=Paper=
{{Paper
|id=Vol-2614/session4_paper2
|storemode=property
|title=Semi-supervised Models via Data Augmentation for Classifying Interactive Affective Responses
|pdfUrl=https://ceur-ws.org/Vol-2614/AffCon20_session4_semisupervised.pdf
|volume=Vol-2614
|authors=Jiaao Chen,Yuwei Wu, Diyi Yang
|dblpUrl=https://dblp.org/rec/conf/aaai/ChenWY20
}}
==Semi-supervised Models via Data Augmentation for Classifying Interactive Affective Responses==
https://ceur-ws.org/Vol-2614/AffCon20_session4_semisupervised.pdf
Semi-Supervised Models via Data Augmentation
for Classifying Interactive Affective Responses
Jiaao Chen ∗1 , Yuwei Wu?2 , and Diyi Yang1
1
Georgia Institute of Technology, Atlanta GA 30318, USA
2
Shanghai Jiao Tong University, Shanghai, China
jiaaochen@gatech.edu, will8821@sjtu.edu.cn, diyi.yang@cc.gatech.edu
Abstract. We present semi-supervised models with data augmentation
(SMDA), a semi-supervised text classification system to classify interac-
tive affective responses. SMDA utilizes recent transformer-based models
to encode each sentence and employs back translation techniques to para-
phrase given sentences as augmented data. For labeled sentences, we per-
formed data augmentations to uniform the label distributions and com-
puted supervised loss during training process. For unlabeled sentences,
we explored self-training by regarding low-entropy predictions over unla-
beled sentences as pseudo labels, assuming high-confidence predictions as
labeled data for training. We further introduced consistency regulariza-
tion as unsupervised loss after data augmentations on unlabeled data,
based on the assumption that the model should predict similar class
distributions with original unlabeled sentences as input and augmented
sentences as input. Via a set of experiments, we demonstrated that our
system outperformed baseline models in terms of F1-score and accuracy.
Keywords: Semi-Supervised Learning · Data Augmentation · Deep Learn-
ing · Social Support · Self-disclosure
1 Introduction
Affect refers to emotion, sentiment, mood, and attitudes including subjective
evaluations, opinions, and speculations [23]. Psychological models of affect have
been utilized by other extensive computational research to operationalize and
measure users’ opinions, intentions, and expressions. Understanding affective
responses with in conversations is an important first step for studying affect
and has attracted a growing amount of research attention recently [20, 4, 19].
The affective understanding of conversations focuses on the problem of how
speakers use emotions to react to a situation and to each other, which can
help better understand human behaviors and build better human-computer-
interaction systems.
However, modeling affective responses within conversations is relatively chal-
lenging since it is hard to quantify the affectiveness [16] and there are no large-
scale labeled dataset about affective levels in responses. In order to facilitate
?
Equal Contribution. This work was done when Yuwei is a visiting student at Georgia
Tech.
Copyright 2020 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0). In: N. Chhaya, K. Jaidka, J. Healey, L. H. Ungar, A. Sinha
(eds.): Proceedings of the 3rd Workshop of Affective Content Analysis, New York, USA, 07-
FEB-2020, published at http://ceur-ws.org
2 C. Jiaao and W. Yuwei et al.
research in modeling interactive affective responses, [8] introduced a conversa-
tion dataset, OffMyChest, building from Reddit, and proposed two tasks: (1)
Semi-supervised learning task: predict labels for Disclosure and Supportiveness
in sentences based on a small amount of labeled and large unlabeled training
data; (2) Unsupervised task: design new characterizations and insights to model
conversation dynamics. The current work focused on the first task.
With limited labeled data and large amount of unlabeled data being given,
to alleviate the dependence on labeled data, we combine recent advances in lan-
guage modeling, semi-supervised learning on text and data augmentations on
text to form Semi-Supervised Models via Data Augmentation (SMDA). SMDA
consists of two parts: supervised learning over labeled data (Section 4.1) and
unsupervised learning over unlabeled data (Section 4.2). Both parts utilize data
augmentations to enhance the learning procedures. Our contributions in this
work can be summarized into three parts: analysed the OffMyChest dataset
in Section 3, proposed a semi-supervised text classification system to classify
interactive affective responses classification in Section 4 and described the ex-
perimental details and results in Section 5.
