior work (via citing them) duly.
Cependant, the nature of citation, in this case, is different from that of contextual citations. These
citations, which heavily rely on a given work or build upon that work, are significant citations.
Cependant, the current citation count metric puts equal weights on all the citations. Donc, it
is inadequate for identifying the papers that have significantly cited a given work and may
have taken the relevant research forward. Identifying such significant citations is hence crucial
to the literature study.

It is not uncommon that authors sometimes fail to acknowledge relevant papers’ role in
influencing their ideas (Rousseau, 2007; Van Noorden, 2017). Par conséquent, researchers spend
a lot of their time searching for the papers most relevant to their research topic, thereby locat-
ing the subsequent papers that carried forward a given scientific idea. It is usually desirable for
a researcher to understand the story behind a prior work and trace the concept’s emergence
and gradual evolution through publications, thereby identifying the knowledge flow.
Researchers ideally curate their literature base by identifying significant references to a given
paper and then hierarchically locating meaningful prior work.

The idea of recognizing significant citations is also important to understand the true impact
of given research or facility. To understand how pervasive particular research was in the com-
munity, it is essential to understand its influence beyond the direct citations it received. To this
end, tracking the transitive influence of research via identifying significant citations could be
one possible solution.

In this work, we develop automatic approaches to trace the lineage of given research via
transitively identifying the significant citations to a given article. The overall objective of our
work is twofold:

(cid:129) Accelerate relevant literature discovery via establishing a research lineage.
(cid:129) Find the true influence of a given work and its pervasiveness in the community beyond

citation counts.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

There are two aspects to the problem: identifying the relevant prior work and identifying the
follow-up works that stemmed from or are influenced by the current work. The first aspect
would facilitate relevant prior literature discovery for a paper. In contrast, the second aspect
would facilitate discovering the knowledge flow in subsequent relevant papers. Obviously, notre
approach would not be a one size fits all approach. Toujours, we believe it is effective to find inves-
tigations that build upon relevant priors and facilitate relevant literature discovery, and thereby
steer towards identifying the pervasiveness of a given piece of research in the community. Nous
base our work on classifying citations as contextual or significant and trace the lineage of
research in a citation graph via identifying significant edges. The major contributions of the
current work are the following:

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

(cid:129) We use a set of novel and rich features to classify citations as significant or contextual.
(cid:129) A graph-based approach to tracing the lineage of a given research work leveraging on

citation classification.

2. RESEARCH LINEAGE

The mechanism of citations in academia is not always transparent (Van Noorden & Singh
Chawla, 2019; Vîiu, 2016; West, Stenius, & Kettunen, 2017). Problems such as coercive

Études scientifiques quantitatives

1512

Towards establishing a research lineage via identification of significant citations

citations (Wilhite & Fong, 2012), anomalous citations (Bai, Xia et al., 2016), citation manipulation
(Bartneck & Kokkelmans, 2011), rich get richer effects (Ronda-Pupo & Pham, 2018), et
discriminatory citation practices (Camacho-Miñano & Núñez-Nickel, 2009) have infested
the academic community. Cependant, in spite of all these known issues, citation counts and
h-indices still remain the measures of research impact and tools for academic incentives,
though long-debated by many (Cerdá, Nieto, & Campos, 2009; Laloë & Mosseri, 2009).
Généralement, we measure the impact of a given paper by the direct citations it receives. Cependant,
a given piece of research may have induced a transitive effect on other papers, which is not
apparent with the current citation count measures. Chiffre 1 shows a sample citation network
where A could be a paper or a research facility. We want to know how pervasive was the
research or facility A in the community. At d = 1 are the direct citations to A. We see that
article B cite A significantly, or B is inspired by A. Other citations to A are background. À
citation depth d = 2, we see that article C and article D significantly cite B (direct citation).
We see that C also cites A significantly. Enfin, at citation depth d = 3, E significantly cites C.
We intend to understand if there is a lineage of research from A to E (A → B → C → E).
Although E does not cite A directly, can we identify A’s influence on E? If E is a seminal work
receiving hundreds of citations, can we infer that A was the prior work that indirectly inspired
E? We are interested in discovering such hidden inspirations to honestly assess the contribu-
tions of a research article or a facility.

3. RELATED WORK

Measuring academic influence has become a research topic because publications are associ-
ated with academic prestige and incentives. Several metrics (impact factor, eigen factor,
h-index, citation counts, altmetrics, etc.) have been devised to comprehend research impact
efficiently. Toujours, each one is motivated by a different aspect and has found varied importance
across disciplines. Zhu et al. (2015) did pioneering work on academic influence prediction
leveraging citation context. Shi, Wang et al. (2019) presented a visual analysis of citation
context-based article influence ranking. Xie, Sun, and Shen (2016) predicted paper influence
in an academic network by taking into account the content and venue of a paper, ainsi que
the reputation of its authors. Shen, Song et al. (2016) used topic modeling to measure
academic influence in scientific literature. Manju, Kavitha, and Geetha (2017) identified influ-
ential researchers in an academic network using a rough-set based selection of time-weighted

Chiffre 1. Research lineage.

Études scientifiques quantitatives

1513

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

academic and social network features. Pileggi (2018) did a citation network analysis to mea-
sure academic influence. Zhang and Wu (2020) used a dynamic academic network to predict
the future influence of papers. Ji, Tang, and Chen (2019) analyzed the impact of academic
papers based on improved PageRank. F. Wang, Jia et al. (2019) assessed the academic influ-
ence of scientific literature via altmetrics. F. Zhao, Zhang et al. (2019) measured academic
influence using heterogeneous author-citation networks. Recently, many deep learning–based
methods have been explored for citation classification. Perier-Camby, Bertin et al. (2019)
attempt to compare deep learning-based methods with rule-based methods. They use deep
learning–based feature extractors such as BCN (McCann, Bradbury et al., 2017) and ELMo
(Peters, Neumann et al., 2018) to extract semantic information and feed it to various classifiers
for classification. They conclude that neural networks could be a potential dimension for cita-
tion classification when a large number of samples are available. Cependant, for a small data set
such as the one we use, rule-based methods clearly hold an advantage. Apart from this, le
features used in rule-based methods are more comprehensible than features extracted from
deep learning methods, thus providing deeper insights into analyzing factors that make a
citation significant or contextual.

The closest literature for our task is that on citation classification. Citation classification has
been explored in the works of Alvarez, Soriano, and Martínez-Barco (2017), Dong and Schäfer
(2011), Qayyum and Afzal (2019), and Teufel, Siddharthan, and Tidhar (2006). These works
use features from the perspective of citation motivation. On the other hand, there are works that
emphasize on features from a semantic perspective. Wang, Zhang et al. (2020) use syntactic
and contextual information of citations for classification. Aljuaid, Iftikhar et al. (2021) et
Amjad and Ihsan (2020) perform classification based on sentiment analysis of in-text citations.
Athar (2011) and Ihsan, Imran et al. (2019) propose sentiment analysis of citations using
linguistic studies of the citance. Plus récemment, several open-source data sets for citation
classification have been developed in the work of Cohan, Ammar et al. (2019) and Pride
and Knoth (2020). Valenzuela, Ha, and Etzioni (2015) explored citation classification into
influential and incidental using machine learning techniques which we adapt as significant
and contextual respectively in this work.

