Content uploaded by Ringo Baumann
Author content
All content in this area was uploaded by Ringo Baumann on Jun 23, 2020
Content may be subject to copyright.
Datenbank-Spektrum manuscript No.
(will be inserted by the editor)
The Road Map to FAME: A Framework for Mining and
Formal Evaluation of Arguments
Ringo Baumann* ·Gregor Wiedemann ·Maximilian Heinrich ·
Ahmad Dawar Hakimi ·Gerhard Heyer
Received: date / Accepted: date
Abstract Two different perspectives on argumenta-
tion have been pursued in computer science research,
namely approaches of argument mining in natural lan-
guage processing on the one hand, and formal argument
evaluation on the other hand. So far these research ar-
eas are largely independent and unrelated. This article
introduces the agenda of our recently started project
“FAME – A framework for argument mining and evalu-
ation”. The main project idea is to link the two perspec-
tives on argumentation and their respective research
agendas by employing controlled natural language as a
convenient form of intermediate knowledge representa-
tion. Our goal is to develop a framework which inte-
grates argument mining and formal argument evalua-
tion to study patterns of empirical argumentation us-
age. If successful, this combination will allow for new
types of queries to be answered by argumentation re-
trieval systems and large-scale content analysis. More-
over, feeding evaluation results as additional knowledge
input to argument mining processes could be utilized to
further improve their results.
Keywords Argument mining ·Argument evaluation ·
Controlled natural Languages
1 Two Worlds of Argumentation
The study of argumentation has attracted increasing
attention by computer scientists over the past decade.
Dr. Ringo Baumann*
Leipzig University, Augustusplatz 10, 04109 Leipzig
E-mail: baumann@informatik.uni-leipzig.de
Dr. Gregor Wiedemann
Hamburg University, Vogt-K¨olln-Str. 30, 22527 Hamburg
E-mail: gwiedemann@informatik.uni-hamburg.de
Two major strands can be distinguished. On the one
hand, argument mining as a sub-field of natural lan-
guage processing (NLP) strives to identify, classify and
relate argumentative structures in texts. On the other
hand, researchers from theoretical computer science in-
vestigate how to model argumentation with symbolic
knowledge representations that can be logically evalu-
ated with a formalism. While the former profit from the
availability of large amounts of empirical data thanks
to the digitization of communication in the internet
era, they struggle to incorporate more complex, sym-
bolic knowledge into their works. The latter, while hav-
ing developed expressive models and tools for logical
representation and reasoning of formal argumentation,
largely lack empirical grounding or application of their
works. In this light, a large mutual benefit could be ex-
pected from a combination of both research strands.1
However, until recently the two research communities
remained surprisingly disconnected, mutually neglect-
ing their publications and rarely meeting for joint events.
With our research project “FAME – A framework for
argument mining and evaluation”, we strive to bridge
the gap between those two worlds. Employing controlled
natural languages (CNLs) as an intermediate represen-
tation for argumentation, we will investigate ways to
transform the empirical use of arguments into a machine-
evaluable form and explore the potentials of automatic
logical reasoning for the analysis of argumentation at
large scale.
This article introduces to the fundamental ideas, se-
lected technologies and targeted goals of our recently
1This is also a major hypothesis underlying the Prior-
ity Programme “Robust Argumentation Machines” (RATIO,
SPP 1999) funded by the German Research Foundation
(DFG) since 2017.
2 Ringo Baumann* et al.
started project.2For this, the upcoming section reflects
on what we consider as an argument against the back-
ground of the two research fields involved. The third
section introduces to formal argument evaluation. In
the fourth section, our approach to using CNL as inter-
mediate representation is described along with illustra-
tive examples. Finally, we explain the planned architec-
ture and give an outlook on the expected outcomes of
our project.