2 Related Work
Transformer-based Models : With transformer-based pre-trained models becom-
ing more and more widely-used, pre-training and fine-tuning framework [7] with
large pre-trained language models are applied into a lot of NLP applications
and achieved state-of-the-art performances [18]. Language models [15, 7, 21] or
masked language models [3, 10] are pre-trained over a large amount of text from
Wikipedia and then fine-tuned on specific tasks like text classifications. Here we
built our SMDA system based on such framework.
Data Augmentation on Text : When the amount of labeled data is limited,
one common technique for handling the shortage of data is to augment given
data and generate more training “augmented” data. Previous work has utilized
simple operations like synonym replacement, random insertion, random swap
and random deletion for text data augmentation [17]. Another line of research
applied neural models for augmenting text by generating paraphrases via back
translations [18] and monotone submodular function maximization [9]. Building
on those prior work, we utilized back translations as our augment methods on
both labeled and unlabeled sentences.
Semi-Supervised Learning on Text Classification : One alternative to deal with
the lack of labeled data is to utilize unlabeled data in the learning process, which
is denoted as Semi-Supervised Learning (SSL), since unlabeled data is usually
easier to get compared to labeled data. Researchers has made use of variational
auto encoders (VAEs) [2, 22, 6], self-training [11, 5, 12], consistency regularization
[14, 13, 18] to introduce extra loss functions over unlabeled data to help the learn-
ing of labeled samples. VAEs utilize latent variables to reconstruct input labeled
Semi-Supervised Models via Data Augmentation 3
and unlabeled sentences and predict sentence labels with these latent variables;
self-training adds unlabeled data with high-confidence predictions as pseudo la-
beled data during training process and consistency regularization forces model to
output consistent predictions after adding adversarial noise or performing data
augmentations to input data. We combined self-training, entropy minimization
and consistency regularization in our system for unlabeled sentences.
3 Data Analysis and Pre-processing
Researching how human initiate and hold conversations has attracted increasing
attention those days, as it can help us better understand how human behave over
conversations and build better AI systems like social chatbot to communicate
with people. In this section, we took a closer look at the conversation dataset,
OffMyChest [8], for better understanding and modeling interactive affective re-
sponses. Specifically, we describe certain characteristics of this dataset and our
pre-processing steps.
3.1 Label Definition
For each comment of a post on Reddit, [8] annotated them with 6 labels: In-
formation disclosure representing some degree of personal information in com-
ments; Emotional disclosure representing comments containing certain positive
or negative emotions; Support referring to comments offering social support like
advice; General support representing that comments are offering general support
through quotes and catch phrases, with Information support offering specific in-
formation like practical advice, and Emotional support offering sympathy, caring
or encouragement. Each comment can belong to multiple categories.
Fig. 1. Distribution of each label in the labeled corpus. The y axis is the number of
sentences that have the corresponding labels.
3.2 Data Statics
In OffMyChest corpus, there are 12,860 labeled sentences and over 420k unla-
beled sentences for training, 5,000 unlabeled sentences for test. The label dis-
tributions of labeled sentences are showed in Fig. 1. To train and evaluate our
4 C. Jiaao and W. Yuwei et al.
Table 1. Dataset split statistics. We utilized both labeled data and unlabeled data for
training, generated dev and test set by sampling from given labeled comments set.
Labeled Train Set Dev Set Test Set Unlabeled Train Set
8,000 2,000 2,860 420,607
Table 2. Paraphrase examples generated via back translation from original sentences
into augmented sentences.
Original Augmented Labels
I’m crying a lot of tears of joy Right now I’m crying a lot of
Emo disclosure
right now. happy tears.
Stepdad will be the one walking It will be my stepfather walking
me down the aisle when I get me down the aisle when I Info disclosure
married. get married.
Hope you have a nice day. I hope you have a good day. Support
Both of you are giving it your
Your best effort, both of you General support
best shot.
Plan your transition back to Plan your move back to a job
Info support
working outside of the home. outside your own home.
I am so freaking happy for you! I’m so excited for you! Emo support
systems, we randomly split the given labeled sentence set into train, development
and test set. The data statics are shown in Table 1. We tuned hyper-parameters
and chose best models based on performance on dev set, and reported model’s
performance on test set.
3.3 Pre-processing
We utilized XLNet-cased-based tokenizer3 to split each sentence into tokens. We
showed the cumulative sentence length distribution in Fig. 2, 95% comments
have less than 64 tokens. Thus we set the maximum sentence length to 64,
and remained the first 64 tokens for sentences that exceed the limit. As for data
augmentations, we made use of back translation with German as middle language
to generate paraphrases for given sentences. Specifically, we loaded translation
model from Fairseq4 , translated given sentences from English to German, and
then translated them back to English. Also to increase the diversity of generated
paraphrases, we employed random sampling with a tunable temperature (0.8)
instead of beam search for the generation. We describe some examples in Table 2.
