Let the cross-entropy loss of the span classifier
be Ls, cross-entropy loss for the component clas-
sification be Lc, binary cross-entropy loss for the
link detection be L(cid:2), and cross-entropy loss for
the relation label classifier be Lr. The objective to
be optimized is L = λsLs + λcLc + λ(cid:2)L(cid:2) + λrLr,
where λ are hyperparameters.

4.3 MT-AM

We propose MT-AM, an end-to-end model us-
ing a two-staged method (multi-task pre-training
and target corpus fine-tuning), which is similar
to the work of Liu et al. (2019a). We modify
the model architecture of ST for the multi-task
pre-training because the output labels are different
for each corpus, and corpus-specific output layers
are implemented while still sharing Longformer
parameters. In the target corpus fine-tuning, we
further train the model on only the target corpus.
Generally speaking, the objective can be the
sum of losses for each corpus in the multi-task
pre-training, and MT-AM should capture transfer-
able features across different corpora. However,
each auxiliary corpus would still have distant
information, which degrades the parsing perfor-
mance for a target corpus. We therefore provide
a smaller loss weight (van der Goot et al., 2021)
(i.e., auxiliary weight) for the auxiliary corpora.
We also provide a different learning rate (LR)
and epochs for the multi-task pre-training and tar-
get corpus fine-tuning. These hyperparameters are
described later.

We provide the following variants of MT-AM:
MT-All, which conducts multi-task pre-training
with all five corpora. An example of MT-All for

Figure 3: Overview of MT-AM.

AAEC is shown in Figure 3a, where the five cor-
pora provide corpus-specific output layers in the
multi-task pre-training. In this case, the auxiliary
corpora are MTC, CDCP, AbstRCT, and AASD.
After a couple of epochs in this stage, we further
fine-tune on the target corpus (i.e., AAEC). Finally,
we evaluate the test data of the target corpus.
→ Corpusy, which uses a single aux-
Corpusx
iliary corpus (i.e., Corpusx) only for analysis
(see Sections 5.3 and 5.4). We use Corpusx and
the training data of Corpusy in the multi-task
pre-training. After a couple of epochs in this
stage, we further fine-tune on the target corpus
(i.e., Corpusy). An example of AAEC → CDCP is
shown in Figure 3b.

5 Experiments

We compared the MT-AM and ST models
(Section 5.2),
investigated the transferability
(Section 5.3), and conducted an in-depth analyses
under low-resource settings (Section 5.4).

5.1 Experimental Setup

Hyperparameters: We apply dropout (Srivastava
et al., 2014) to the MLP layers. We use the Adam
(Kingma and Ba, 2015) optimizer with a linear
warmup scheduler (Howard and Ruder, 2018).
We tune the LR and λ values as well as the
number of epochs in the multi-task pre-training
and the auxiliary weight by Optuna (Akiba et al.,
2019), a hyperparameter optimization framework.

645

l

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

p

:
/
/

d
i
r
e
c
t
.

m

i
t
.

e
d
u

/
t

a
c
l
/

l

a
r
t
i
c
e
–
p
d

f
/

d
o

i
/

.

1
0
1
1
6
2

/
t

l

a
c
_
a
_
0
0
4
8
1
2
0
2
2
9
6
5

/

/
t

l

a
c
_
a
_
0
0
4
8
1
p
d

.

f

b
y
g
u
e
s
t

t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

Five-fold cross-validation (CV) is applied to the
training data for the hyperparameter optimization.
MTC and AASD use CV in the experiments, so
instead of tuning the hyperparameters we use
the same hyperparameters as AAEC. To stabilize
the tuning, we fix the total number of training
epochs to 20, set λs = 1, and tune the other λ
values by sampling from a uniform distribution
within the range of [0.01, 2.0]. We first tune the
hyperparameters related to the ST model and then
those related to the multi-task pre-training (e.g.,
LR, auxiliary weight, and number of epochs). The
number of epochs in the multi-task pre-training
is tuned by sampling from 2, 4, 6, and 8 while
keeping the total number of epochs to 20.

The tuned and fixed hyperparameters are shown
in Table 3. We found that the LR in the multi-task
pre-training was generally lower than that of
the target corpus fine-tuning, suggesting that
multi-task pre-training degrades the target corpus
representation if we use a higher LR.

Implementation Details: After applying the
relation classifiers, we use Chu–Liu/Edmonds’
algorithm (Chu and Liu, 1965) for AAEC, MTC,
AbstRCT,5 and AASD to make the predicted links
(i.e., the score matrix produced by the link detec-
tion) into a tree. We use given training and test
data for AAEC, CDCP, and AbstRCT, examine
30 different seeds, report the average score, and
use ten sets of five-fold CV (Kuribayashi et al.,
2019) for MTC and AASD. Unless otherwise spec-
ified, we use the essay-level data for AAEC. The
development data are randomly sampled from the
training data except for AbstRCT. We use given
development data for AbstRCT. We select the
optimal model on the basis of the development
score by evaluating once every two epochs.

Evaluation Metrics: Let Gtask and Stask be a
set of gold and system outputs for a task, re-
spectively. Precision P = |Gtask ∩ Stask|/|Stask|,
recall R = |Gtask ∩ Stask|/|Gtask|, and F =
2P R/(P + R) are then defined. For example,
(cid:4)id, s, e(cid:5) ∈ Gspan, where id indicates an ID inher-
ent to a text. Similarly, (cid:4)id, s, e, c(cid:5) ∈ Gcomponent,
and (cid:4)id, ssrc, esrc, stgt, etgt, r(cid:5) ∈ Grelation. We also
introduce the link score (Kuribayashi et al., 2019),
which is used to measure F-scores for determining
the existence of relations regardless of their labels.

5This is done since most of the AbstRCT components

have only one outgoing relation (Mayer et al., 2020).