2 Argumentation in NLP
For a couple of years now, the study of argumentation
in natural language attracts lots of attention from re-
searchers. In a seminal paper by Palau and Moens [15],
argument mining is introduced as a number of consec-
utive NLP tasks for automatic extraction of argument
structures from unstructured documents such as news-
paper texts, blogs, user comments, etc. The chain of
tasks comprises 1) argument unit segmentation of the
input text, 2) classification of units into functional types
such as premise or claim, and 3) the classification of
structures between them, e.g. whether they support or
attack each other or are pro/contra regarding a cer-
tain topic. Single steps usually rely on supervised ma-
chine learning in which a classification model is trained
based on a manually annotated corpus. Corpora are
annotated with different annotation schemes of vary-
ing complexity. Simpler approaches are claim detection
[14] or stance classification [1] where pro/con positions
towards certain issues are classified. More complex ap-
proaches try to adapt theoretically derived annotation
schemes such as the Toulmin model [17] or Walton
schemes [21]. In these studies, the complexity of annota-
tion schemes usually is reduced by synthesizing subsets
of the theoretically defined categories [16].
During the conceptualization of our project, we iden-
tified two major gaps in the current research on NLP-
based argument mining. First, the field lacks a common
definition of its subject and suffers from simplification
of its operationalization. Although most works acknowl-
edge a semantic core of an argument, i.e. that a premise
is required to underpin the plausibility of a claim [18],
supervised learning to detect such functional types of
argument units usually boils down to statistical learn-
ing of superficial language cues. This is due to the fact
that high quality of training data for machine learning
requires high inter-annotator agreement for the argu-
ment constituents. Usually, this can only be achieved
by narrowing down either genre (e.g. student essays),
2The project started in summer 2019 and, due to the the-
matic fit, was given the opportunity to join the DFG Priority
Programme RATIO, SPP 1999.
topic (e.g. gay marriage), or argument type (e.g. study
evidence) for the manual annotation task. Still, anno-
tation is costly such that typical argument mining cor-
pora comprise only some hundred to a few thousand
texts. Any model simply striving for the detection of
functional types based on manually labelled resources
will eventually overemphasize corpus-specific language
patterns such as topic words or discourse markers [9],
and neglect semantic features as well as their logical
dependencies. Consequently, generalization of trained
models to new datasets is poor, as can be tested with
tool Targer by Chernodub et al. [7]. Second, since ar-
gument mining often boils down to automatic classifi-
cation of functional argument unit types based on lin-
guistic patterns, it refrains from the incorporation of
any semantic knowledge. Unfortunately, this impedes
not only the use of structured background knowledge,
semantic priors or logical constraints which could pro-
vide helpful information to the mining process. It also
prevents interesting use cases of automated evaluation
of empirical argumentation.
In light of these shortcomings, we argue that a prac-
tically useful argument mining system requires a more
expressive encoding of arguments based on semantics.
This would also allow modeling implicit background as-
sumptions which often cannot be observed from empir-
ical arguments directly, but can be encoded in knowl-
edge bases (KB) curated by domain experts. In our
project, we explore ways and opportunities for using
such a KB for argument mining and formal evaluation.
3 Formal Argument Evaluation
Computational models of argumentation have become
one central topic in leading AI conferences. Moreover,
since 2006, there is a biennial conference called Compu-
tational Models of Argument. The formal analysis of ar-
gumentation studies how to model arguments and their
relationships, as well as the necessary conflict resolution
in the presence of diverging opinions. One can distin-
guish two major lines of research in the field: logic-based
and abstract approaches. The former takes the logical
structure of arguments into account and defines notions
like attack, undercut, defensibility etc. in terms of log-
ical properties of the chosen argument structures (cf.
[4,5] for excellent overviews). For instance, in case of
propositional logic, an argument is a premise/claim pair
A= (Φ, φ) where Φis a set of formulas and φa single
propositional formula, s.t. i) Φ|=φ, ii) Φis consis-
tent and iii) Φis ⊆-minimal w.r.t. i) and ii). The pairs
({p, ¬q→ ¬p}, q) and ({(s∨q)→t, ¬t},¬s∧ ¬q)) are
examples of logical arguments. Moreover, we may say
The Road Map to FAME: A Framework for Mining and Formal Evaluation of Arguments 3
b
a
F:
d
c
Fig. 1 Example of a simple argumentation framework
that both arguments attack each other since the union
of their claims, namely {q, ¬s∧ ¬q}is inconsistent.