4 Method
We convert this 6-class affective response classification task into 6 binary clas-
sification tasks, namely whether each sentence belongs to each category or not
3
https://huggingface.co/transformers/model_doc/xlnet.html#xlnettokenizer
4
https://github.com/pytorch/fairseq
Semi-Supervised Models via Data Augmentation 5
Fig. 2. Cumulative distribution of sentence length in the given labeled sentence set.
The y axis represents the portion over all sentences.
(labeled with 1 or 0). For each binary classification task, given a set of labeled
sentences consisting of n samples S = {s1 , ..., sn } with labels L = {l1 , ..., ln },
where li ∈ {0, 1}2 , and a set of unlabeled sentences Su = {su1 , ..., sum }, our goal is
to learn the classifier f (ˆl|s, θi ), i ∈ [1, 6]. Our SMDA model contains several com-
ponents: Supervised Learning (Section 4.1) for labeled sentences, Unsupervised
Learning (Section 4.2) for unlabeled sentences, and Semi-Supervised Objective
Function (Section 4.3) to combine labeled and unlabeled sentences.
4.1 Supervised Learning
Generating Balanced Labeled Training Set As shown in Fig. 1, the dis-
tribution is very unbalanced with respect to General support, Info support and
Emo support. In order to get more training sentences with these three types of
support and make these three binary classification sub-tasks learn-able with a
more balanced training set, we performed data augmentations over sentences
with these three labels. Specifically, we paraphrased each sentence by 4 times
via back translations and regarded that the augmented sentences have the same
labels as original sentences. The comparison distributions are shown in Fig. 3
Supervised Learning for Labeled Sentences For each input labeled sen-
tence si , we used XLNet [21] g(.) to encode it into hidden representation hi =
g(si ), and then passed them though a 2-layer MLP to predict the class distribu-
tion lˆi = f (hi ). Since these sentences have specific labels, we optimize the cross
entropy loss as supervised loss term:
X
LS (si , li ) = − li log f (g(si )) (1)
6 C. Jiaao and W. Yuwei et al.
(a) Before Augmentation
(b) After Augmentation
Fig. 3. Distributions before and after performing augmentations over labeled sentences
belonging to General support, Info support and Emo support. 0 means sentences don’t
use corresponding types of support, while 1 represents sentences use corresponding
types of support. y axis is the number of sentences.
4.2 Unsupervised Learning
Paraphrasing Unlabeled Sentences We first performed back translations
once for each unlabeled sentence sui ∈ Su to generate the augmented sentence
set Su,a = {su,a u,a
1 , ..., sm } in the same manner we described before.
Guessing Labels for Unlabeled Sentences For an unlabeled sentence sui ,
we utilized g(.) and f (.) in Section 4.1 to predict the class distribution:
ˆlu = f (g(su )) (2)
i i
To avoid the prediction being so close to uniform distribution, we generate low-
entropy guessing labels ˜liu by a sharpening function [1]:
1
˜lu = (ˆliu ) T
i 1 (3)
||(ˆlu ) T ||1
i
Semi-Supervised Models via Data Augmentation 7
where ||.||1 is l1 -norm of the vector. When T → 0, the guessed label becomes an
one-hot vector.
Self-training for Original Sentences Inspired by self-training where model is
also trained over unlabeled data with high-confidence predictions as their labels,
in SMDA, with our guessed labels ˜liu with respect to original unlabeled sentence
sui , we added such pair (sui , ˜liu ) into training by minimize the KL Divergence
between them:
Ls (sui ) = KL(f (g(sui ))||˜liu ) (4)
Entropy Minimization for Original Sentences One common assumption in
many semi-supervised learning methods is that a classifier’s decision boundary
should not pass through high-density regions of the marginal data distribution
[5]. Thus for original unlabeled sentence sui , we added another loss term to
minimize the entropy of model’s output:
X
Le (sui ) = − f (g(sui )) log f (g(sui )) (5)
Consistency Regularization for Augmented Sentences With the assump-
tion that the model should predict similar distributions with input sentences be-
fore and after augmentations, we minimized the KL Divergence between outputs
with original sentence sui as input and augmented sentence su,ai as input:
Lc (sui ) = KL(ˆliu ||f (g(su,a
i ))) (6)
Combining all the loss terms for unlabeled sentences, we defined our unsuper-
vised loss terms as:
LU (sui ) = Ls (sui ) + Le (sui ) + Lc (sui ) (7)
4.3 Semi-Supervised Objective Function
We combined the supervised and unsupervised learning described above to form
our overall semi-supervised objective function:
L = E(si ,li )∈(S,L) LS (si , li ) + γEsui ∈Su LU (sui ) (8)
where γ is the balanced weight between supervised and unsupervised loss term.