Table 4: F-scores [%]. MT-All (shown in bold)
generally outperformed the ST models. ‘‘OS’’
denotes oracle span setting.
repre-
sents macro-averaged F-score. * shows statistical
significance p < 0.05. ‘‘Macro’’ Note that the scores of component and relation classifications are affected by the performance of the span identification task. 5.2 Comparison of MT-All with ST Model 5.2.1 Effectiveness of MT-AM To determine the effectiveness of MT-All, we compared it against the ST model. Table 4 shows the overall results on the five AM corpora. We also show oracle span (OS) results, where the compo- nents and relations are predicted on gold spans, to determine the improvement for each component and relation classification. As we can see in the table, in most corpora, MT-All outperformed the ST model on average. The largest improvement by MT-All was observed for MTC and AASD in the component and relation classification. This would be because these two corpora are smaller than the others and MT-All benefited more from the aux- iliary corpora. However, the ST model performed better than MT-All for some tasks in AbstRCT and CDCP. In CDCP, labels that do not have cor- responding labels in other corpora, such as Policy and Testimony (see Table 1), would be more challenging to improve by multi-task learning. 646 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 5.2.2 MT-All under a Low-resource Setting The above results indicate that the parsing per- formance of MT-AM improves for low-resource corpora. Because AM corpora in real settings are generally low-resource, we further investigated the effectiveness of MT-AM under the following three extremely low-resource settings inspired by Schulz et al. (2018): 1. Sample n [%] texts of the training and development data in a target corpus. 2. For MT-AM, conduct multi-task pre-training with auxiliary corpora and the sampled data of the target corpus. For the target-corpus fine-tuning, use the sampled data. For ST, conduct fine-tuning on the sampled data. 3. Evaluate test data in the target corpus. Similar to the study by Liu et al. (2019a), we changed the sample amount n to 1%, 10%, or 100%. When n = 1%, the amount of sampled training data of AbstRCT was 3 (texts) and that of MTC and AASD was 1. To investigate the improvement of MT-AM for an ST model, we used the following metric (larger is better), similar to the study by Morishita et al. (2020): Error reduction [%] = (cid:3)(ST) − (cid:3)(MT-AM), where (cid:3) is a function that computes the error rate, that is, 100 − F-score [%]. However, it is more robust to use a degree of error reduction for errors produced in an ST model as a metric. For example, error reduction by MT-AM (95 F-score) for an ST model (90 F-score) is the same as that by MT-AM (55 F-score) for an ST model (50 F-score). Intuitively, improving the F-score from 90 to 95 is more challenging than improving it from 50 to 55. Thus, we define the error reduction rate (ERR; larger is better) as ERR [%] = 100 × Error reduction (cid:3)(ST) . Following the above setting, Figure 4 repre- sents the ERR by MT-All for an ST model with three different training data amounts.6 Overall, we found that when there was less training data in the target corpus, the ERR was larger. This suggests that MT-AM is especially effective in the low-resource setting, which is likely in line 6We compute ERRs for each CV or seed. Figure 4: ERR (with OS for components and relations) by MT-All for ST with different training data amounts (n%). We averaged the results of five corpora, re- moving outliers for visibility. Red line shows average values over three tasks. with the previous work (Schulz et al., 2018). On the other hand, when n = 100%, ERR for relation and component tasks was larger than that for the span identification task. However, when n = 1% or 10%, the ERR for the span identification task was larger than that for the other tasks. Given that the span identification task is less semantic7 than component and relation classifications, we conclude that the ERR of spans can be improved with less training data. In other words, it is diffi- cult for the span identification task to benefit from multi-task learning when the amount of training data is sufficient. In contrast, relation and compo- nent tasks are more challenging to train; therefore, the two tasks require more training data. 5.2.3 Evaluation of the ST Model To investigate whether our ST model (and hence the MT-AM model) performs reasonably, we compared it with the following state-of-the-art end-to-end AAEC parsers: • BLCC (Eger et al., 2017), which parses the argument structure as a sequence tagging task. • LSTM-ER (Eger et al., 2017), which uses LSTM-ER (Miwa and Bansal, 2016) (originally used to parse relations of entities). • BiPAM-syn (Ye and Teufel, 2021), a state-of-the-art parser based on a biaffine operation and syntactic information. In addition to the essay-level results, we pro- vide paragraph-level results. We also provide 7For example, some corpora annotate spans based on (semi-)automatically split segments for ease of annotation. 647 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 2017)-based AM parser that uses RoBERTa (Liu et al., 2019b) or SciBERT (Beltagy et al., 2019).9 All the scores are drawn from the original papers of the above baselines except for Rel.RoBERTa and Rel.SciBERT. Table 5 shows the F-scores, where we can see that in both essay- and paragraph-level AAEC, our ST model outperformed other end-to-end models (i.e., non-OS models). Compared with existing methods like BiPAM-syn, our model signifi- cantly improved the relation classification score. Although our model is not tailored to OS ar- chitectures or hyperparameters, when combined with OS it showed comparative or better parsing scores than the other models. Interestingly, BERT Trans. had better scores than our model in AAEC, while our model was better in link detection on CDCP. Since the architectures of our model and BERT Trans. are different (cf. graph-based vs. transition-based; Falenska et al., 2020), we con- clude that incorporating the advantages of both methods may improve the parsing performance. 