Abstract approaches, in contrast, abstract away from
the internal structure of arguments and consider them
as atomic items, focusing entirely on the attack rela-
tion among arguments. This means it is assumed that
the reason why something is an argument was already
identified beforehand. Such a reason can be an explicit
construction from a given background KB as sketched
above or simply an argument mining process applied
to real-world data. This means the abstract approach
is not standalone, it rather depends on methods for
generating abstract arguments in the first place and a
subsequent instantiation process. The main aim of ab-
stract argumentation is to provide possibilities to evalu-
ate conflicting scenarios, i.e. to return reasonable sets of
arguments that represent acceptable positions one may
take in the light of the available arguments.
From AFs to ADFs: At the heart of the abstract ap-
proach are currently Dung’s widely used argumentation
frameworks (AFs) [8] and their associated semantics. In
a nutshell, an AF is a directed graph F= (A, R) with a
set of vertices Abeing the abstract arguments and a set
of directed edges R⊆A×Acorresponding to attacks
between arguments. Consider the AF Fin Fig. 1 con-
sisting of four arguments a,b,c, and d. Such a conflict-
ing scenario is resolved by so-called semantics. By now
a variety of argumentation semantics has been defined,
each one encoding different desiderata for acceptable
sets of arguments, so-called extensions (cf. [2, Section
3.1.2] for a compact summary). One of the most promi-
nent ones was already defined by Dung in 1995, so-
called stable semantics. Informally, a set of arguments
is a stable extension if there are no conflicts between
them and moreover, all other arguments are attacked
by at least one argument of the set. This means, in
case of Fwe obtain one single stable extension, namely
E={a, c}. The position Edoes not contain any conflict
and furthermore, all remaining arguments (i.e. band c)
are refuted. Obviously, such positions are very desirable
in a debate or argumentation scenarios in general.
Dung’s argumentation frameworks, as well as most
of the available semantics, are very intuitive and easy
to understand. However, they suffer from various short-
comings. One main drawback is that they are rather
limited in their expressive capabilities implying that
they are not necessarily the right target systems for in-
stantiation. More precisely, modeling the relations be-
tween arguments is problematic if these relations are
more complex than a simple binary attack between two
arguments. For this reason, Abstract Dialectical Frame-
works (ADFs) were introduced [6]. They are a natural
generalization of classical Dung-style AFs. The funda-
mental idea behind them is to stick to abstract argu-
ments which remain atomic entities and thus are not
further analyzed as in AFs, yet to allow for much more
flexible relationships among arguments. In particular,
arguments can not only attack each other, but they
also may provide support for other arguments which is
not expressible in classical AFs. This expressive power
is achieved by adding acceptance conditions to the ar-
guments which allow for the specification of arbitrary
relationships between arguments and their parents in
the argument graph. Acceptance conditions are usually
expressed in terms of propositional formulas. Conse-
quently, an ADF can be represented as a pair D=
(S, Φ) where Sis a set of statements and Φa set of ac-
ceptance functions. A semantics-preserving translation
from AFs to ADFs takes former arguments as state-
ments and formalizes the acceptance function of a cer-
tain statement aas the conjunction of all negated at-
tackers of a. This means a statement acan be ac-
cepted if and only if none of its attackers is accepted.
For the introduced AF Fwe obtain the ADF DF=
({a, b, c, d},{φa, φb, φc, φd}) where φa=>,φb=¬a∧
¬d,φc=¬band φd=¬c. We mention that the def-
initions for semantics in case of ADFs are more in-
volved than in case of AFs since they rely on differ-
ent (pre-)fixpoints of associated consequence operators
(cf. [20] for more information).
We already mentioned that ADFs may express more
than single attacks as in case of AFs. For instance, the
acceptance function φa=cencodes single support, i.e.
statement ashould be accepted, if statement cis. In
other words, the acceptance of cleads to the acceptance
of a. Another form of support is collective support. The
acceptance functions φa=b∧cis such an example.
It encodes that ashould be accepted if both band c
are accepted. This means the acceptance of only one of
them is not sufficient for supporting a.