In this work, we propose a rich set of features informed from both citation and context
(semantics) perspectives, leveraging advantages of both types, thus performing better than
all of the methods mentioned above. Cependant, our problem is motivated beyond citation clas-
sification. We restrict our classification labels to significant and contextual, unlike Valenzuela
et autres. (2015), as these labels are enough to trace the lineage of a work. En outre, to the best
of our knowledge, we did not find any work leveraging citation classification for finding a
research lineage. Ainsi, we only compare our performance for the citation significance
detection subtask with other approaches.

4. DATA SET DESCRIPTION

We experiment with the Valenzuela data set (Valenzuela et al., 2015) for our task. The data set
consists of incidental/influential human judgments on 630 citing-cited paper pairs for articles
drawn from the 2013 ACL anthology, the full texts of which are publicly available. Two expert
human annotators determined the judgment for each citation, and each citation was assigned
a label. Using the author’s binary classification, 396 citation pairs were ranked as incidental
citations, et 69 (14.3%) were ranked as influential (important) citations. For demonstrating
our research lineage idea, we explore knowledge flow on certain papers of Document-Level
Novelty Detection (Ghosal, Salam et al., 2018b) and the High Performance Computing (HPC)

Études scientifiques quantitatives

1514

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

algorithm MENNDL (Jeune, Rose et al., 2015). The actual authors of these two topics helped
us with manual annotation of their paper’s lineage.

5. METHODOLOGY

To identify significant citations, we pursue a feature-engineering approach to curate several
features from cited-citing paper pairs. The objective is to classify the citations received by a
given paper into SIGNIFICANT and CONTEXTUAL. The original cited citing papers in the
Valenzuela data set are in PDF. We convert the PDFs to corresponding XMLs using GROBID
(Lopez, 2009). We use GROBID to parse our PDFs into XMLs as well as manually correcting a
few inconsistent files so that there is no discrepancy.

1. Citation frequency inside the body of citing paper (F1): We measure the number of
times the cited paper is referenced from within the citing paper’s body. The intuition is
that if a paper is cited multiple times, the cited paper may be significant to the citing
papier.

2. Are the authors of citing and cited paper the same? (Boolean) (F2): We check if the
authors of the citing and cited papers are the same. This might be a case of self-citation
or can also signal the extension of the work.

3. Author overlap ratio (F3): This measures number of common authors in the citing and
cited papers normalized to the total number of authors in the citing paper. The intui-
tion is similar to F2.
Is the citation occurring in a table or figure caption? (Boolean) (F4): The intuition is
that most of the citations in tables and figures appear for comparison/significantly
referencing existing work. Ainsi, the citing paper might be an extension of the cited
article or may have compared it with earlier significant work.
Is the citation occurring in groups? (Boolean) (F5): We check if the citation is occur-
ring along with other citations in a group. The intuition is that such citations generally
appear in related works to highlight a background detail; hence, they might not be a
significant citation.

6. Number of citations to the cited paper normalized by the total number of citations
made by the citing paper (F6): This measures the number of citations to the cited
paper by the citing paper normalized by the total number of citation instances in
the citing paper. This measures how frequently the cited paper is mentioned compared
to other cited papers in the citing paper.

7. Number of citations to the cited paper normalized by the total number of bibliography
items in the citing paper (F7): This measures the number of citations to the cited paper
normalized to the total number of bibliography items in the citing paper. The intuition is
similar to F6.
tf-idf similarity between abstracts of the cited and citing paper (F8): We take cosine
similarity between the tf-idf representations of the abstracts of cited and citing papers.
The intuition is that if the similarity is higher, the citing paper may be inspired/extended
from the cited paper.
tf-idf similarity between titles of the cited and citing paper (F9): We take cosine sim-
ilarity between the tf-idf representations of the titles of cited and citing papers.
10. Average tf-idf similarity between citance and abstract of the cited paper (F10): Nous
calculate the similarity of each citance with the abstract of the cited article and take
the average of it. Citances are sentences containing the citations in the citing paper.
Citances reveal the purpose of the cited paper in the citing paper. Abstracts contain the

Études scientifiques quantitatives

1515

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

contribution/purpose statements of a given paper. Hence similarity with citances may
suggest that the cited paper has been used significantly in the current paper.

11. Maximum tf-idf similarity between citance and abstract of the cited paper (F11): Nous
take the maximum of similarity of the citances (there could be multiple citation
instances of the same paper in a given paper) with the abstract of the cited paper.
12. Average tf-idf similarity between citance and title of the cited paper (F12): We calculate

the similarity of each citance with the title of the cited paper and take an average of it.

13. Maximum tf-idf similarity between citance and title of the cited paper (F13): We take

the maximum of similarity of the citances with the title of the cited paper.

14. Average length of the citance (F14): Average length of the citances (in words) pour
multiple citances. The intuition is that if the citing paper has spent many words on
the cited article, it may have significantly cited the corresponding article.

17.

15. Maximum length of the citance (F15): Maximum length of the citances (in words).
16. Non. of words between citances (F16): We take the average of the number of words
between each pair of consecutive citances of the cited paper. This is set to 0 dans le
case of a single citance.
In how many different sections does the citation appear in the citing paper? (F17):
We take the number of different sections in which the citation to a cited paper occurs
and normalize it with the total number of sections present in the citing paper. Le
intuition is that if a citation occurs in most sections, it might be a significant citation.
18. Number of common references in citing and cited paper normalized by the total
number of references in citing article (F18): We count the number of common
bibliographic items present in the citing and cited papers and normalize it with total
bibliographic items present in the citing paper.

19. Number of common keywords between abstracts of the cited and citing paper
extracted by YAKE (Campos, Mangaravite et al., 2018) (F19): We compare the num-
ber of common keywords between the abstracts of the citing and cited papers
extracted using YAKE. Our instinct is that a greater number of common keywords
would denote a greater similarity between abstracts.

20. Number of common keywords between titles of the cited and citing paper extracted
by YAKE (F20): We compare the number of common keywords between the titles of
the citing and cited papers extracted using YAKE.

21. Number of common keywords between the body of the cited and citing papers
extracted by YAKE (F21): We compare the number of common keywords between
the body of the citing and cited papers extracted using YAKE.

22. Word Mover’s Distance ( WMD) (Huang, Guo et al., 2016) between the abstracts of
the cited and citing papers (F22): We measure the WMD between the abstracts of the
citing and cited papers. The essence of this feature is to calculate semantic
distance/similarity between abstracts of the two papers.

23. WMD between titles of the cited and citing papers (F23): We measure the WMD

between the titles of the citing and cited papers.

24. WMD between the bodies of the cited and citing papers (F24): We measure the

WMD between the bodies of the citing and cited papers.

25. Average WMD between citance and abstract of the cited and citing papers (F25): Nous
take the average of WMDs between the citance and abstract of the cited paper.
26. Maximum WMD between citance and abstract of the cited and citing papers (F26):
We take the maximum of WMDs between the citance and abstract of the cited paper.
27. Average VADER (Gilbert & Hutto, 2014) polarity index—Positive (F27), Negative
(F28), Neutral (F29), Compound (F30): We measure the VADER polarity index of

Études scientifiques quantitatives

1516

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

all the citances of the cited paper, and take their average for each sentiment (positive,
negative, neutral, and compound).