5.3 Discussion: Transferability among Corpora To investigate transferability in MT-AM, we first examined the corpus-wise transferability by all → Corpusy to determine the pairs of Corpusx capabilities of each corpus. We then analyzed the label-wise transferability for relations. Corpus Preference: Figure 5 shows a matrix plot created by computing the error reduction (n = → Corpusy pairs. The 100%) for all Corpusx vertical axis represents the auxiliary corpus and the horizontal axis represents the target corpus. As shown in the figure, each corpus pair has a different corpus preference. The error reduction in each column varies in accordance with the → Corpusy pairs. For example, when Corpusx AAEC was an auxiliary corpus (i.e., AAEC row), the error reduction of the relation for AbstRCT was 0.1 while that for CDCP was 0.95. From the ‘‘avg’’ columns in Figure 5, we can evaluate how each corpus works as an auxil- iary corpus on average. As shown, the lower resource corpora, MTC and AASD, are less effec- tive as an auxiliary corpus for MT-AM because Table 5: Comparison of our ST model and other models. * is computed by C-F1 (100%) and R-F1 (100%) introduced by Eger et al. (2017). † calcu- lates scores on the basis of label sets introduced by Kuribayashi et al. (2019) for comparison, where we map AAEC’s Claim:Against and Claim:For la- bels into Claim. We also map an MTC’s component labels to Claim or Premise and relation labels to Support or Attack. the following state-of-the-art OS baselines for reference: • ILP Joint (Stab and Gurevych, 2017), which uses integer linear programming (ILP) for parsing. • Ptr. Net. (Potash et al., 2017), which uses a pointer network (Ptr. Net.) to predict relations. • Span Repr. (Kuribayashi et al., 2019), which uses distinct encoders to represent argumentative markers. • BERT Trans. (Bao et al., 2021), transition-based parser with BERT. • TSP-PLBA (Morio biaffine-based parser. et al., 2020), a a • Rel.RoBERTa/SciBERT (Mayer et al., (Vaswani et al., 2020),8 a Transformer 8https://gitlab.com/tomaye/ecai2020 9We only use the relation classification model, distin- -transformer based am. guishing Attack and Partial-attack labels. 648 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 → Corpusy) for ST models (with OS for components and Figure 5: Error reduction (not ERR) by MT-AM (Corpusx relations), showing each auxiliary corpus (vertical; Corpusx) and target corpus (horizontal; Corpusy) combination. For example, error reduction in the AAEC →AASD relation is 4.37. ‘‘All’’ represents MT-All. Darker green shows positive effects of MT-AM and darker blue shows negative effects. Note that the darkness is normalized in each task (i.e., span, component, or relation). they produce lower error reduction (AASD-‘‘avg’’ cells and MTC-‘‘avg’’ cells). In both compo- nent and relation classifications, the average error reductions of AAEC and CDCP as an auxiliary cor- pus (i.e., AAEC-‘‘avg’’ and CDCP-‘‘avg’’ cells) were better among all corpora. Because CDCP forms less constrained argument structures than tree-constrained argument structures such as MTC and AASD, we presume that one factor for the high transferability of CDCP stems from its ability to represent broad argumentative relations that may appear in a target corpus. AbstRCT also showed better error reductions, namely, 1.49 in compo- nent and 2.21 in relation, as an auxiliary corpus (i.e., AbstRCT-‘‘avg’’ cells). However, some auxiliary corpora barely improved error reduc- tions on the AbstRCT components and relations. This asymmetry result can be due to a domain difference. The above results indicate that the average error reduction varies depending on the auxiliary corpus (Corpusx). On the other hand, the average performance of MT-All (i.e., the ‘‘All’’-‘‘avg’’ cells) shows stable error reductions. For example, when CDCP was a target corpus, the average error reduction of MT-All (i.e., ‘‘All’’-CDCP cells) was better, namely, 0.14 in span and 1.61 in relation, → CDCP pairs. when compared with all Corpusx We presume that MT-All stabilized the fluctuation of performance produced by each auxiliary corpus. Label-wise Analysis: The label-wise best cor- pus combinations for the relation perspective are shown in Table 6 at n = 100%. We report the best auxiliary corpus with score transition from ST to MT-AM for each target corpus and its re- lation labels. For example, the ST model trained on AAEC produced a 44.54 F-score for Attack Target corpus (Corpusy) AAEC MTC CDCP AbstRCT AASD Relation Attack Support Exa Reb Sup Und Evidence Reason Best auxi- riary corpus (Corpusx) CDCP CDCP CDCP AAEC AAEC AAEC AAEC AAEC CDCP Attack Partial-Attack MTC Support AASD Additional Detail Sequence Support CDCP CDCP AbstRCT AAEC ST → MT ERR 44.54 → 46.60 67.37 → 68.35 3.71 3.01 3.47 → 21.33 18.51 53.56 → 59.58 12.96 59.68 → 64.22 11.26 40.38 → 47.79 12.43 0.00 → 0.67 40.79 → 41.73 0.67 1.59 34.26 → 37.91 5.55 48.37 → 52.28 7.56 58.72 → 58.46 –0.63 67.34 → 70.11 8.49 50.49 → 57.54 14.26 6.21 → 17.19 11.71 66.75 → 71.56 14.46 Table 6: Best auxiliary corpus for relations (with OS). ST → MT shows score transition from ST to Corpusx → Corpusy. relations, the MT-AM (CDCP→AAEC) that was best among all corpus combinations produced a 46.60 F-score, and the ERR between the ST and MT-AM was 3.71. The results in the table show that, similar to the corpus-wise analyses, there is a combination preference. For example, AAEC and CDCP seem to be the best combina- tion, where injecting auxiliary CDCP into AAEC produced a 3.71 ERR in Attack relation and inject- ing auxiliary AAEC into CDCP produced a 1.59 ERR in Reason relation. CDCP and AAEC are also better auxiliary corpora for MTC, for exam- ple, 18.51 ERR for Exa relation by CDCP→MTC was obtained. On the other hand, CDCP, Abs- tRCT, and AAEC are better auxiliary corpora for specific labels of AASD. While CDCP is bet- ter for Additional and Detail relations, AAEC is 649 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 better for the Support relation and AbstRCT is better for the Sequence relation. This suggests there could be corpus preferences for specific relation labels. The results in Table 6 also indicate that the ERRs of Attack and Partial-Attack