Censorship of Online Encyclopedias: Implications for NLP Models

Eddie Yang; Margaret E. Roberts

RECORD

Censorship of Online Encyclopedias: Implications for NLP Models

Eddie Yang, Margaret E. Roberts

Computation and Language · Submitted 22 Jan 2021

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim Our paper shows how government repression, censorship, and self-censorship may impact training data and the applications that draw from them.

how it is stated · hedged possibility 0.82 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 4,979 input tokens.

What the abstract itself states, as probabilities: claims first 0.12 · states limitation 0.10 · human participants 0.06 · controlled comparison 0.86 · randomized assignment 0.01 · states quantitative result 0.04 · compares with prior results 0.68 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

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.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	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.	background 1.00	0.04	0.04	asserted 0.98
s1	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.	gap 1.00	0.04	0.04	asserted 0.99
s2	We describe how censorship has affected the development of Wikipedia corpuses, text data which are regularly used for pre-trained inputs into NLP algorithms.	contribution 0.97	0.77	0.07	asserted 0.98
s3	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.	result 0.99	0.95	0.46	asserted 1.00
s4	We examine the implications of these discrepancies by studying their use in downstream AI applications.	contribution 0.71	0.86	0.06	unclear 0.51
s5	Our paper shows how government repression, censorship, and self-censorship may impact training data and the applications that draw from them.	interpretation 0.76	0.81	0.78	hedged possibility 0.82

What kind of paper the classifier reports this is: empirical study 1.00 · evidence from analysis of existing data 0.74.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

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.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

What this paper does with the arXiv papers it cites

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

What it cites · 5 sentences without a confident label

Uses data or values from 1 paper

We use word embeddings made available by [ref] .[<https://github.com/Embedding/Chinese-Word-Vectors>] [ref] train 300-dimensional word embeddings on both Baidu Baike and Chinese language Wikipedia using the same algorithm, Word2Vec [ref] .

Analogical Reasoning on Chinese Morphological and Semantic Relations · Distance from Democracy: Comparison Between Baidu Baike and Wikipedia Embeddings · 0.80

Cites as support 1 paper

Pre-trained word embeddings have been shown to achieve higher accuracy faster [ref] .

When and Why are Pre-trained Word Embeddings Useful for Neural Machine Translation? · Pre-Trained Word Embeddings and Wikipedia Corpuses · 0.98

Cites as background 1 paper

The literature in computer science has taken on the problem of bias in training data by looking for ways to de-bias it – for example, through data augmentation [ref] , de-biasing word embeddings [ref] , and adversarial learning [ref] .[Although methods for de-biasing have also been shown to often be inadequate [ref] .] However, it is unclear how to think about de-biasing attitudes toward democracy, freedom, surveillance, and social control.

Language (Technology) is Power: A Critical Survey of "Bias" in NLP · Conclusion · 0.88

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.