28. Maximum VADER polarity index—Positive (F31), Negative (F32), Neutral (F33),
Compound (F34) of citances: We measure the VADER polarity index of all the
citances of the cited paper, and take the maximum among them for each sentiment
(positive, negative, neutral, and compound). The intuition to use sentiment informa-
tion is to understand how the citing paper cites the cited paper.

29. Number of common venues in the bibliographies of the citing and cited papers (F35):
We count the number of common venues mentioned in the bibliographies of the citing
and cited papers and normalize it with the number of unique venues in the citing
papier. Higher venue overlap would signify that the papers are in the same domain
(Ghosal et al., 2019b).

30. Number of common authors in the bibliographies of the citing and cited papers
(F36): We count the number of common authors mentioned in the bibliographies of
the citing and cited papers and normalize it with the number of unique authors in the
citing paper (Ghosal et al., 2019b).

As mentioned earlier, only 14.3% of total citations are labeled as significant, which poses a
class imbalance problem. To address this issue, we use SMOTE (Chawla, Bowyer et al., 2002)
along with random undersampling of the majority (contextual citation) class. We first split the
data set into 60% training and 40% testing data. Then we undersample the majority class by
50%, and oversample the minority class by 40% on the training partition of the data set.

6. EVALUATION

Our evaluation consists of two stages: D'abord, we evaluate our approach on the citation signifi-
cance task. Suivant, we try to see if we can identify the research lineage via tracing significant
citations across the two research topics (Document-level Novelty and MENNDL). We ask the
original authors to annotate the lineage and verify it with our automatic method. We train our
model on the Valenzuela data set and use that trained model to predict significant citations of
Document-Level Novelty and MENNDL papers, and thereby try to visualize the research lin-
eage across the citing papers. We curate a small citation graph to demonstrate our idea. Note
that our task in concern is Citation Significance Detection, which is different from Citation
Classification in Literature. Whereas Citation Classification focuses on identifying the citation’s
intent, Citation Significance aims to identify the value associated with the citation. Obviously,
the two tasks are related to each other, but the objectives are different.

6.1. Citation Significance Detection

The goal of this task is to identify whether a citation was SIGNIFICANT or CONTEXTUAL. Nous
experiment with several classifiers for the binary classification task such as kNN (k = 3),
Support Vector Machines (kernel = RBF), Decision Trees (max depth = 10) and Random Forest
(n estimators = 15, max depth = 10). We found Random Forest to be the best performing one
with our feature set. Tableau 1 shows our current results against the earlier reported results on the
Valenzuela data set. We attain promising results compared to earlier approaches with a rela-
tive improvement of 20 points in precision. As the data set is small, neither earlier works nor
we attempted a deep neural approach for citation classification on this dataset. Like us,
Qayyum and Afzal (2019) also used Random Forest as the classifier; cependant, they relied
on metadata features rather than content-based features for their work. Their experiments tried
to answer the following questions: To what extent can the similarities and dissimilarities

Études scientifiques quantitatives

1517

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Tableau 1.

Results on Citation Significance Detection on the Valenzuela data set

Methods
Valenzuela et al. (2015)

Qayyum and Afzal (2019)

Nazir, Asif, and Ahmad (2020un)

Nazir, Asif et al. (2020b)

Current Approach

Precision
0.65

0.72

0.75

0.85

0.92

between metadata parameters serve as useful indicators for important citation tracking? Et:
Which metadata parameters or their combinations are helpful in achieving good results? Nous
specifically work with full-text content-based features; hence our approach leverages richer
information because it takes into consideration the full text of the works, whereas Qayyum
and Afzal (2019) is solely based on metadata which helps us to achieve better performance.

Tableau 2 shows the classification results of the various classifiers we experimented with.
Clairement, our features are highly interdependent (Section 5), which explains the better perfor-
mance of Random Forests.

Chiffre 2 shows the importance of the top 10 features ranked by their information gain.
Cependant, our experimental data set is small and our features corelated; hence it seems that
some features have marginal contributions. We deem that in a bigger real-life data set, le
feature significance would be more visible. Ici, we can see that features such as distance
between citances, the number of concerned citations normalized by the total number of cita-
tion, similarity between cited-citing abstracts, in-text citation frequency, and the similarity
between citance and cited abstract play an important role in the classification. The other fea-
tures in the top 10 are distance between citance, number of citations from citing to cited nor-
malized by the total citations made by the citing paper, the similarity between cited-citing
abstracts, in-text citation frequency, the average similarity between citance and cited abstract,
number of citations from citing to cited normalized by the total references made by the citing
papier, number of common YAKE keywords between the body of citing and cited paper, le
average similarity between citance and title of cited paper, the max similarity between citance
and abstract of cited paper, and neutral sentiment polarity of citance. We explain the possible
reasons behind the performance of these features in the subsequent sections. The precision
using only the top 10 features is 0.73. Ainsi, other features play a significant role as well.
A complete list of features and the corresponding information gain is given in Table 3.

Tableau 2.

Classification result of various classifiers for Citation Significance

Methods
kNN

SVM

Decision Tree

Random Forest

Precision
0.80

0.79

0.80

0.92

Recall
0.87

0.67

0.82

F1 score
0.83

Accuracy
0.81

0.73

0.81

0.87

0.81

0.86

0.90

1518

Études scientifiques quantitatives

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Chiffre 2.

Feature importance ranked via information gain.

To analyze the contribution of each feature, we evaluate our model using single feature at a
time similar to Valenzuela et al. (2015). The precision after considering each feature individ-
ually is shown in Table 4. It is seen that the first 28 features in the table contribute significantly
to the classification, and the overall precision after considering all the features is even better
(an improvement of 14 points). F1, F4 (suggesting that significant citations do occur in tables or
figures), and F21, followed by F7, F19, and F3 are the best performing features. This indicates
that features obtained from a citation perspective are more useful. On the other hand, the worst
performing features are F20 (perhaps due to the small size of the data set), F13, F5 (suggérant
that significant citations also occur in groups), F17, F2, F22, and F33. Most of our observations
are in line with Valenzuela et al. (2015).

Pride and Knoth (2017b) found Number of Direct Citations, Author Overlap, and Abstract
Similarity to be the most important features. Our approach performs well enough to proceed
with the next stage.