of AbstRCT significantly improved (5.55 and 7.56 ERR, re- spectively). In addition, we observed a significant improvement for labels such as Exa, Reb, and Und of MTC (18.51, 12.96, and 12.43 ERR, respectively). Since the numbers of these la- bels are smaller in each corpus (as shown in Table 1), we conjecture that the prediction per- formance of such lower-resource labels improved thanks to the multi-task learning. Figure 6 shows the case of an MTC sub-graph at n = 100%. The ST model trained on MTC sometimes de- tected a false-positive Und relation from Although Ukraine... to Despite the..., while MT-AM suc- cessfully removed the false-positive relation. We explain this through the ST model trained on AAEC that was used to parse the MTC text. We found the predicted graph of the ST model (AAEC) was similar to the gold graph, since both graphs rep- resent the second component (i.e., the EU...) as a root component (i.e., MajorClaim). We conclude that this transferability between the auxiliary and target corpora successfully removed the false- positive relation. The ERRs for certain relations did not improve with MT-AM. For example, the ERR of AbstRCT Support was −0.63 (see Table 6). We presume that there are two reasons for this: (i) the number of Support labels is large enough to achieve a reason- able prediction performance and (ii) AbstRCT is a more challenging target corpus due to domain differences. For example, medical knowledge of AbstRCT cannot be obtained from other corpora. Also, a different scheme for relations, for exam- ple, a lower number of non-crossing relations (as shown in Table 2), could not be transferred from other corpora. 5.4 Discussion: Typology of Transferability in AM We further discuss the results under low-resource settings to examine the typology of transferability. We hypothesize that there are three possible types of transferability: data/training sufficiency, an- notation compatibility, and semantically induced compatibility, as follows. Figure 6: Case study: Predicted results of an MTC sub-graph with OS. For visibility, we combine model outputs for several CVs or seeds. 1. Data/Training Sufficiency: Models will not output minor classes such as ‘‘B’’ in the BIO tags and positive link classes when insufficient training data is available or a small number of training steps is applied. We found that MT-AM alleviates this problem. Figure 7 shows the number of predicted components and relations, regardless of whether the outputs are correct or not. For relations, we can see in Figure 7b that MT-All helps increase the link outputs for both AAEC and CDCP under low-resource settings, suggesting that the minor positive link class is more often predicted in MT-All. For components, we can see in Figure 7a that the number of predicted components increases for CDCP at n = 1%, while AAEC shows a different phenomenon where the 650 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Figure 7: Number of predicted components and rela- tions for each training data amount (n%). The red line shows the gold number. Table 7: Predicted AASD spans. The red box shows a span start character, and the blue box shows the end. When we add up all the CV results, the darker box is the more predicted. See also the original abstract in Gao et al. (2014). l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Figure 8: Relationship between span similarity (F-score of ST model; horizontal axis) for target corpus AASD and ERR. number of predicted components by ST is larger than those by MT-All at n = 1%. This is because there is a greater number of ‘‘O’’ tags in AAEC since the span of the corpus is not segment-based. 2. AnnotationCompatibility: Under low-resource settings, because a model cannot exploit much information of a target corpus in the training, compatibility of the annotation design between the auxiliary corpus and target corpus would di- rectly help multi-task training improve the parsing performance. To demonstrate this, we define the span or link similarity10 that approximates the compatibility between an auxiliary corpus and target corpus. For span similarity, we compute the span F-score for a target Corpusy by using an ST model trained on another Corpusx. Similarly, 10Because component-label types of an auxiliary corpus differ from those of a target corpus, we only investigated span and link (i.e., unlabeled) similarities. Figure 9: Relationship between link similarity (F-score of ST model) for target corpus AASD and ERR. Figure 10: F-score of the root component detection (with OS) with AbstRCT→AASD for each training data amount (n%). we compute the link F-score with OS for a target Corpusy. Since AASD is the smallest corpus in this study, we examined the relationship between similarity and the ERR of the corpus. Figures 8 and 9 show the relationships between the computed similar- → AASD.11 Under ities and ERRs by Corpusx 11We exclude MTC as an auxiliary corpus because it is too small to compare with the other corpora. 651 Table 8: Predicted root components (with OS). We add up all CV or seed results, so the darker the color, the more predicted. Gold root component is fourth (i.e., 3. In this...). the extreme low-resource setting (n = 1%), as shown in Figures 8a and 9a, we observe the potential for positive correlations between the similarities and ERRs. This implies that the an- notation compatibility directly helps multi-task learning. Specifically, in terms of the span, CDCP showed the best compatibility with AASD in Figure 8a (n = 1%) because both span similarity and ERR are highest. AbstRCT showed the worst compati- bility with AASD because both span similarity and ERR are lower. This is illustrated in Table 7, which represents the span predictions. At n = 1%, the ST (AASD) often produced few segments, while the CDCP → AASD split multiple sentences into segments that were almost compatible with the gold segments. Compared with CDCP → AASD, AbstRCT → AASD produced fewer segments. This is because AbstRCT spans are partially se- lected from a text, and are not compatible with AASD. On the other hand, when n = 100% (as shown in Figure 8b), we found fewer positive cor- relations than in n = 1%, suggesting that MT-AM does not benefit from the compatibility when the training data amount is sufficient. In fact, as we can see with AbstRCT → AASD (n = 100%) in Table 7, the issue of fewer segments produced by AbstRCT → AASD (n = 1%) has already been resolved. In terms of links, CDCP and AAEC in Figure 9a (n = 1%) showed better compatibility while Abs- tRCT showed the worst. In contrast with the span, the correlation can still be observed when n = 100% (Figure 9b). 