Super Characters: A Conversion from Sentiment Classification to Image Classification

Sun, Baoqing; Yang, Lin; Dong, Patrick; Zhang, Wenhan; Dong, Jason; Young, Charles

doi:10.18653/v1/w18-6245

Cited by 25 publications

(16 citation statements)

References 28 publications

(28 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For fine-tuning all models on a target task, we follow the same parameters that were used in the original implementation. 121314 (2017) and Sun et al (2018) 6 Results…”

Section: Fine-tuningmentioning

confidence: 99%

“…To the best of our knowledge, the current state-of-theart for Japanese text classification uses shallow (context-free) word embeddings for text classification (Peinan and Mamoru, 2015;Nio and Murakami, 2018). Sun et al (2018) proposed the Super Characters method that converts sentence classification into image classification by projecting text into images. Zhang and LeCun (2017) did an extensive study of different ways of encoding Chinese/Japanese/Korean (CJK) and English languages, covering 14 datasets and 473 combinations of different encodings including one-hot, character glyphs, and embeddings and linear, fasttext and CNN models.…”

Section: Text Classificationmentioning

confidence: 99%

“…This paper investigates transfer learning-based methods for sentiment analysis that is comparable to above mentioned models including Zhang and LeCun (2017) and Sun et al (2018) for the Japanese language. Pre-training a language model is the most expensive part but we train it only once and fine-tune on a target task.…”

Section: Text Classificationmentioning

confidence: 99%

See 2 more Smart Citations

An Investigation of Transfer Learning-Based Sentiment Analysis in Japanese

Bataa¹,

Wu²

2019

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

Text classification approaches have usually required task-specific model architectures and huge labeled datasets. Recently, thanks to the rise of text-based transfer learning techniques, it is possible to pre-train a language model in an unsupervised manner and leverage them to perform effectively on downstream tasks. In this work we focus on Japanese and show the potential use of transfer learning techniques in text classification. Specifically, we perform binary and multi-class sentiment classification on the Rakuten product review and Yahoo movie review datasets. We show that transfer learning-based approaches perform better than task-specific models trained on 3 times as much data. Furthermore, these approaches perform just as well for language modeling pre-trained on 1 30 of Wikipedia. We release our pre-trained models and code as open source. 10 github.com/google-research/bert/issues/130 11 nlpprogress.com/english/sentiment analysis.html 12 github.com/fastai/fastai 13 github.com/allenai/allennlp 14 github.com/google-research/bert#fine-tuning-with-bert 15 cloud.google.com/tpu/

show abstract

“…For fine-tuning all models on a target task, we follow the same parameters that were used in the original implementation. 121314 (2017) and Sun et al (2018) 6 Results…”

Section: Fine-tuningmentioning

confidence: 99%

Section: Text Classificationmentioning

confidence: 99%

See 1 more Smart Citation

An Investigation of Transfer Learning-Based Sentiment Analysis in Japanese

Bataa¹,

Wu²

2019

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

show abstract

“…Compared with CNN and SVM, the Bi-LSTM model shows good performance in this experiment. However, compared with the current SOTA (State of the Arts) model in the field of sentiment analysis, there are still some deficiencies [55] 85 It can be seen from the above table that the Bi-LSTM model mentioned in this paper has a certain gap in Accuracy, Recall, F1-Score and other indicators compared with SOTA. The reason for the difference in accuracy should be attributed to the insufficient number of samples in the training set, and the quality is also a factor limiting the accuracy.…”

Section: Comparative Experiments With Sotamentioning

confidence: 93%

Improved Danmaku Emotion Analysis and Its Application Based on Bi-LSTM Model

et al. 2020

View full text Add to dashboard Cite

With the rapid development of social media, danmaku video provides a platform for users to communicate online. To some extent, danmaku video provides emotional timing information and an innovative method to analyze video data. In the age of big data, studying the characteristics of danmaku and its emotional tendencies can not only help us understand the psychological characteristics of users but also feedback the effective information of users to video platforms, which can help the platforms optimize related short video recommendations so that it can provide a more accurate solution for the selection of audiences during video production. However, danmaku is different from traditional comments. Current emotion classification methods are only suitable for two-dimensional classification which are not suitable for danmaku emotion analysis. Aiming at the problems such as the colloquialism, diversity, spelling errors, structural non-linearity informal language on the Internet, diversity of social topics, and context dependency of emotion analysis of the danmaku data, this paper proposes an improved emotion analysis model based on Bi-LSTM model to classify the further four-dimensional emotions of Pleasure, Anger, Sorrow and Joy. Furthermore, we add tags such as comment time and user name to the danmaku information. Experimental results show that the improved model has higher Accuracy, Recall, Precision, and F1-Score under the same conditions compared with the CNN and SVM. The classification effect of improved model is close to the SOTA. Experimental results also show that the improved model can be effectively applied to the analysis of irregular danmaku emotion.

show abstract

“…However, this reliance upon one-dimensional embeddings may soon come to change. Recent NLP research has shown that the two-dimensional embedding of the Super Characters method [33] is capable of achieving stateof-the-art results on large dataset benchmarks. The Super Characters method is a two-step method that was initially designed for text classification problems.…”

Section: Introductionmentioning

confidence: 99%

SuperTML: Two-Dimensional Word Embedding for the Precognition on Structured Tabular Data

Sun¹,

Yang²,

Zhang³

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

Self Cite

View full text Add to dashboard Cite

Tabular data is the most commonly used form of data in industry according to a Kaggle ML and DS Survey. Gradient Boosting Trees, Support Vector Machine, Random Forest, and Logistic Regression are typically used for classification tasks on tabular data. DNN models using categorical embeddings are also applied in this task, but all attempts thus far have used one-dimensional embeddings. The recent work of Super Characters method using twodimensional word embeddings achieved state-of-the-art results in text classification tasks, showcasing the promise of this new approach. In this paper, we propose the Su-perTML method, which borrows the idea of Super Characters method and two-dimensional embeddings to address the problem of classification on tabular data. For each input of tabular data, the features are first projected into twodimensional embeddings like an image, and then this image is fed into fine-tuned two-dimensional CNN models for classification. The proposed SuperTML method handles the categorical data and missing values in tabular data automatically, without any need to pre-process into numerical values. Comparisons of model performance are conducted on one of the largest and most active competitions on the Kaggle platform, as well as on the top three most popular data sets in the UCI Machine Learning Repository. Experimental results have shown that the proposed SuperTML method have achieved state-of-the-art results on both large and small datasets.

show abstract

Super Characters: A Conversion from Sentiment Classification to Image Classification

Cited by 25 publications

References 28 publications

An Investigation of Transfer Learning-Based Sentiment Analysis in Japanese

An Investigation of Transfer Learning-Based Sentiment Analysis in Japanese

Improved Danmaku Emotion Analysis and Its Application Based on Bi-LSTM Model

SuperTML: Two-Dimensional Word Embedding for the Precognition on Structured Tabular Data

Contact Info

Product

Resources

About