Tableau 3.
information gain

Information gain (IG) due to each feature. Features are ranked in decreasing order of

Feature
F16

F10

F21

F12

F13

F33

F35

F26

IG
0.147

0.070

0.065

0.061

0.041

0.033

0.031

0.030

0.025

0.024

Feature
F24

F13

F33

F18

F23

F35

F34

F19

F22

F28

F25

IG
0.024

0.022

0.021

0.020

0.019

0.017

0.016

Feature
F30

F27

F31

F32

F15

F17

F36

F20

IG
0.015

0.015

0.014

0.013

0.011

0.006

0.004

0.003

1519

Études scientifiques quantitatives

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Tableau 4.
are listed in decreasing order of the precision

Performance of Random Forest model by using individual features at a time. The features

Feature
F1

Precision
0.78

Feature
F15

Precision
0.28

Feature
F25

Precision
0.15

F21

F19

F16

F28

F35

F32

F33

Total

0.76

0.71

0.68

0.61

0.50

0.47

0.43

0.37

0.33

0.29

F10

F27

F23

F36

F35

F12

F34

F30

F18

0.27

0.25

0.23

0.20

0.19

0.17

0.15

F11

F24

F31

F26

F33

F22

F17

F13

F20

0.14

0.13

0.11

0.10

0.08

0.07

0.04

0.03

0.01

0.92

It is important to note that despite so many features, it is possible that some features might
be correlated. Ainsi, we find the Pearson’s correlation coefficient between each pair of fea-
tures to see how dependent they are on each other. The heatmap of the correlation matrix is
shown in Figure 3.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Chiffre 3. Heatmap of correlation between various pair of features.

Études scientifiques quantitatives

1520

Towards establishing a research lineage via identification of significant citations

Tableau 5.

Feature pairs with high correlation coefficient

Feature pair
F10 & F11

Correlation coefficient
0.937

Feature pair
F25 & F26

Correlation coefficient
0.910

F30 & F34

F28 & F32

F27 & F31

F12 & F13

0.919

0.917

0.914

0.910

F9 & F20

F29 & F33

F1 & F15

F27 & F29

0.907

0.905

0.835

0.832

We find that the average correlation coefficient between all the features is 0.074. Cependant,
there are few pairs of features that have high correlation coefficients. We have listed such pairs
in Table 5.

From Table 5 we can see that feature pairs such as F10 & F11, F30 & F34, F28 & F32, F27 &
F31, F12 & F13, F25 & F26, and F29 & F33 have high correlation, which is understandable, comme
these pairs are nothing but the maximum and average of the same quantity measured through-
out the corresponding literature. Ainsi, to reduce the complexity of the classifier, one may
use just one of the features from each pair. The results after combining these features are
shown in Table 6. It can be seen that even after combining these features, there is no significant
degradation in performance of our model.

6.2. The 3C Data Set

As we mention earlier, the data set used is small, due to which the significance of each feature
might not be visible explicitly. Ainsi, we also test our method on the 3C data set, which is
larger. The 3C Citation Context Classification1 Shared Task organized as part of the Second
Workshop on Scholarly Document Processing @ NAACL 2021 is a classification challenge
where each citation context is categorized based on its purpose and influence. It consists of
two subtasks:

(cid:129) Task A: Multiclass classification of citation contexts based on purpose with categories
BACKGROUND, USES, COMPARES CONTRASTS, MOTIVATION, EXTENSION, et
FUTURE.

(cid:129) Task B: Binary classification of citations into INCIDENTAL or INFLUENTIAL classes (c'est à dire.,

a task for identifying the importance of a citation).

The training and test data sets used for Task A and Task B are the same. The training data
and test data consist of 3,000 et 1,000 instances, respectivement. We use the data for Task B in
our experiments. Cependant, the 3C data set does not provide us with full text, so we are only
able to test only 19 of our features. We achieved an F1 score of 0.5358 with these 19 features
on the privately held 3C test set. Our relevant features in use here are F1, F2, F9, F10, F11,
F12, F13, F14, F15, F20, F23, F27, F28, F29, F30, F31, F32, F33, and F34. We provide the
results on the validation set using a Random Forest Classifier in Table 7. The best performing
system in 3C achieved an F1 score of 0.60 while the baseline F1 score was 0.30.

1 https://sdproc.org/2021/sharedtasks.html#3c

Études scientifiques quantitatives

1521

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Tableau 6.

Citation significance results after combining features on the Valenzuela data set

Precision
0.91

Recall
0.80

F1 score
0.85

Accuracy
0.89

6.3. Research Lineage: Études de cas

Our end goal is not just citation classification but to make use of a highly accurate citation
significance detection approach to trace significant citations and, thereafter, to try to establish
a lineage of the given research. As explained in Section 2, by research lineage we aim to iden-
tify the idea propagation via tracking the significant citations. Pour y parvenir, we create a
Significant Citation Graph (SCG). This is a graph-like structure in which where each node rep-
resents a research paper. There is a directed edge between each cited-citing pair, whose direc-
tion is from cited paper node to citing paper node, indicating the flow of knowledge from cited
paper to citing paper. In the usual case, all citations have equal weights in a citation graph.
Cependant, in our case, each edge is labeled as either significant or contextual, using the
approach we discussed in the previous section. Our idea is similar to that of existing scholarly
graph databases; cependant, we go one step further and depict how a particular concept or
knowledge has propagated with consecutive citations.

Algorithm 1 shows the method to create the adjacency list for the SCG. The Citation
Significance Detection ML model is trained on a given data set ( Valenzuela in our case). À
demonstrate the effectiveness of our method, we present an SCG for a set of papers on
Document-Level Novelty Detection and MENNDL. Being the authors of the papers on these
topics, we have identified the significant citations of each paper and used it to test the effec-
tiveness of our proposed method to create an SCG.

Algorithm 1.

Algorithm to Create Significance Citation Graph

Input: Trained Model & concerned research document, P.

Output: Adjacency List for Citation Graph

Initialize adjacency list, UN

Initialize an empty queue, Q

Q.add(P.)

while Q is not empty do

for Each citation, C in Q[0] do

Extract features (F1-F36) for C

if C is Significant and C is not in Q then

Q.add(C )

UN[Q[0]].add(C )

Q.pop()

return A

Études scientifiques quantitatives

1522

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Tableau 7.
Classifier

Precision
0.569

Citation Influence Classification results on 3C Validation Set using a Random Forest

Recall
0.575

F1 score
0.572

Accuracy
0.606

6.3.1. Case study I: Document-level novelty detection

Chiffre 4 denotes an excerpt of a SCG from our Document-Level Novelty Detection papers.
The red edges denote significant citations, whereas black edges denote contextual citations.
Our approach determined if a citation edge is significant or contextual. In the citation graph,
we are interested in the lineage among four textual novelty detection papers (P1, P2, P3, P4),
which the original authors annotate. We annotated that P1 is the pivot paper that introduced
their document-level novelty detection data set, and their other papers P2, P3, and P4 are
based on P1. While P2 and P4 address novelty classification, P3 aims to quantify textual
novelty. Our approach conforms to the annotation by the original authors. With P1 as the pivot
we can see that there are significant edges from P1 to each of P2, P3, and P4. There is also a
significant edge between P2 and P4. Cependant, there is no edge between P2 and P3, as they
were contemporaneous submissions and their objective was different (P2 was about novelty
classification and P3 was about novelty scoring). P1 → P2 → P4 forms a research lineage as P2
extends on P1 and P4 extends on P2. En outre, we see that P12, P25, P24, and P22 (tran-
sitively) are some influential papers for P1. We verified from the authors that P25 was the
paper to introduce the first document-level novelty detection data set but from an information
retrieval perspective. P25 inspired the authors to create the data set in P1 for ML experiments.
We infer that P12, P22, and P24 had significant influence on their investigations with P1.
Ainsi, our approach (trained on a different set of papers in the Valenzuela data set) proved
successful to identify the significant citations and thereby also identify the corresponding
lineage.