5 Experiments
5.1 Model Setup
In SMDA 5 , we only used single model for each task without jointly training and
parameter sharing. That is, we trained six separate classifiers on these tasks.
5
The codes and data split will be released later.
8 C. Jiaao and W. Yuwei et al.
Inspired by recent success in pre-trained language models, we utilized the pre-
trained weights of XLNet and followed the same fine-tuning procedure as XLNet.
We set the initial learning rate for XLNet encoder as 1e-5 and other linear layers
as 1e-3. The batch size was selected in {32, 64, 128, 256}. The maximum number
of epochs is set as 20. Hyper-parameters were selected using the performance
on development set. The sharpen temperature T was selected in {0.3, 0.5, 0.8}
depending on different tasks. The balanced weight γ between supervised learning
loss and unsupervised learning loss term started from a small number and grew
through training process to 1.
5.2 Results
Our experimental results are shown in Table 3. We compared our proposed
SMDA with BERT and XLNet in terms of accuracy(%) and Macro F1 score.
BERT and XLNet achieved similar performance since they both obey the pre-
training and fine-tuning manner. When combining with augmented and more
balanced labeled data, massive unlabeled data, our SMDA achieved best perfor-
mance across six binary-classification tasks. And we submitted the classification
results on given unlabeled test set.
Table 3. Results on test set. Our baseline is our implementation of XLNet-cased-base.
Task Emo disc Info disc Support Gen supp Info supp Emo supp
acc F1 acc F1 acc F1 acc F1 acc F1 acc F1
BERTBASE 71.3 65.7 71.1 68.7 81.9 75.6 90.6 63.9 88.9 69.8 92.9 73.8
XLNetBASE 72.4 67.9 72.2 69.3 83.4 77.3 92.7 65.0 87.9 70.3 93.4 73.8
SMDA 75.2 68.5 74.3 71.0 83.5 77.7 91.7 63.7 89.9 70.5 93.6 76.2
6 Conclusion
In this work, we focused on identifying disclosure and supportiveness in conver-
sation responses based on a small labeled and large unlabeled training data via
our proposed semi-supervised text classification system : Semi-Supervised Mod-
els via Data Augmentation (SMDA). SMDA utilized supervised learning over
labeled data and conducted self-training, entropy minimization and consistency
regularization over unlabeled data. Experimental results demonstrated that our
system outperformed baseline models significantly.
References
1. Berthelot, D., Carlini, N., Goodfellow, I.J., Papernot, N., Oliver, A., Raf-
fel, C.: Mixmatch: A holistic approach to semi-supervised learning. CoRR
abs/1905.02249 (2019)
Semi-Supervised Models via Data Augmentation 9
2. Chen, M., Tang, Q., Livescu, K., Gimpel, K.: Variational sequential labelers for
semi-supervised learning. In: Proc. of EMNLP (2018)
3. Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: 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
Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers).
pp. 4171–4186. Association for Computational Linguistics, Minneapolis, Minnesota
(Jun 2019)
4. Ernala, S.K., Rizvi, A.F., Birnbaum, M.L., Kane, J.M., De Choudhury, M.:
Linguistic markers indicating therapeutic outcomes of social media disclosures
of schizophrenia. Proceedings of the ACM on Human-Computer Interaction
1(CSCW), 43 (2017)
5. Grandvalet, Y., Bengio, Y.: Semi-supervised learning by entropy minimization. In:
Proceedings of the 17th International Conference on Neural Information Processing
Systems. pp. 529–536. NIPS’04, MIT Press, Cambridge, MA, USA (2004)
6. Gururangan, S., Dang, T., Card, D., Smith, N.A.: Variational pretraining for semi-
supervised text classification. CoRR abs/1906.02242 (2019)