Temporal aspects of argumentation: The classical
definition of ADFs does not provide one with temporal
notions. However, in daily life we are often faced with
statements/laws which are valid for a certain time only
or depend on the past development, e.g. “You can con-
tinue working in the company as long as the Brexit is
not delivered”, “From the beginning of next year it will
4 Ringo Baumann* et al.
be not allowed to build a nightclub near a residential
area” or “I will spend my holidays in France given that I
get a salary increase this year.” In order to encode such
expressions, we need to be able to distinguish between
different time states related via a certain ordering. We,
therefore, introduce so-called timed Abstract Dialectical
Frameworks (tADFs) [3]3which are powerful enough to
model many frequently occurring temporal restrictions.
More precisely, a timed Abstract Dialectical Framework
(tADF) will be a classical ADF equipped with a count-
able set Tof time states. Moreover, we assume that this
set is totally ordered, i.e. there is a binary relation ≤
over Twhich is antisymmetric, transitive and connex.
For simplicity, we may assume that Tis a subset of the
first natural numbers with the inherited standard or-
dering. Now, a certain time state nmight stand for an
hour, a day, a month or whatever granularity is needed.
In doing so we are able to speak about the same state-
ment sat different time points tin the future, denoted
as st. Accordingly, we will have timed acceptance con-
ditions φstfor any statement sat any time point t. For
instance, the condition φs5=a1∨a2∨a3∨a4encodes
a support of sat time point 5 via the statement afor
any time point between 1 and 4. If the numbers are in-
terpreted as the first months of the year and if sand
aare standing for “I am on vacation in France” or “I
have a salary increase”, respectively, then φs5expresses
the statement “I will be vacationing in France in May
if I get a salary increase between January and April.”
4 Controlled Natural Languages
As we have seen, computational models of argumenta-
tion provide us with a set of tools for logical reason-
ing about abstract arguments. However, for any use-
ful empirical application, we need instantiations of ab-
stract arguments with real-world arguments as com-
monly expressed and understood in human communi-
cation. Unfortunately, natural language argumentation
comprises a large variety to express semantically equiv-
alent statements in heterogeneous ways, and cannot be
parsed unambiguously by machines. To facilitate the
necessary linking between abstract and real-world argu-
ments, we suggest employing a controlled natural lan-
guage. In general, a CNL can be seen as a small subset
of a natural language [13]. This shrinking can happen
in order to increase comprehension or formal precision.
For our project, we rely on Attempto Controlled En-
glish (ACE), which is one of the most mature CNLs
[12]. ACE sentences are notated in a simplified English,
3Accepted for COMMA 2020 (see https://comma2020.
dmi.unipg.it/).
Fig. 2 Reasoning with RACE over an inconsistent set of
ACE statements
which allows a intuitive understanding by human read-
ers. At the same time, each sentence has exactly one
logical representation, which prevents ambiguity. For
more details see e.g. [12].
For reasoning with ACE two steps are performed.
First, a parser translates the ACE-sentences into a log-
ical representation. Then, a reasoner is applied to the
preprocessed data in order to get the actual logical in-
ferences. For the first step, the Attempto Parsing En-
gine (APE) is used, which processes the input into a dis-
course representation structure (DRS) [11]. The DRS
representation can then be further translated, e.g. into
a first-order representation, or into semantic web lan-
guages like OWL or SWRL, which enables the use of
several reasoners [10]. As Reasoner we use for the fol-
lowing examples the web interface of RACE (see Fig. 2).4
This sophisticated reasoner allows us to check consis-
tency, logical entailment and can be used for query an-
swering. RACE works on the level of logical formulas
and does not include the abstract argumentation ap-
proach, which we want to use for FAME. The idea of
the following examples is to illustrate, how an ACE
representation of natural language can be used for log-
ical reasoning and query answering. In addition, possi-
ble ideas on how the examples could be modeled with
the means of abstract-argumentation are provided. In
a mid-term perspective, it is planned that the abstract
argumentation representation is directly processed by
its own customized reasoner framework.
An illustrative example: Let us illustrate the ap-
plication of our approach on some simplified examples
about the introduction of a minimum wage. An argu-
ment a1that favors the introduction of a minimum
wage might sound like the following: ‘Minimum wage is
4http://attempto.ifi.uzh.ch/race/
The Road Map to FAME: A Framework for Mining and Formal Evaluation of Arguments 5
a baseline for salaries. Therefore, it defines a living stan-
dard and guarantees social security.’ Translated into the
ACE language this argument might sound like the fol-
lowing: ‘The minimum-wage is a baseline for all salaries.