3. Semantic Compatibility: Although the above-mentioned on annotation transfer between auxiliary and target corpus, we argue that another possible transfer- ability also exists. That is, in addition to the direct transferability focused transfer based on the link and span similarity between the auxiliary corpus and target corpus in the annotation compatibility, we want to examine the transfer of potential features of the argument structure. Through our analysis, we found that MT-AM sometimes improves the root component detection significantly when n = 10%. Figure 10 shows the root component detection by AbstRCT→AASD for each training data amount. We can see that, compared with the ST model, auxiliary AbstRCT significantly improves the score when n = 10%. We suppose this is derived from semantic compatibility such as implicit argumentative knowledge. the F-score for To validate this, we show the predicted root components of AbstRCT→AASD in Table 8. As mentioned above, annotation compatibility could be important for improving the parsing perfor- mance under low-resource settings. Since AASD and AbstRCT do not have much compatibility, as shown in Figure 9a, the gold root compo- nent (i.e., the fourth component) of AASD is less compatible with the ST model trained on Abs- tRCT (Table 8(a)). This makes the prediction of AbstRCT → AASD at n = 1% (Table 8(d)) less similar to the gold. However, as shown in Table 8 (b), we found that AbstRCT→AASD at n = 10% predicted the fourth component as the root. Surprisingly, this result was not observed in the ST model trained on AASD (Table 8 (c)). This suggests that AbstRCT is able to transfer implicit argumentative knowledge that can determine the root component of a small target corpus. 5.5 Limitation While we have shown both the effectiveness and the transferability capability of MT-AM, the results might change depending on the model architecture, implementation, experimental de- sign, hyperparameters, and unrecognized factors 652 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 of annotations. In addition, the three types of trans- ferability are not likely to be independent of each other and are no better than hypotheses. More ex- periments and discussions will be our future work. 6 Conclusion In this paper, we focused on argument mining (AM) for handling a scarcity of resources and proposed an end-to-end multi-task AM (MT-AM) model for various corpora. Experiments showed that the proposed MT-AM generally outperformed single-task models and further improved the parsing performance under low-resource settings. Our extensive analyses suggest that the ad- vantage of MT-AM is due to its three types of transferability: data/training sufficiency, annota- tion compatibility, and semantic compatibility. In future work, we will develop more accu- rate MT-AM parsers on the basis of these transferability hypotheses. Also, it has been suggested by a reviewer that using a trained parser to generate a silver corpus could be a means of extending the training data. Future work should thus address the develop- ment of a silver corpus. We are also interested in multi-lingual training using the proposed system. Acknowledgments Three anonymous reviewers and the action edi- tor gave us insightful comments and suggestions for our paper. Computational resources of AI Bridging Cloud Infrastructure (ABCI) provided by National Institute of Advanced Industrial Science and Technology (AIST) were used. Dr. Naoaki Okazaki at Tokyo Institute of Technology gave us comments for our paper. References Pablo Accuosto and Horacio Saggion. 2019. Transferring knowledge from discourse to argu- ments: A case study with scientific abstracts. In Proceedings of the 6th Workshop on Argument Mining, pages 41–51, Florence, Italy. Associa- tion for Computational Linguistics. https:// doi.org/10.18653/v1/W19-4505 Pablo Accuosto and Horacio Saggion. 2020. Mining arguments in scientific abstracts with embeddings. Data Knowl- discourse-level edge Engineering, 129:101840. https:// doi.org/10.1016/j.datak.2020.101840 Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. Optuna: A next-generation hyperparameter op- timization framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD ’19, pages 2623–2631, New York, NY, USA. ACM. https://doi.org/10 .1145/3292500.3330701 Jianzhu Bao, Chuang Fan, Jipeng Wu, Yixue Dang, Jiachen Du, and Ruifeng Xu. 2021. A neural transition-based model for argumenta- tion mining. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 6354–6364, Online. Association for Computational Linguistics. Iz Beltagy, Kyle Lo, and Arman Cohan. 2019. SciBERT: A pretrained language model for scientific text. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Pro- cessing (EMNLP-IJCNLP), pages 3615–3620, Hong Kong, China. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/D19-1371 Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. Longformer: The long-document trans- former. arXiv:2004.05150. Filip Boltuˇzi´c and Jan ˇSnajder. 2020. Structured prediction models for argumentative claim parsing from text. Automatika, 61(3):361–370. https://doi.org/10.1080/00051144 .2020.1761101 Lynn Carlson, Daniel Marcu, and Mary Ellen Okurovsky. 2001. Building a discourse-tagged corpus in the framework of Rhetorical Structure Theory. In Proceedings of the Second SIGdial Workshop on Discourse and Dialogue. Tuhin Chakrabarty, Christopher Hidey, and Kathy McKeown. 2019. IMHO fine-tuning improves claim detection. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguis- tics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 558–563, Minneapolis, Minnesota. Association for Computational Linguistics. 