Significant Citation Graph for a set of papers on Document-Level Novelty Detection.
Chiffre 4.
Please refer to the bibliography for the paper details. P1 → Ghosal et al. (2018b), P2 → Ghosal,
Edithal et al. (2018un), P3 → Ghosal, Shukla et al. (2019un), P4 → Ghosal, Edithal et al. (2020), P6 →
Soboroff and Harman (2003), P7 → Zhao and Lee (2016), P8 → Colomo-Palacios, Tsai, and Chan
(2010), P9 → Tang, Tsai, and Chen (2010), P11 → Kusner, Sun et al. (2015), P12 → Li and Croft
(2005), P17 → Schiffman and McKeown (2005), P22 → Allan, Wade, and Bolivar (2003), P23 →
Soboroff and Harman (2005), P24 → Karkali, Rousseau et al. (2013), P25 → Zhang, Callan, et
Minka (2002), P28 → Zhang, Zheng et al. (2015).

Études scientifiques quantitatives

1523

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Significant Citation Graph for a set of papers on the MENNDL HPC algorithm. P1 →
Chiffre 5.
Patton, Johnston et al. (2018), P4 → Young, Rose et al. (2017), P6 → Patton, Johnston et al.
(2019), P7 → Johnston, Young et al. (2019), P8 → Johnston, Young et al. (2017), P9 → Young
et autres. (2015), P12 → Chae, Schuman et al. (2019), P13 → Jia, Shelhamer et al. (2014), P14 → Saltz,
Gupta et al. (2018), P15 → Thorsson, Gibbs et al. (2018), P16 → Bottou (2010), P17 → Noh, Hong,
and Han (2015), P18 → Lucchi, Márquez-Neila et al. (2014), P19 → Baldi, Sadowski, and Whiteson
(2014), P20 → Ciregan, Meier, and Schmidhuber (2012), P25 → Y. Zhang et al. (2002), P28 →
Zhang et al. (2015).

6.3.2. Case study II: MENNDL HPC algorithm

We tested our approach’s efficacy to predict the lineage of the high-performance computing
algorithm MENNDL. We show the research lineage of MENNDL (Young et al., 2015) dans
Chiffre 5. We asked the original authors to annotate their research progression with MENNDL.
According to the authors, the first paper to describe the MENNDL algorithm was published in
2015, which is deemed the pivot (P9). The follow-up paper that carried forward the work in P9
was P4 in 2017. Then P1 came in 2018, building upon P4. P7 and P12 came as extensions of P4.
Suivant, P6 came in 2019 that took forward the work from P1. With P9 as the source, our approach
correctly predicted the lineage as P9 → P4 → P1 → P6. Aussi, the lineage P9 → P4 → P12 and
P9 → P4 → P7 via tracing significant citations could be visible in the SCG at Figure 4. Nous
annotate P8 as an application of P9; hence no significant link exists between P9 and P8.

From the above experiments and case studies, it is clear that our proposed method works
reasonably well when a paper cites the influencing paper meaningfully. Cependant, there are
cases where some papers do not cite the papers by which they are inspired. Dans de tels cas, notre
method would not work.

7. CONCLUSION AND FUTURE WORK

In this work, we present our novel idea towards finding a research lineage to accelerate liter-
ature review. We achieve state-of-the-art performance on citation significance detection,
which is a crucial component to form the lineage. We leverage that and show the efficacy
of our approach on two completely different research topics. Our approach is simple and
could be easily implemented on a large-scale citation graph (given the papers’ full text).
The training data set is built from NLP papers. Cependant, we demonstrate our approach’s effi-
cacy by testing on two topics: one from NLP and the other from HPC, hence establishing that
our approach is domain agnostic. Identifying significant citations to form a research lineage
would also help the community to understand the real impact of a research beyond simple
citation counts. We would look forward to experimenting with deep neural architectures to
identify meaningful features for the current task automatically. Our next foray would be to
identify the missing citations for papers that may have played an instrumental role in certain
papers but unfortunately are not cited. We release all the codes related to our experiment at
https://figshare.com/s/2388c54ba01d2df25f38.

Études scientifiques quantitatives

1524

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

CONTRIBUTIONS DES AUTEURS

Tirthankar Ghosal: Conceptualisation, Conservation des données, Analyse formelle, Enquête, Method-
ology, Logiciel, Writing—Original draft. Piyush Tiwary: Analyse formelle, Implementation,
Writing—Original draft. Robert Patton: Acquisition de financement, Surveillance. Christopher Stahl:
Conceptualisation, Conservation des données, Gestion de projet, Rédaction—Révision & édition.

REMERCIEMENTS

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-
00OR22725 with the U.S. Department of Energy (DOE). The views expressed in the article
do not necessarily represent the views of the DOE or the U.S. government. The U.S. govern-
ment retains and the publisher, by accepting the article for publication, acknowledges that the
U.S. government retains a nonexclusive, paid-up, irrevocable, world-wide license to publish
or reproduce the published form of this manuscript, or allow others to do so, for U.S. govern-
ment purposes. The Department of Energy will provide public access to these results of
federally sponsored research in accordance with the DOE Public Access Plan (https://energy
.gov/downloads/doe-public-access-plan).

TG also thanks the Oak Ridge Institute for Science and Education (ORISE) for sponsorship
for the Advanced Short-Term Research Opportunity (ASTRO) program at the Oak Ridge
National Laboratory (ORNL). The ASTRO program is administered by the Oak Ridge Institute
for Science and Education (ORISE) for the U.S. Department of Energy. TG also acknowledges
the Visvesvaraya PhD fellowship award VISPHD-MEITY-2518 from Digital India Corporation
under Ministry of Electronics and Information Technology, Government of India
.

INFORMATIONS SUR LE FINANCEMENT

TG was sponsored by ORNL under the ORISE ASTRO Internship Program.

DATA AVAILABILITY

We release all the codes related to our experiment at https://figshare.com/s/2388c54ba01d2df25f38.

RÉFÉRENCES

Aljuaid, H., Iftikhar, R., Ahmad, S., Asif, M., & Afzal, M.. T. (2021).
Important citation identification using sentiment analysis of in-text
citations. Telematics and Informatics, 56, 101492. https://doi.org
/10.1016/j.tele.2020.101492

Allan, J., Wade, C., & Bolivar, UN. (2003). Retrieval and novelty
detection at the sentence level. Proceedings of the 26th Annual
International ACM SIGIR Conference on Research and Develop-
ment in Information Retrieval (pp. 314–321). https://est ce que je.org/10
.1145/860435.860493

Álvarez, M.. H., Soriano, J.. M.. G., & Martínez-Barco, P.. (2017). Citation
fonction, polarity and influence classification. Natural Language
E ng i ne er in g, 23 ( 4 ) , 5 6 1 –58 8. h t t p s : / / d o i . o rg / 1 0 . 1 0 1 7
/S1351324916000346

Amjad, Z., & Ihsan, je. (2020). VerbNet based citation sentiment
class assignment using machine learning. Revue internationale de
Advanced Computer Science and Applications, 11(9), 621–627.
https://doi.org/10.14569/IJACSA.2020.0110973

Athar, UN. (2011). Sentiment analysis of citations using sentence
structure-based features. Proceedings of the ACL 2011 Student
Session (pp. 81–87).