A baseline for all salaries enables a living-standard and
guarantees some social-security.’ However, this ACE
sentence still generates an error during parsing because
it uses words unknown to APE. In general, there is
the possibility to specify a dictionary for APE, or, al-
ternatively, one can specify unknown words with pre-
fixes in RACE directly. For example ‘v:’ is used to de-
note verbs, and ‘n:’ is used to denote nouns, yielding
this modification: ‘The n:minimum-wage is a n:baseline
for all n:salaries. A n:baseline for all n:salaries
v:enables a n:living-standard and v:guarantees some
n:social-security.’ Such an argument might be confronted
with an argument a2expressing ‘Minimum wage is so
low, that it will not change anything.’. This can be
translated into proper ACE as: ‘The n:minimum-wage
is not a n:baseline for all n:salaries.’
To analyze the given arguments as well as to obtain
an abstract argument representation RACE offers sev-
eral possibilities. First, we may check consistency. If the
previous ACE statements are input to RACE, one gets
notified that the sentences are inconsistent (see Fig. 2).
This can be interpreted as the arguments a1and a2at-
tacking each other. Moroever, RACE shows a minimal
subset causing the inconsistency. This means, that al-
ready a proper part of a1, namely ‘Minimum wage is
a baseline for salaries’ (so to speak a subargument of
a1) is in conflict with a2. Moreover, it can be checked
that the sentence ‘A baseline for all salaries enables a
living-standard and guarantees some social-security.’ is
consistent with a2. Such information can be used for a
more fine-grained perspective on arguments. More pre-
cisely, the argument a1might be splitted into two sur-
barguments a1
1and a2
1and only the latter conflicts with
a2.
Let us consider another example where somebody
states the argument: ‘A minimum wage and therefore
higher salaries will only increase automatization or will
cost some jobs.’ The intention here is that the or in the
argument expresses that the person, who uses this argu-
ment is a bit unsure about the possible consequences of
a minimum wage. Therefore two possibilities are listed
which do not have to be true at the same time. Trans-
lated to ACE this argument might be represented as: ‘If
there is a n:minimum-wage then there are some n:higher-
salaries. Every n:higher-salaries v:lead to some n:auto-
matization or v:cost some jobs. There is a n:minimum-
wage.’ Note that here an extra sentence is added which
states the existence of a minimum wage in order to ini-
tiate the logical reasoning with the ‘if-then’ part of the
Issue specific
Argumen-
tative texts
News texts
News texts
Issue specific
News texts
1. Extraction
2. Mapping
6. Verbalization
Natural language Formal argumentation
CNL
Knowledge
Base
Abstract
Argumentation
5. Evaluation
4. Translation
3. Analysis
7. User-Interface
● Issue based queries:
○ Pro/con arguments
○ Documents with certain stances / arguments
● Document based queries:
○ Completion: implicit background assumptions
○ Refutation: assumptions to reject
○ Coherence: contradicting arguments
Fig. 3 FAME architecture
first sentence. From an abstract point of view this state-
ment can be modeled as an argument a3resembling
a minimum-wage, and the arguments a4and a5repre-
senting the arising of automatization or the loss of jobs,
respectively. One suitable way to derive a support rela-
tions between arguments is simply logical consequence,
i.e. n1supports n2, if the latter is a consequence of the
former. Thus, we may state that a3supports a4and a5
given that we include some background knowledge like
the general trend of technification in economy, which
reduces the need for manual labour.
RACE also offers the function query answering. Re-
garding our example we may ask ‘Are there some jobs?’.
RACE states that this question cannot be answered.
The reason for this is that it is not known whether the
first or the second disjunct are true. Let us assume that
a new argument arises saying that there is no automa-
tion due to a high machine tax. This means, we add the
sentence ‘There is no n:automatization.’. Now, RACE
states that the initial question can be answered and
lists also the minimal subset necessary for this. Con-
sequently, with the help of query answering we may
complete our KB and derive further support relations
on the abstract level. Moreover, we may detect missing
(implicit) background assumptions in order to answer
a question and provide this information to the user.