653 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Yoeng-Jin Chu and Tseng-Hong Liu. 1965. On the shortest arborescence of a directed graph. Science Sinica, 14:1396–1400. Oana Cocarascu, Elena Cabrio, Serena Villata, and Francesca Toni. 2020. Dataset independent baselines for relation prediction in argument mining. In Frontiers in Artificial Intelligence and Applications, Volume 326: Computa- tional Models of Argument (COMMA 2020), pages 45–52. IOS Press. Michael A. Covington. 2001. A fundamental algorithm for dependency parsing. In In Pro- ceedings of the 39th Annual ACM Southeast Conference, pages 95–102. Johannes Daxenberger, Steffen Eger, Ivan Habernal, Christian Stab, and Iryna Gurevych. 2017. What is the essence of a claim? Cross-domain claim identification. In Proceedings of the 2017 Conference on Empirical Methods in Natu- ral Language Processing, pages 2055–2066, Copenhagen, Denmark. Association for Com- putational Linguistics. https://doi.org /10.18653/v1/D17-1218 Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguis- tics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics. Timothy Dozat and Christopher D. Manning. 2017. Deep biaffine attention for neural depen- dency parsing. In 5th International Conference ICLR 2017, on Learning Representations, Toulon, France, April 24-26, 2017, Conference Track Proceedings. Timothy Dozat and Christopher D. Manning. 2018. Simpler but more accurate semantic de- pendency parsing. In Proceedings of the 56th Annual Meeting of the Association for Compu- tational Linguistics (Volume 2: Short Papers), pages 484–490, Melbourne, Australia. Associa- tion for Computational Linguistics. https:// doi.org/10.18653/v1/P18-2077 Markus Eberts 2020. Span-based joint entity and relation extraction and Adrian Ulges. with transformer pre-training. In Frontiers in Artificial Intelligence and Applications, 24th European Conference on Artificial Intelligence (ECAI 2020), pages 2006–2013. IOS Press. Judith Eckle-Kohler, Roland Kluge, and Iryna Gurevych. 2015. On the role of discourse mark- ers for discriminating claims and premises in argumentative discourse. In Proceedings of the 2015 Conference on Empirical Methods in Nat- ural Language Processing, pages 2236–2242, Lisbon, Portugal. Association for Computa- tional Linguistics. Frans H. Van Eemeren, Rob Grootendorst, and Francisca Snoeck Henkemans. 1996. Funda- mentals of Argumentation Theory a Handbook of Historical Backgrounds and Contempo- rary Developments. https://doi.org/10 .2307/358423 Steffen Eger, Johannes Daxenberger, and Iryna Gurevych. 2017. Neural end-to-end learning for computational argumentation mining. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguis- tics (Volume 1: Long Papers), pages 11–22, Vancouver, Canada. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/P17-1002 Agnieszka Falenska, Anders Bj¨orkelund, and Jonas Kuhn. 2020. Integrating graph-based and transition-based dependency parsers in the deep contextualized era. In Proceedings of the 16th International Conference on Parsing Tech- nologies and the IWPT 2020 Shared Task on Parsing into Enhanced Universal Depen- dencies, pages 25–39, Online. Association for Computational Linguistics. https://doi .org/10.18653/v1/2020.iwpt-1.4 James B. Freeman. 2011. Argument Structure: Representation and Theory, Argumentation Library (18), Springer. https://doi.org /10.1007/978-94-007-0357-5 Andrea Galassi, Marco Lippi, and Paolo Torroni. 2018. Argumentative link prediction using residual networks and multi-objective learn- ing. In Proceedings of the 5th Workshop on Argument Mining, pages 1–10, Brussels, Bel- gium. Association for Computational Linguis- tics. https://doi.org/10.18653/v1 /W18-5201 654 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Andrea Galassi, Marco Lippi, and Paolo Torroni. 2021. Multi-task attentive residual networks for argument mining. CoRR, abs/2102.12227. Jianfeng Gao, Patrick Pantel, Michael Gamon, Xiaodong He, and Li Deng. 2014. Model- ing interestingness with deep neural networks. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Pro- cessing (EMNLP), pages 2–13, Doha, Qatar. Association for Computational Linguistics. Alex Graves and J¨ugen Schmidhuber. 2005. Framewise phoneme classification with bidi- rectional LSTM and other neural network architectures. Neural Networks, 18(5):602–610. https://doi.org/10.1016/j.neunet .2005.06.042, PubMed: 16112549 Grigorii Guz, Patrick Huber, and Giuseppe Carenini. 2020. Unleashing the power of neu- ral discourse parsers - a context and structure aware approach using large scale pretrain- ing. In Proceedings of the 28th International Conference on Computational Linguistics, pages 3794–3805, Barcelona, Spain (Online). International Committee on Computational Linguistics. Ivan Habernal and Iryna Gurevych. 2017. user-generated Argumentation mining web discourse. Computational Linguistics, 43(1):125–179. https://doi.org/10.1162 /COLI a 00276 in In Proceedings of Jeremy Howard and Sebastian Ruder. 2018. Uni- language model fine-tuning for text versal the 56th classification. Annual Meeting of the Association for Compu- tational Linguistics (Volume 1: Long Papers), pages 328–339, Melbourne, Australia. Associ- ation for Computational Linguistics. https: //doi.org/10.18653/v1/P18-1031 Patrick Huber and Giuseppe Carenini. 2019. Predicting discourse structure using distant supervision from sentiment. In Proceedings of the 2019 Conference on Empirical Meth- ods in Natural Language Processing and the 9th International Joint Conference on Natu- ral Language Processing (EMNLP-IJCNLP), pages 2306–2316, Hong Kong, China. Associ- ation for Computational Linguistics. https:// doi.org/10.18653/v1/D19-1235 parsing. In Proceedings of the 52nd Annual the Association for Computa- Meeting of tional Linguistics (Volume 1: Long Papers), pages 13–24, Baltimore, Maryland. Association for Computational Linguistics. Zhengbao Jiang, Wei Xu, Jun Araki, and Graham Neubig. 2020. Generalizing natu- language analysis through span-relation ral the 58th representations. In Proceedings of Annual Meeting of the Association for Computational Linguistics, pages 2120–2133, Online. Association for Computational Linguis- tics. https://doi.org/10.18653/v1 /2020.acl-main.192 Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In Pro- ceedings of the 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015. Christian Kirschner, Judith Eckle-Kohler, and Iryna Gurevych. 