Analogical Reasoning on Chinese Morphological and Semantic Relations

Shen, Li; Zhao, Zhe; Hu, Renfen; Wensi, Li; Liu, Tao; Du, Xiaoyong

doi:10.48550/arxiv.1805.06504

Cited by 26 publications

(28 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Main results on the text dataset THUNEWS-LT We choose a commonly-used subset of Chinese news dataset THUCNews (Sun et al 2016) which contains ten classes 1 . We build its long-tailed version with the imbalance factor at 100 and use the pre-trained embedding model provided by Li et al (2018). We choose three models including TextCNN (Kalchbrenner, Grefenstette, and Blunsom 2014), TextRNN (Liu, Qiu, and Huang 2016) and Transformer (Vaswani et al 2017) as the backbone.…”

Section: Main Results On Inaturalist Tablementioning

confidence: 99%

Label-Aware Distribution Calibration for Long-tailed Classification

Wang¹,

Shuzheng²,

Gao³

et al. 2021

Preprint

View full text Add to dashboard Cite

Real-world data usually present long-tailed distributions. Training on imbalanced data tends to render neural networks perform well on head classes while much worse on tail classes. The severe sparseness of training instances for the tail classes is the main challenge, which results in biased distribution estimation during training. Plenty of efforts have been devoted to ameliorating the challenge, including data re-sampling and synthesizing new training instances for tail classes. However, no prior research has exploited the transferable knowledge from head classes to tail classes for calibrating the distribution of tail classes. In this paper, we suppose that tail classes can be enriched by similar head classes and propose a novel distribution calibration approach named as Label-Aware Distribution Calibration (LADC). LADC transfers the statistics from relevant head classes to infer the distribution of tail classes. Sampling from calibrated distribution further facilitates rebalancing the classifier. Experiments on both image and text long-tailed datasets demonstrate that LADC significantly outperforms existing methods. The visualization also shows that LADC provides a more accurate distribution estimation.

show abstract

Section: Main Results On Inaturalist Tablementioning

confidence: 99%

Label-Aware Distribution Calibration for Long-tailed Classification

Wang¹,

Shuzheng²,

Gao³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The collection dataset is obtained on a MOOC platform with complex data dimensions. Therefore, a large Chinese corpus is used to capture the semantic information between different words [28]. Besides, we use the Word2Vec tool to convert the text corpus as input and convert the title and course profile content into a 300-dimensional feature vector.…”

Section: Text Feature Extractionmentioning

confidence: 99%

A Deep Learning Framework for Multimodal Course Recommendation Based on LSTM+Attention

et al. 2022

View full text Add to dashboard Cite

With the impact of COVID-19 on education, online education is booming, enabling learners to access various courses. However, due to the overload of courses and redundant information, it is challenging for users to quickly locate courses they are interested in when faced with a massive number of courses. To solve this problem, we propose a deep course recommendation model with multimodal feature extraction based on the Long- and Short-Term Memory network (LSTM) and Attention mechanism. The model uses course video, audio, and title and introduction for multimodal fusion. To build a complete learner portrait, user demographic information, explicit and implicit feedback data were added. We conducted extensive and exhaustive experiments based on real datasets, and the results show that the AUC obtained a score of 79.89%, which is significantly higher than similar algorithms and can provide users with more accurate recommendation results in course recommendation scenarios.

show abstract

“…Perhaps because of the size difference between the two corpuses, increasingly researchers developing cutting edge Chinese langauge NLP models are drawing on the Baidu Baike corpus [38,43]. Baidu Baike word embeddings have been shown to perform better on certain tasks [21]. Here, we assess the downstream implications of this choice on the representation of democratic concepts, social control, and historical events and figures.…”

Section: Censorship Of Chinese Language Wikipedia and Implications Fo...mentioning

confidence: 99%

“…In this section, we consider the differences in word associations among word embeddings trained with Chinese language Wikipedia and Baidu Baike. We use word embeddings made available by Li et al [21]. 4 Li et al [21] train 300-dimensional word embeddings on both Baidu Baike and Chinese language Wikipedia using the same algorithm, Word2Vec [24].…”

Section: Distance From Democracy: Comparison Between Baidu Baike and ...mentioning

confidence: 99%

Censorship of Online Encyclopedias

Yang

Roberts

2021

Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency

View full text Add to dashboard Cite

While artificial intelligence provides the backbone for many tools people use around the world, recent work has brought to attention that the algorithms powering AI are not free of politics, stereotypes, and bias. While most work in this area has focused on the ways in which AI can exacerbate existing inequalities and discrimination, very little work has studied how governments actively shape training data. We describe how censorship has affected the development of Wikipedia corpuses, text data which are regularly used for pre-trained inputs into NLP algorithms. We show that word embeddings trained on Baidu Baike, an online Chinese encyclopedia, have very different associations between adjectives and a range of concepts about democracy, freedom, collective action, equality, and people and historical events in China than its regularly blocked but uncensored counterpart -Chinese language Wikipedia. We examine the implications of these discrepancies by studying their use in downstream AI applications. Our paper shows how government repression, censorship, and self-censorship may impact training data and the applications that draw from them. CCS CONCEPTS• Computing methodologies → Supervised learning by classification; • Information systems → Content analysis and feature selection; • Social and professional topics → Political speech.

show abstract

Analogical Reasoning on Chinese Morphological and Semantic Relations

Cited by 26 publications

References 9 publications

Label-Aware Distribution Calibration for Long-tailed Classification

Label-Aware Distribution Calibration for Long-tailed Classification

A Deep Learning Framework for Multimodal Course Recommendation Based on LSTM+Attention

Censorship of Online Encyclopedias

Contact Info

Product

Resources

About