Bai, X., Xia, F., Lee, JE., Zhang, J., & Ning, Z. (2016). Identifying
anomalous citations for objective evaluation of scholarly article
impact. PLOS ONE, 11(9), e0162364. https://doi.org/10.1371
/journal.pone.0162364, PubMed: 27606817

Baldi, P., Sadowski, P., & Whiteson, D. (2014). Searching for exotic
particles in high-energy physics with deep learning. Nature
Communications, 5(1), 1–9. https://doi.org/10.1038/ncomms5308,
PubMed: 24986233

Bartneck, C., & Kokkelmans, S. (2011). Detecting h-index manipula-
tion through selfcitation analysis. Scientometrics, 87(1), 85–98.
https://doi.org/10.1007/s11192-010-0306-5, PubMed: 21472020
Bottou, L. (2010). Large-scale machine learning with stochastic gra-
dient descent. Proceedings of Compstat 2010 (pp. 177–186).
Springer. https://doi.org/10.1007/978-3-7908-2604-3_16

Camacho-Miñano, M., & Núñez-Nickel, M.. (2009). The multilay-
ered nature of reference selection. Journal of the American Society
for Information Science and Technology, 60(4), 754–777. https://
doi.org/10.1002/asi.21018

Campos, R., Mangaravite, V., Pasquali, UN., Jorge, UN. M., Nunes, C.,
& Jatowt, UN. (2018). YAKE! Collection-independent automatic

Études scientifiques quantitatives

1525

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

keyword extractor. European Conference on Information
Retrieval (pp. 806–810). https://doi.org/10.1007/978-3-319
-76941-7_80

Cerdá, J.. H. C., Nieto, E. M., & Campos, M.. L. (2009). What’s wrong
with citation counts? D-Lib Magazine, 15(3/4). https://est ce que je.org/10
.1045/march2009-canos

Chae, J., Schuman, C. D., Jeune, S. R., Johnston, J.. T., Rose, D. C.,
… Potok, T. E. (2019). Visualization system for evolutionary
neural networks for deep learning. 2019 IEEE International Con-
ference on Big Data (pp. 4498–4502). https://doi.org/10.1109
/BigData47090.2019.9006470

Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P..
(2002). SMOTE: synthetic minority over-sampling technique.
Journal of Artificial Intelligence Research, 16, 321–357. https://
doi.org/10.1613/jair.953

Ciregan, D., Meier, U., & Schmidhuber, J.. (2012). Multi-column deep
neural networks for image classification. 2012 IEEE Conference on
Computer Vision and Pattern Recognition (pp. 3642–3649).
https://doi.org/10.1109/CVPR.2012.6248110

Cohan, UN., Ammar, W., van Zuylen, M., & Cady, F. (2019). Struc-
tural scaffolds for citation intent classification in scientific publi-
cations. In J. Burstein, C. Doran, & T. Solorio (Éd.), Procédure
of the 2019 Conference of the North American Chapter of the
Association for Computational Linguistics: Human Language
Technologies (pp. 3586–3596). Association for Computational
Linguistics. https://doi.org/10.18653/v1/n19-1361

Colomo-Palacios, R., Tsai, F. S., & Chan, K. L. (2010). Redundancy
and novelty mining in the business blogosphere. The Learning
Organization.

Dong, C., & Schäfer, U. (2011). Ensemble-style self-training on
citation classification. Fifth International Joint Conference on
Natural Language Processing (pp. 623–631). The Association
for Computer Linguistics. https://www.aclweb.org/anthology/I11
-1070/.

Ghosal, T., Edithal, V., Ekbal, UN., Bhattacharyya, P., Chivukula,
S. S. S. K., & Tsatsaronis, G. (2020). Is your document novel? Let
attention guide you. An attention-based model for document-
level novelty detection. Natural Language Engineering, 1–28.
https://doi.org/10.1017/S1351324920000194

Ghosal, T., Edithal, V., Ekbal, UN., Bhattacharyya, P., Tsatsaronis, G.,
& Chivukula, S. S. S. K. (2018un). Novelty goes deep. A deep neu-
ral solution to document level novelty detection. Proceedings of
the 27th International Conference on Computational Linguistics
(pp. 2802–2813).

Ghosal, T., Salam, UN., Tiwari, S., Ekbal, UN., & Bhattacharyya, P..
(2018b). TAP-DLND 1.0: A corpus for document level novelty
detection. arXiv, arXiv:1802.06950.

Ghosal, T., Shukla, UN., Ekbal, UN., & Bhattacharyya, P.. (2019un). À
comprehend the new: On measuring the freshness of a docu-
ment. 2019 International Joint Conference on Neural Networks
(pp. 1–8). https://doi.org/10.1109/IJCNN.2019.8851857

Ghosal, T., Sonam, R., Ekbal, UN., Saha, S., & Bhattacharyya, P..
(2019b). Is the paper within scope? Are you fishing in the right
pond? En M. Bonn, D. Wu, J.. S. Downie, & UN. Martaus (Éd.), 19ème
ACM/IEEE Joint Conference on Digital Libraries (pp. 237–240).
IEEE. https://doi.org/10.1109/JCDL.2019.00040

Gilbert, C., & Hutto, E. (2014). VADER: A parsimonious rule-based
model for sentiment analysis of social media text. 8th Interna-
tional Conference on Weblogs and Social Media (Vol. 81, p. 82).
https://comp.social.gatech.edu/papers/icwsm14.vader.hutto.pdf.
Huang, G., Guo, C., Kusner, M.. J., Sun, Y., Sha, F., & Weinberger,
K. Q. (2016). Supervised word mover’s distance. Advances in
Neural Information Processing Systems (pp. 4862–4870).

Ihsan, JE., Imran, S., Ahmed, O., & Qadir, M.. UN. (2019). UN
corpus-based study of reporting verbs in citation texts using
natural language processing: Study of reporting verbs in citation
texts using natural language processing. CORPORUM: Journal de
Corpus Linguistics, 2(1), 25–36.

Ji, C., Tang, Y., & Chen, G. (2019). Analyzing the influence of
academic papers based on improved PageRank. In E. Popescu,
T. Hao, T. Hsu, H. Xie, M.. Temperini, & W. Chen (Éd.), Emerging
Technologies for Education – 4th International Symposium
(Vol. 11984, pp. 214–225). Springer. https://doi.org/10.1007/978
-3-030-38778-5_24

Jia, Y., Shelhamer, E., Donahue, J., Karayev, S., Long, J., … Darrell,
T. (2014). Caffe: Convolutional architecture for fast feature
embedding. Proceedings of the 22nd ACM International Confer-
ence on Multimedia (pp. 675–678). https://doi.org/10.1145
/2647868.2654889

Johnston, J.. T., Jeune, S. R., Schuman, C. D., Chae, J., Mars,
D. D., … Potok, T. E. (2019). Fine-grained exploitation of mixed
precision for faster CNN training. 2019 IEEE/ACM Workshop on
Machine Learning in High Performance Computing Environments
(pp. 9–18). https://doi.org/10.1109/MLHPC49564.2019.00007
Johnston, T., Jeune, S. R., Hughes, D., Patton, R.. M., & Blanc, D.
(2017). Optimizing convolutional neural networks for cloud
detection. Proceedings of Machine Learning on HPC Environ-
ments (pp. 1–9). https://doi.org/10.1145/3146347.3146352

Karkali, M., Rousseau, F., Ntoulas, UN., & Vazirgiannis, M.. (2013).
Efficient online novelty detection in news streams. International
Conference on Web Information Systems Engineering (pp. 57–71).
https://doi.org/10.1007/978-3-642-41230-1_5

Kusner, M., Sun, Y., Kolkin, N., & Weinberger, K. (2015). Depuis
word embeddings to document distances. International Confer-
ence on Machine Learning (pp. 957–966).