7. Howard, J., Ruder, S.: Universal language model fine-tuning for text classification.
In: Proceedings of the 56th Annual Meeting of the Association for Computational
Linguistics (Volume 1: Long Papers). pp. 328–339. Association for Computational
Linguistics, Melbourne, Australia (Jul 2018)
8. Jaidka, K., Singh, I., Jiahui, L., Chhaya, N., Ungar, L.: A report of the CL-Aff
OffMyChest Shared Task at Affective Content Workshop @ AAAI. In: Proceedings
of the 3rd Workshop on Affective Content Analysis @ AAAI (AffCon2020). New
York, New York (February 2020)
9. Kumar, A., Bhattamishra, S., Bhandari, M., Talukdar, P.: Submodular
optimization-based diverse paraphrasing and its effectiveness in data augmenta-
tion. In: Proceedings of the 2019 Conference of the North American Chapter of the
Association for Computational Linguistics: Human Language Technologies, Vol-
ume 1 (Long and Short Papers). pp. 3609–3619. Association for Computational Lin-
guistics, Minneapolis, Minnesota (Jun 2019). https://doi.org/10.18653/v1/N19-
1363, https://www.aclweb.org/anthology/N19-1363
10. Lample, G., Conneau, A.: Cross-lingual language model pretraining. CoRR
abs/1901.07291 (2019)
11. Lee, D.H.: Pseudo-label : The simple and efficient semi-supervised learning method
for deep neural networks. ICML 2013 Workshop : Challenges in Representation
Learning (WREPL) (07 2013)
12. Meng, Y., Shen, J., Zhang, C., Han, J.: Weakly-supervised neural text classifica-
tion. In: Proceedings of the 27th ACM International Conference on Information
and Knowledge Management. pp. 983–992. CIKM ’18, ACM, New York, NY, USA
(2018). https://doi.org/10.1145/3269206.3271737, http://doi.acm.org/10.1145/
3269206.3271737
13. Miyato, T., Dai, A.M., Goodfellow, I.: Adversarial training methods for semi-
supervised text classification. In: International Conference on Learning Represen-
tations (2017)
14. Miyato, T., Maeda, S., Koyama, M., Ishii, S.: Virtual adversarial training: A regu-
larization method for supervised and semi-supervised learning. IEEE Trans. Pat-
tern Anal. Mach. Intell. 41(8), 1979–1993 (2019)
15. Peters, M., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., Zettlemoyer,
L.: Deep contextualized word representations. In: Proceedings of the 2018 Confer-
10 C. Jiaao and W. Yuwei et al.
ence of the North American Chapter of the Association for Computational Lin-
guistics: Human Language Technologies, Volume 1 (Long Papers). pp. 2227–2237.
Association for Computational Linguistics, New Orleans, Louisiana (Jun 2018)
16. Warriner, A.B., Shore, D.I., Schmidt, L.A., Imbault, C.L., Kuperman, V.:
Sliding into happiness: A new tool for measuring affective responses to words.
Canadian Journal of Experimental Psychology/Revue canadienne de psy-
chologie expérimentale 71(1), 71 (2017). https://doi.org/10.1037/cep0000112,
https://app.dimensions.ai/details/publication/pub.1084126691andhttp:
//europepmc.org/articles/pmc5334777?pdf=render
17. Wei, J.W., Zou, K.: EDA: easy data augmentation techniques for boosting perfor-
mance on text classification tasks. CoRR abs/1901.11196 (2019)
18. Xie, Q., Dai, Z., Hovy, E., Luong, M.T., Le, Q.V.: Unsupervised data augmentation
for consistency training. arXiv preprint arXiv:1904.12848 (2019)
19. Yang, D., Kraut, R.E., Smith, T., Mayfield, E., Jurafsky, D.: Seekers, providers,
welcomers, and storytellers: Modeling social roles in online health communities. In:
Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems.
p. 344. ACM (2019)
20. Yang, D., Yao, Z., Seering, J., Kraut, R.: The channel matters: Self-disclosure,
reciprocity and social support in online cancer support groups. In: Proceedings of
the 2019 CHI Conference on Human Factors in Computing Systems. p. 31. ACM
(2019)
21. Yang, Z., Dai, Z., Yang, Y., Carbonell, J.G., Salakhutdinov, R., Le, Q.V.: Xl-
net: Generalized autoregressive pretraining for language understanding. CoRR
abs/1906.08237 (2019)
22. Yang, Z., Hu, Z., Salakhutdinov, R., Berg-Kirkpatrick, T.: Improved variational au-
toencoders for text modeling using dilated convolutions. CoRR abs/1702.08139
(2017)
23. Zhang, P.: The affective response model: A theoretical framework of
affective concepts and their relationships in the ict context. Manage-
ment Information Systems Quarterly (MISQ) 37, 247–274 (03 2013).
https://doi.org/10.25300/MISQ/2013/37.1.11