2015. Linking the thoughts: Analysis of argumentation structures in sci- the entific publications. In Proceedings of 2nd Workshop on Argumentation Mining, pages 1–11, Denver, CO. Association for Computational Linguistics. https://doi .org/10.3115/v1/W15-0501 Tatsuki Kuribayashi, Hiroki Ouchi, Naoya Inoue, Paul Reisert, Toshinori Miyoshi, Jun Suzuki, and Kentaro Inui. 2019. An empirical study of span representations in argumentation structure pars- ing. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguis- tics, pages 4691–4698, Florence, Italy. Associ- ation for Computational Linguistics. https:// doi.org/10.18653/v1/P19-1464 Anne Lauscher, Goran Glavaˇs, and Simone Paolo Ponzetto. 2018. An argument-annotated cor- pus of scientific publications. In Proceedings of the 5th Workshop on Argument Mining, pages 40–46, Brussels, Belgium. Associa- tion for Computational Linguistics. https:// doi.org/10.18653/v1/W18-5206 John Lawrence and Chris Reed. 2019. Ar- gument mining: A survey. Computational Linguistics, 45(4):765–818. https://doi .org/10.1162/coli_a_00364 Yangfeng Ji and Jacob Eisenstein. 2014. Rep- resentation learning for text-level discourse Xiaodong Liu, Pengcheng He, Weizhu Chen, and Jianfeng Gao. 2019a. Multi-task deep neural 655 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 networks for natural language understanding. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4487–4496, Florence, Italy. Association for Computational Linguistics. Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pages 985–995, Vancouver, Canada. Associa- tion for Computational Linguistics. https:// doi.org/10.18653/v1/P17-1091 Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019b. RoBERTa: A robustly op- timized BERT pretraining approach. arXiv preprint arXiv:1907.11692. Tobias Mayer, Elena Cabrio, and Serena Villata. 2020. Transformer-based argument mining for healthcare applications. In Frontiers in Ar- tificial Intelligence and Applications, 24th European Conference on Artificial Intelligence (ECAI 2020), pages 2108–2115. IOS Press. Makoto Miwa and Mohit Bansal. 2016. End-to-end relation extraction using LSTMs In Pro- on sequences and tree structures. ceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1105–1116, Berlin, Germany. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/P16-1105 Gaku Morio, Hiroaki Ozaki, Terufumi Morishita, Yuta Koreeda, and Kohsuke Yanai. 2020. Towards better non-tree argument mining: Proposition-level biaffine parsing with task-specific parameterization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3259–3266, Online. Association for Computational Lin- guistics. https://doi.org/10.18653 /v1/2020.acl-main.298 Terufumi Morishita, Gaku Morio, Hiroaki Ozaki, and Toshinori Miyoshi. 2020. Hitachi at SemEval-2020 task 7: Stacking at scale with heterogeneous language models for humor the Four- In Proceedings of recognition. teenth Workshop on Semantic Evaluation, pages 791–803, Barcelona (online). Interna- tional Committee for Computational Linguis- tics. https://doi.org/10.18653/v1 /2020.semeval-1.101 Vlad Niculae, Joonsuk Park, and Claire Cardie. 2017. Argument mining with structured SVMs and RNNs. In Proceedings of the 55th Annual Stephan Oepen, Omri Abend, Lasha Abzianidze, Johan Bos, Jan Hajic, Daniel Hershcovich, Bin Li, Tim O’Gorman, Nianwen Xue, and Daniel Zeman. 2020. MRP 2020: The sec- ond shared task on cross-framework and cross-lingual meaning representation pars- the CoNLL 2020 ing. Shared Task: Cross-Framework Meaning Representation Parsing, pages 1–22, On- line. Association for Computational Linguis- tics. https://doi.org/10.18653/v1 /2020.conll-shared.1 In Proceedings of Stephan Oepen, Omri Abend, parsing. Jan Hajic, Daniel Hershcovich, Marco Kuhlmann, Tim O’Gorman, Nianwen Xue, Jayeol Chun, Milan Straka, and Zdenka Uresova. 2019. MRP 2019: Cross-framework meaning represen- the tation Shared Task on Cross-Framework Mean- ing Representation Parsing at the 2019 Conference on Natural Language Learn- ing, pages 1–27, Hong Kong. Association for Computational Linguistics. https:// doi.org/10.18653/v1/K19-2001 In Proceedings of Joonsuk Park, Cheryl Blake, and Claire Cardie. 2015a. Toward machine-assisted participation in erulemaking: An argumentation model of evaluability. In Proceedings of the 15th Inter- national Conference on Artificial Intelligence and Law, ICAIL ’15, pages 206–210, New York, NY, USA. Association for Computing Machinery. https://doi.org/10.1145 /2746090.2746118 Joonsuk Park and Claire Cardie. 2018. A corpus of eRulemaking user comments for measur- ing evaluability of arguments. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. European Language Resources Association (ELRA). Joonsuk Park, Arzoo Katiyar, and Bishan Yang. 2015b. Conditional random fields for identify- ing appropriate types of support for propositions in online user comments. In Proceedings of the 2nd Workshop on Argumentation Mining, 656 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 pages 39–44, Denver, CO. Association for Computational Linguistics. https://doi .org/10.3115/v1/W15-0506 In Proceedings of Andreas Peldszus. 2014. Towards segment-based recognition of argumentation structure in short the First Work- texts. shop on Argumentation Mining, pages 88–97, Baltimore, Maryland. Association for Compu- tational Linguistics. https://doi.org/10 .3115/v1/W14-2112 Andreas Peldszus and Manfred Stede. 2015. Joint prediction in MST-style discourse parsing for argumentation mining. In Proceedings of the 2015 Conference on Empirical Methods in Nat- ural Language Processing, pages 938–948, Lisbon, Portugal. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/D15-1110 Andreas Peldszus and Manfred Stede. 2016. An annotated corpus of argumentative micro- texts. In Argumentation and Reasoned Action: Proceedings of the 1st European Conference on Argumentation, Lisbon 2015 / Vol. 2, pages 801–815, London. College Publications. Isaac Persing and Vincent Ng. 2016. End-to-end argumentation mining in student essays. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Lan- guage Technologies, pages 1384–1394, San Diego, California. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/N16-1164 Peter Potash, Alexey Romanov, and Anna Rumshisky. 2017. Here’s my point: Joint pointer architecture for argument mining. the 2017 Conference on In Proceedings of Empirical Methods in Natural Language Processing, pages 1364–1373, Copenhagen, Denmark. Association for Computational Lin- guistics. https://doi.org/10.18653 /v1/D17-1143 Jan Wira Gotama Putra, Simone Teufel, and Takenobu Tokunaga. 2021a. Multi-task and multi-corpora training strategies to enhance ar- gumentative sentence linking performance. In Proceedings of the 8th Workshop on Argument Mining, pages 12–23, Punta Cana, Domini- can Republic. Association for Computational Linguistics. Jan Wira Gotama Putra, Simone Teufel, and Takenobu Tokunaga. 2021b. Parsing argumen- tative structure in English-as-foreign-language essays. In Proceedings of the 16th Workshop on Innovative Use of NLP for Building Edu- cational Applications, pages 97–109, Online. Association for Computational Linguistics. Ruty Rinott, Lena Dankin, Carlos Alzate Perez, Mitesh M. Khapra, Ehud Aharoni, and Noam Slonim. 2015. Show me your evidence - an automatic method for context dependent evi- dence detection. In Proceedings of the 2015 Conference on Empirical Methods in Nat- ural Language Processing, pages 440–450, Lisbon, Portugal. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/D15-1050 Claudia Schulz, Steffen Eger, Johannes Daxenberger, Tobias Kahse, and Iryna Gurevych. 2018. Multi-task learning for argumentation mining in low-resource settings. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 35–41, New Orleans, Louisiana. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/N18-2006 Anders Søgaard and Yoav Goldberg. 2016. Deep multi-task learning with low level tasks super- vised at lower layers. In Proceedings of the 54th Annual Meeting of the Association for Compu- tational Linguistics (Volume 2: Short Papers), pages 231–235, Berlin, Germany. Associa- tion for Computational Linguistics. https:// doi.org/10.18653/v1/P16-2038 Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, and Ruslan Ilya Sutskever, Salakhutdinov. 2014. Dropout: A simple way to prevent neural networks from overfit- ting. Journal of Machine Learning Research, 15(56):1929–1958. Christian Stab and Iryna Gurevych. 2014. Iden- tifying argumentative discourse structures in persuasive essays. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 46–56, Doha, Qatar. Association for Computational Linguistics. https://doi.org/10.3115 /v1/D14-1006 657 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Christian Stab and Iryna Gurevych. 2017. Pars- ing argumentation structures in persuasive es- says. Computational Linguistics, 43(3):619–659. https://doi.org/10.1162/COLI a 00295 Manfred Stede and Jodi Schneider. 2018. Argu- mentation Mining, volume 11. Morgan Clay- pool Publishers. https://doi.org/10 .2200/S00883ED1V01Y201811HLT040 Stephen E. Toulmin. 2003. The Uses of Argument, second ed. Cambridge Univer- sity Press. https://doi.org/10.1017 /CBO9780511840005 Dietrich Trautmann. 2020. Aspect-based argu- ment mining. In Proceedings of the 7th Work- shop on Argument Mining, pages 41–52, Online. Association for Computational Linguistics. Dietrich Trautmann, Johannes Daxenberger, Christian Stab, Hinrich Sch¨utze, and Iryna Gurevych. 2020. Fine-grained argument unit recognition and classification. Proceedings of the AAAI Conference on Artificial Intelligence, 34(05):9048–9056. https://doi.org/10 .1609/aaai.v34i05.6438 Rob van der Goot, Ahmet ¨Ust¨un, Alan Ramponi, Ibrahim Sharaf, and Barbara Plank. 2021. Massive choice, ample tasks (MaChAmp): A toolkit for multi-task learning in NLP. In Proceedings of the European Chapter of the Association for Com- putational Linguistics: System Demonstrations, pages 176–197, Online. Association for Com- putational Linguistics. https://doi.org /10.18653/v1/2021.eacl-demos.22 the 16th Conference of Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Information Pro- In Advances cessing Systems 30, Curran Associates, Inc., pages 5998–6008. in Neural David Vilares and Carlos G´omez-Rodr´ıguez. 2018. A transition-based algorithm for un- restricted AMR parsing. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 142–149, New Orleans, Louisiana. Association for Computa- tional Linguistics. https://doi.org/10 .18653/v1/N18-2023 Oriol Vinyals, Meire Fortunato, and Navdeep Jaitly. 2015. Pointer networks. In Advances in Neural Information Processing Systems, volume 28, pages 2692–2700. Curran Asso- ciates, Inc. In Proceedings of Henning Wachsmuth, Giovanni Da San Martino, Dora Kiesel, and Benno Stein. 2017. The im- pact of modeling overall argumentation with the 2017 tree kernels. Conference on Empirical Methods in Natu- ral Language Processing, pages 2379–2389, Copenhagen, Denmark. Association for Com- putational Linguistics. https://doi.org /10.18653/v1/D17-1253 In Proceedings of Hao Wang, Zhen Huang, Yong Dou, and Yu Hong. 2020. Argumentation mining on essays at the 28th multi scales. International Conference on Computational Linguistics, pages 5480–5493, Barcelona, Spain (Online). International Committee on Compu- tational Linguistics. https://doi.org/10 .18653/v1/2020.coling-main.478 An Yang and Sujian Li. 2018. SciDTB: Discourse dependency TreeBank for scientific abstracts. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguis- tics (Volume 2: Short Papers), pages 444–449, Melbourne, Australia. Association for Compu- tational Linguistics. https://doi.org/10 .18653/v1/P18-2071 Yuxiao Ye and Simone Teufel. 2021. End-to-end argument mining as biaffine dependency pars- ing. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Vol- ume, pages 669–678, Online. Association for Computational Linguistics. 658 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u / t a c l / l a r t i c e - p d f / d o i / . 1 0 1 1 6 2 / t l a c _ a _ 0 0 4 8 1 2 0 2 2 9 6 5 / / t l a c _ a _ 0 0 4 8 1 p d . f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3

Download pdf