Laloë, F., & Mosseri, R.. (2009). Bibliometric evaluation of individual
chercheurs: Not even right… not even wrong! Europhysics News,
40(5), 26–29. https://doi.org/10.1051/epn/2009704

Li, X., & Croft, W. B. (2005). Novelty detection based on sentence
level patterns. Proceedings of the 14th ACM International
Conference on Information and Knowledge Management
(pp. 744–751). https://doi.org/10.1145/1099554.1099734

Lopez, P.. (2009). GROBID: Combining automatic bibliographic data
recognition and term extraction for scholarship publications. Inter-
national Conference on Theory and Practice of Digital Libraries
(pp. 473–474). https://doi.org/10.1007/978-3-642-04346-8_62
Lucchi, UN., Márquez-Neila, P., Becker, C., Li, Y., Forgeron, K., … Fua, P..
(2014). Learning structured models for segmentation of 2-D and
3-D imagery. IEEE Transactions on Medical Imaging, 34(5),
1096–1110. https://doi.org/10.1109/ TMI.2014.2376274,
PubMed: 25438309

Manju, G., Kavitha, V., & Geetha, T. V. (2017). Influential
researcher identification in academic network using rough set
based selection of time-weighted academic and social network
features. International Journal of Intelligent Information Technol-
ogies, 13(1), 1–25. https://doi.org/10.4018/IJIIT.2017010101
McCann, B., Bradbury, J., Xiong, C., & Socher, R.. (2017). Learned
in translation: Contextualized word vectors. arXiv, arXiv:1708
.00107.

Nazir, S., Asif, M., & Ahmad, S. (2020un). Important citation identi-
fication by exploiting the optimal in-text citation frequency.
2020 International Conference on Engineering and Emerging
Technologies (pp. 1–6). https://doi.org/10.1109/ ICEET48479
.2020.9048224

Nazir, S., Asif, M., Ahmad, S., Bukhari, F., Afzal, M.. T., & Aljuaid,
H. (2020b). Important citation identification by exploiting

Études scientifiques quantitatives

1526

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

content and section-wise in-text citation count. PLOS ONE, 15(3),
e0228885. https://doi.org/10.1371/journal.pone.0228885,
PubMed: 32134940

Noh, H., Hong, S., & Han, B. (2015). Learning deconvolution net-
work for semantic segmentation. Proceedings of the IEEE Interna-
tional Conference on Computer Vision (pp. 1520–1528). https://
doi.org/10.1109/ICCV.2015.178

Patton, R.. M., Johnston, J.. T., Jeune, S. R., Schuman, C. D., Mars,
D. D., … Kalinin, S. V. (2018). 167-PFlops deep learning for elec-
tron microscopy: From learning physics to atomic manipulation.
SC18: International Conference for High Performance Computing,
Networking, Storage and Analysis (pp. 638–648). https://doi.org
/10.1109/SC.2018.00053

Patton, R.. M., Johnston, J.. T., Jeune, S. R., Schuman, C. D., Potok,
T. E., … Saltz, J.. (2019). Exascale deep learning to accelerate
cancer research. 2019 IEEE International Conference on Big Data
(pp. 1488–1496). https://doi.org/10.1109/ BigData47090.2019
.9006467

Perier-Camby, J., Bertin, M., Atanassova, JE., & Armetta, F. (2019). UN
preliminary study to compare deep learning with rule-based
approaches for citation classification. 8th International Workshop
on Bibliometric-Enhanced Information Retrieval (BIR) co-located
with the 41st European Conference on Information Retrieval
(pp. 125–131).

Peters, M.. E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., …
Zettlemoyer, L. (2018). Deep contextualized word representa-
tion. arXiv, arXiv:1802.05365. https://doi.org/10.18653/v1/N18
-1202

Pileggi, S. F. (2018). Looking deeper into academic citations
through network analysis: popularity, influence and impact.
Universal Access in the Information Society, 17(3), 541–548.
https://doi.org/10.1007/s10209-017-0565-5

Pride, D., & Knoth, P.. (2017un). Incidental or influential? A decade of
using text-mining for citation function classification. In J. Qiu, R..
Rousseau, C. R.. Sugimoto, & F. Xin (Éd.), Proceedings of the 16th
International Conference on Scientometrics and Informetrics
(pp. 1357–1367). ISSI Society.

Pride, D., & Knoth, P.. (2017b). Incidental or influential? Défis
in automatically detecting citation importance using publication
full texts. In J. Kamps, G. Tsakonas, Oui. Manolopoulos, L. S. Iliadis,
& je. Karydis (Éd.), Research and Advanced Technology for
Digital Libraries—21st International Conference on Theory and
Practice of Digital Libraries (pp. 572–578). Springer. https://est ce que je
.org/10.1007/978-3-319-67008-9_48

Pride, D., & Knoth, P.. (2020). An authoritative approach to citation
classification. In R. Huang, D. Wu, G. Marchionini, D. Il, S. J..
Cunningham, & P.. Hansen (Éd.), JCDL ’20: Actes du
ACM/ IEEE Joint Conference on Digital Libraries in 2020
(pp. 337–340). ACM. https://doi.org/10.1145/3383583.3398617
Qayyum, F., & Afzal, M.. T. (2019). Identification of important
citations by exploiting research articles’ metadata and cue-terms
from content. Scientometrics, 118(1), 21–43. https://est ce que je.org/10
.1007/s11192-018-2961-x

Ronda-Pupo, G. UN., & Pham, T. (2018). The evolutions of the rich
get richer and the fit get richer phenomena in scholarly networks:
The case of the strategic management journal. Scientometrics,
116(1), 363–383. https://doi.org/10.1007/s11192-018-2761-3
Rousseau, R.. (2007). The influence of missing publications on the
Hirsch index. Journal of Informetrics, 1(1), 2–7. https://est ce que je.org/10
.1016/j.joi.2006.05.001

Saltz, J., Gupta, R., Hou, L., Kurc, T., Singh, P., … Thorsson, V.
(2018). Spatial organization and molecular correlation of
tumor-infiltrating lymphocytes using deep learning on pathology

images. Cell Reports, 23(1), 181–193. https://est ce que je.org/10.1016/j
.celrep.2018.03.086, PubMed: 29617659

Schiffman, B., & McKeown, K. (2005). Context and learning in
novelty detection. Proceedings of Human Language Technology
Conference and Conference on Empirical Methods in Natural
Language Processing (pp. 716–723). https://doi.org/10.3115
/1220575.1220665

Shen, J., Song, Z., Li, S., Tan, Z., Mao, Y., … Wang, X. (2016).
Modeling topic-level academic influence in scientific literatures.
En M. Khabsa, C. L. Giles, & UN. D. Wade (Éd.), Scholarly Big
Données: AI Perspectives, Défis, and Ideas—2016 AAAI
Workshop. AAAI Press. https://www.aaai.org/ocs/index.php/ WS
/AAAIW16/paper/view/12598.