5 The Road Map to FAME
A central goal of the FAME project is to convert ar-
gumentation from empirical texts into a formal repre-
sentation. However, translating arguments from natural
language directly to ACE is an AI-complete problem
out of the scope of our project. Instead, we will sim-
plify this to a mapping-problem which can be solved
6 Ringo Baumann* et al.
in a supervised learning paradigm. On this basis, we
strive to answer two research questions in the course
of the project: (1) how can we effectively map argu-
ments in empirical text to known ACE statements in a
knowledge base, and (2) how can we derive an abstract
frameworks (e.g. AF, ADF or tADF) from subsets of
this knowledge base for automatic evaluation? Fig. 3
presents an overview of the major steps of the project
architecture. From the NLP perspective, we start with
collecting and extracting issue-specific samples of real-
world argumentation from existing resources such as
the UKP SAM corpus [19]. Semantically equivalent sets
of arguments are grouped together in a computer-aided
manner. For this, we employ semantic similarity search
technologies based on contextualized word embeddings
[22]. Table 1 shows an example of equivalent argument
sentences retrieved from the UKP SAM corpus and in
line with our previous minimum-wage example. Accord-
ingly, in step 3 corresponding ACE statements to the
retrieved arguments from step 1 are formulated. The
resulting KB then needs to be translated into a suit-
able abstract representation including attack and/or
support relations as well as timed arguments if neces-
sary (step 4). In Section 4, we sketched how such rela-
tion (with the help of RACE) can be found. Moreover,
in Section 3 we presented suitable formalisms for this
endeavour. Step 5 is enganged with determining accept-
able positions, i.e. acceptable subsets of arguments for a
given scenario w.r.t. an argumentation semantics. The
main aim of the verbalization step 6 is the retransla-
tion in CNL of the evaluation outcome extending the
KB with exactly this information. What do we gain
from this? A major goal from the NLP perspective is
to develop a precise enough mapping from arguments
in empirical text collections such as newspaper articles
to arguments in our KB such that we can support vari-
ous argument analysis and retrieval tasks (7). Potential
queries could be ‘Is the argumentation in a given doc-
ument consistent?’, or ‘What implicit background as-
sumptions does the author assume given the presented
arguments in his/her document?’. We believe such a
system will allow for many exciting possibilities for do-
main experts to analyze not only single documents but
also argumentation patterns in large document collec-
tions. We look forward to sharing our insights, technolo-
gies resources and with the two argument communities
and bringing them more closely together in the near
future.
Acknowledgements We thank DFG (Deutsche Forschungs-
gemeinschaft) for funding this work (project no. 406289255,
funding period: 05/2019–04/2022), and the coordinators of
the SPP 1999 for our subsequent admission to the Priority
Programme.
References
1. Bar-Haim, R., Bhattacharya, I., Dinuzzo, F., Saha, A.,
Slonim, N.: Stance classification of context-dependent
claims. In: Proc. ACL, pp. 251–261. ACL, Valencia, Spain
(2017)
2. Baumann, R.: On the existence of characterization logics
and fundamental properties of argumentation semantics.
Habilitation treatise, Leipzig University (2019). DOI
10.13140/RG.2.2.21831.24483
3. Baumann, R., Heinrich, M.: Timed abstract dialectical
frameworks. In: Proc. of COMMA 2020, Perugia, Italy,
September 8-11, 2020, p. to appear (2020)
4. Besnard, P., Garcia, A., Hunter, A., Modgil, S., Prakken,
H., Simari, G., Toni, F.: Special issue: Tutorials on struc-
tured argumentation. Argument and Computation 5(1),
1–117 (2014). DOI 10.1080/19462166.2013.869764
5. Besnard, P., Hunter, A.: Elements of Argumentation.
MIT Press, Cambridge, USA (2008)