Shi, C., Wang, H., Chen, B., Liu, Y., & Zhou, Z. (2019). Visual analysis
of citation context-based article influence ranking. IEEE Access, 7,
113853–113866. https://doi.org/10.1109/ACCESS.2019.2932051
Soboroff, JE., & Harman, D. (2003). Overview of the TREC 2003

novelty track. In TREC (pp. 38–53).

Soboroff, JE., & Harman, D. (2005). Novelty detection: The TREC expe-
rience. Proceedings of Human Language Technology Conference
and Conference on Empirical Methods in Natural Language Pro-
cessation (pp. 105–112). https://doi.org/10.3115/1220575.1220589
Tang, W., Tsai, F. S., & Chen, L. (2010). Blended metrics for novel
sentence mining. Expert Systems with Applications, 37(7),
5172–5177. https://doi.org/10.1016/j.eswa.2009.12.075

Teufel, S., Siddharthan, UN., & Tidhar, D. (2006). Automatic classifi-
cation of citation function. In D. Jurafsky & É. Gaussier (Éd.),
Actes du 2006 Conference on Empirical Methods in
Natural Language Processing (pp. 103–110). ACL. https://www
.aclweb.org/anthology/ W06-1613/. https://doi.org/10.3115
/1610075.1610091

Thorsson, V., Gibbs, D. L., Brun, S. D., Loup, D., Bortone, D. S., …
Shmulevich, je. (2018). The immune landscape of cancer. Immu-
nity, 48(4), 812–830. https://doi.org/10.1016/j.immuni.2018.03
.023, PubMed: 29628290

Valenzuela, M., Ha, V., & Etzioni, Ô. (2015). Identifying meaningful
citations. In C. Caragea et al. (Éd.), Scholarly Big Data: AI
Perspectives, Défis, and Ideas. AAAI Press. https://aaai.org
/ocs/index.php/ WS/AAAIW15/paper/view/10185.

Van Noorden, R.. (2017). The science that’s never been cited.

Nature, 552. https://doi.org/10.1038/d41586-017-08404-0

Van Noorden, R., & Singh Chawla, D. (2019). Hundreds of extreme
self-citing scientists revealed in new database. Nature, 572(7771),
578–579. https://doi.org/10.1038/d41586-019-02479-7,
PubMed: 31455906

Vîiu, G.-A. (2016). A theoretical evaluation of Hirsch-type bibliometric
indicators confronted with extreme self-citation. Journal of Infor-
metrics, 10(2), 552–566. https://doi.org/10.1016/j.joi.2016.04.010
Wang, F., Jia, C., Liu, J., & Liu, J.. (2019). Dynamic assessment of the
academic influence of scientific literature from the perspective of
altmetrics. In G. Catalano, C. Daraio, M.. Gregori, H. F. Moed, &
G. Ruocco (Éd.), Proceedings of the 17th International Confer-
ence on Scientometrics and Informetrics (pp. 2528–2529). ISSI.
Wang, M., Zhang, J., Jiao, S., Zhang, X., Zhu, N., & Chen, G.
(2020). Important citation identification by exploiting the syntactic
and contextual information of citations. Scientometrics, 125(3),
2109–2129. https://doi.org/10.1007/s11192-020-03677-1

West, R., Stenius, K., & Kettunen, T. (2017). Use and abuse of cita-
tion. Addiction science: A guide for the perplexed, 191. https://
doi.org/10.5334/bbd.j

Wilhite, UN. W., & Fong, E. UN. (2012). Coercive citation in academic
édition. Science, 335(6068), 542–543. https://est ce que je.org/10
.1126/science.1212540, PubMed: 22301307

Études scientifiques quantitatives

1527

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
q
s
s
/
un
r
t
je
c
e
–
p
d

F
/

2
4
1
5
1
1
2
0
3
4
4
8
1
q
s
s
_
un
_
0
0
1
7
0
p
d

b
oui
g
toi
e
s
t

o
n
0
9
S
e
p
e
m
b
e
r
2
0
2
3

Towards establishing a research lineage via identification of significant citations

Xie, Y., Sun, Y., & Shen, L. (2016). Predicating paper influence in
academic network. 20th IEEE International Conference on Com-
puter Supported Cooperative Work in Design (pp. 539–544).
IEEE. https://doi.org/10.1109/CSCWD.2016.7566047

Jeune, S. R., Rose, D. C., Johnston, T., Heller, W. T., Karnowski,
T. P., … Miller, J.. (2017). Evolving deep networks using HPC. Pro-
ceedings of the Machine Learning on HPC Environments (pp. 1–7).
https://doi.org/10.1145/3146347.3146355

Jeune, S. R., Rose, D. C., Karnowski, T. P., Lim, S.-H., & Patton,
R.. M.. (2015). Optimizing deep learning hyper-parameters
through an evolutionary algorithm. Proceedings of the Workshop
on Machine Learning in High-Performance Computing Environ-
ments (pp. 1–5). https://doi.org/10.1145/2834892.2834896
Zhang, F., & Wu, S. (2020). Predicting future influence of papers,
chercheurs, and venues in a dynamic academic network. Journal
of Informetrics, 14(2), 101035. https://doi.org/10.1016/j.joi.2020
.101035

Zhang, F., Zheng, K., Yuan, N. J., Xie, X., Chen, E., & Zhou, X.
(2015). A novelty-seeking based dining recommender system.

Proceedings of the 24th International Conference on World Wide
Web (pp. 1362–1372). https://doi.org/10.1145/2736277.2741095
Zhang, Y., Callan, J., & Minka, T. (2002). Novelty and redundancy
detection in adaptive filtering. Proceedings of the 25th Annual
International ACM SIGIR Conference on Research and Develop-
ment in Information Retrieval (pp. 81–88). https://est ce que je.org/10
.1145/564376.564393

Zhao, F., Zhang, Y., Lu, J., & Shai, Ô. (2019). Measuring academic influ-
ence using heterogeneous author-citation networks. Scientometrics,
118(3), 1119–1140. https://doi.org/10.1007/s11192-019-03010-5
Zhao, P., & Lee, D. L. (2016). How much novelty is relevant? Il
depends on your curiosity. Proceedings of the 39th International
ACM SIGIR Conference on Research and Development in Infor-
mation Retrieval (pp. 315–324). https://doi.org/10.1145/2911451
.2911488

Zhu, X., Turney, P.. D., Lemire, D., & Vellino, UN. (2015). Measuring
academic influence: Not all citations are equal. Journal of the
Association for Information Science and Technology, 66(2),
408–427. https://doi.org/10.1002/asi.23179