6. Brewka, G., Woltran, S.: Abstract dialectical frameworks.
In: Proc. KR, pp. 102–111 (2010)
7. Chernodub, A., Oliynyk, O., Heidenreich, P., Bon-
darenko, A., Hagen, M., Biemann, C., Panchenko, A.:
Targer: Neural argument mining at your fingertips. In:
Proc. ACL. Florence, Italy (2019)
8. Dung, P.M.: On the acceptability of arguments and its
fundamental role in nonmonotonic reasoning, logic pro-
gramming and n-person games. Artificial Intelligence
77(2), 321 – 357 (1995)
9. Eckle-Kohler, J., Kluge, R., Gurevych, I.: On the role of
discourse markers for discriminating claims and premises
in argumentative discourse. In: Proc. EMNLP, pp. 2236–
2242. ACL, Lisbon, Portugal (2015). DOI 10.18653/v1/
D15-1267
10. Fuchs, N.E.: First-order reasoning for Attempto Con-
trolled English. In: M. Rosner, N.E. Fuchs (eds.) Proc.
CNL, vol. 7175, pp. 73–94. Springer (2012)
11. Fuchs, N.E., Kaljurand, K., Kuhn, T.: Discourse Rep-
resentation Structures for ACE 6.6. Tech. Rep. ifi-
2010.0010, Department of Informatics, University of
Zurich (2010)
12. Kuhn, T.: Controlled english for knowledge representa-
tion. Ph.D. thesis, Faculty of Economics, Business Ad-
ministration and Information Technology of the Univer-
sity of Zurich (2010)
13. Kuhn, T.: A survey and classification of controlled natu-
ral languages. Computational Linguistics 40(1), 121–170
(2014). DOI 10.1162/COLI\a\00168
14. Levy, R., Gretz, S., Sznajder, B., Hummel, S., Aharonov,
R., Slonim, N.: Unsupervised corpus–wide claim detec-
tion. In: Proc. Workshop on Argument Mining, pp. 79–
84. ACL, Copenhagen, Denmark (2017). DOI 10.18653/
v1/W17-5110
15. Palau, R.M., Moens, M.F.: Argumentation mining: The
detection, classification and structure of arguments in
text. In: Proc. ICAIL, pp. 98–107. ACM, New York,
USA (2009). DOI 10.1145/1568234.1568246
16. Peldszus, A., Stede, M.: From argument diagrams to ar-
gumentation mining in texts: A survey. International
Journal of Cognitive Informatics and Natural Intelligence
7(1), 1–31 (2013). DOI 10.4018/jcini.2013010101
17. Reisert, P., Inoue, N., Okazaki, N., Inui, K.: A com-
putational approach for generating Toulmin model ar-
gumentation. In: Proc. Workshop on Argumentation
Mining, pp. 45–55. ACL, Denver, USA (2015). DOI
10.3115/v1/W15-0507
The Road Map to FAME: A Framework for Mining and Formal Evaluation of Arguments 7
Table 1 Example for mapping of ACE statements to natural language arguments
ACE If there is a n:minimum-wage then there are some n:higher-salaries. Every n:higher-salaries v:lead to some
n:automatization or v:cost some jobs.
NL
equivalent
If the minimum wage is increased, companies may use more robots and automated processes to replace service
employees.
Ordering businesses to pay entry-level workers more will make them hire fewer of them, and consider replacing
more workers with robots or computers.
After that, it doesnt encourage higher wages, and only encourages unemployment and automation.
18. Stab, C., Gurevych, I.: Identifying argumentative dis-
course structures in persuasive essays. In: Proc. EMNLP
2014, pp. 46–56 (2014)
19. Stab, C., Miller, T., Schiller, B., Rai, P., Gurevych,
I.: Cross-topic argument mining from heterogeneous
sources. In: Proc. EMNLP, pp. 3664–3674. ACL, Brus-
sels, Belgium (2018). DOI 10.18653/v1/D18-1402
20. Strass, H.: Approximating operators and semantics for
abstract dialectical frameworks. Artificial Intelligence
205, 39–70 (2013). DOI 10.1016/j.artint.2013.09.004
21. Walton, D., Reed, C., Macagno, F.: Argumentation
Schemes. Cambridge Univ. Press, Cambridge, USA
(2008)
22. Wiedemann, G., Remus, S., Chawla, A., Biemann, C.:
Does BERT make any sense? Interpretable word sense
disambiguation with contextualized embeddings. In:
Proc. KONVENS, pp. 161–170. Erlangen (2019)