Wikipedia:Wikipedia Signpost/Single/2024-08-14

The Oregonian reported that Portland city commissioner Rene Gonzalez spent $6,400 of city taxpayer dollars "to spruce up his profile on Wikipedia" as part of his mayoral bid, by hiring a contractor, WhiteHatWiki, who "helped craft eight requested edits" which were then posted on the article's talk page by a staffer. Only half of these were approved by the volunteer editor who reviewed the request, with one of the rejected ones asking for the removal of a mention that "Gonzalez tagged a member of the right-wing group Patriot Prayer in a Twitter post thanking supporters after his race for City Council in 2022."

In contrast, the newspaper reports that neither Portland's current mayor nor any of Gonzalez' colleagues on the Portland City Council "have paid money to spruce up their Wikipedia entries, according to their offices." It also quotes a political consultant calling the practice "highly unusual" ("I haven’t seen that before"). However, Gonzalez’s chief of staff offered what The Oregonian called "a full-throated defense" of the practice, arguing for a need to be "innovative in how we manage our public profile and how we invest in educating our staff."

Unfortunately this is only one of many such incidents recorded at Conflict-of-interest editing on Wikipedia. – B, HaeB

R&D World says you can "write a Wikipedia-style article draft in a few minutes for less than a penny using STORM". STORM stands for 'Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking', and is described as "an open-source artificial intelligence system that promises to generate Wikipedia-style articles on pretty much any topic using large language models and web search."

The R&D World author tested this promise, using the topic of "double descent." Did it work? On the plus side, STORM quickly produced a Wikipedia-like article at a cost of about half a cent. To this non-expert on the topic, it appears to be at least OK. On the negative side, STORM had the advantage that Wikipedia already had an article on double descent, which STORM used to create a new article. For its next trick, I suggest R&D World try to create an article about Bronx cheer.

Fortunately, STORM was written by a team from Stanford University, not from R&D World. See further coverage in this issue's Recent research. – S

Stephen Harrison in Slate covers the oldest living Olympians, while noting that they are no longer automatically considered to be notable and deserving of an English Wikipedia article unless they've won a medal. His source material is from Paul Tchir, a San Diego State University sports historian, which can be viewed here.

The New York Times reports on Earl M. Washington, a convicted art forger who is now serving a 52-month term in Federal prison. Washington sold woodblock prints and the intricately carved woodblocks themselves, "more than 3,000 blocks and more than a million prints," sometimes claiming they were antiques dating back to the 16th–17th centuries. After a 2004 Forbes magazine questioned the authenticity of the prints and reported accusations of Washington copying M.C. Escher prints, Washington took a break from the scam until about 2010.

Dr. Douglas Arbittier, who owns a private museum of antique medical instruments, bought 130 prints from Washington from 2013 to 2016 for about $118,810, according to these articles. Washington and Arbittier then lost contact for a few years. Arbittier began to suspect that the works were forgeries. Around 2018, he began a Javert-like pursuit of information about Washington and his forgeries.

In 2020, Arbittier read the Wikipedia article about Washington, which at the time looked like this. "The world comes crashing down at that point," he told the Times. "It was gut wrenching because, oh my God, why did I spend all that money, but also it was a betrayal of the trust and relationship that we had."

Soon he sent a 286-page report to the FBI. Washington was indicted in January 2023, later reaching a plea deal and confessing to reduced charges last summer. He was then sentenced in April 2024.

The Wikipedia article's history, as checked by this reporter, is quite surprising. It was created in 2006, based on the 2004 Forbes story. It has remained quite critical of Washington since then and readers would have seen that Washington's honesty had been questioned. About 2008, the article explicitly included accusations that Washington had forged M.C. Escher prints. Two deletion discussions (in 2006 and 2008, respectively) made clear that Washington might be a scammer... but was he a notable scammer? Washington was accused several times in edit comments and on the talk page of editing or whitewashing the article himself. There were several clumsy attempts to remove negative details in the article, but none of them approached a complete whitewashing. – S

Three center-right Jewish media outlets have criticized the English Wikipedia's coverage of the Israel–Hamas war as being biased against Israel.

The Jewish News Syndicate (JNS) says, "Wikipedia hates Israel and Jews":

Wikipedia’s antisemitism is practically ubiquitous across the website. It features an extensive article accusing Israel of “war crimes,” “indiscriminate attacks” and “genocide” as Israel seeks to eliminate the Hamas terror organization.

The JNS article goes on to portray Wikipedia and Wikimedia as "big tech's antisemitic propaganda arm":

The connection between Wikipedia and Big Tech is easy to establish. Though Wikipedia likes to beg regular users for money, its Wikimedia Foundation (WMF) features a who’s who of Big Tech donors: Apple, Google, Facebook, Microsoft, Adobe, Salesforce and more. Besides direct grants from these firms, woke programmers, engineers and other staff at companies like Apple and Google, as well as LinkedIn, Intel and Netflix have used matching gift programs to multiply their contributions.

The truth is that WMF, together with the Wikimedia Endowment, holds over $350 million in assets. This means that with no further donations or investments, Wikipedia can continue operating comfortably for over a century. Yet its relationship with Big Tech has only deepened and diversified.

In 2021, Wikimedia launched Wikimedia Enterprise, providing paid services for companies and organizations that reuse Wikipedia content on a large scale. A routine search using Google, Alexa or Siri often brings you a highlighted result drawn from Wikipedia like the Google “Knowledge Panel” at the top of search results. This is, in part, why a Google search for "apartheid" features recurring instances of antisemitic fiction on its first page of results.

Besides Big Tech, there is one more Wikipedia donor to consider: the Soros-funded Tides Foundation. Tides has also given millions of dollars to groups that instigated and supported the antisemitic protests across America since Oct. 7. This is the company Wikimedia keeps.

The article concludes by calling on Big Tech to rein Wikipedia in.

Another article, in Tablet magazine, comments on the English Wikipedia community's recent decision to designate the Anti-Defamation League "generally unreliable" on matters pertaining to the Israel-Palestinian conflict. The Tablet writer expresses the view that –

Wikipedia's articles are now badly distorted, feeding billions of people—and large-language models that regularly train on the site, such as ChatGPT—with inaccurate research and dangerously skewed narratives about Jews, Jewish history, Israel, Zionism, and contemporary threats to Jewish lives.

The article also reviews the Grabowski/Klein paper (see previous Signpost coverage: 1, 2) and likens the present situation in the English Wikipedia to the historic right-wing takeover of the Croatian Wikipedia, describing it as "incomprehensible" that the Wikimedia Foundation took so long to address that situation.

In a third article, the Jewish Journal gave a very detailed description of the move discussion for the article Allegations of genocide in the 2023 Israeli attack on Gaza, which resulted in that article being renamed Gaza genocide. The Journal notes that the article Allegations of genocide in the 2023 Hamas-led attack on Israel continues to feature the word "allegations" in its title and wonders if this indicates a double standard, pointing out that a discussion to remove the word from that article's title as well appears to have stalled.

The Journal quotes two Wikipedians arguing that there is such a double standard and a third arguing that it may not necessarily be a double standard "if the academic sources don't refer to the Oct. 7 massacre as a genocide but do refer to Israel's actions in Gaza as such." The article ends with a discussion of what Middle East scholar Asaf Romirowsky views as the "Palestinization of the Academy", which he says has led to a problematic focus on Diversity, Equity and Inclusion (DEI) narratives where Palestinians are portrayed as victims worthy of support rather than as perpetrators of crimes. The Journal quotes a Wikipedian who told the publication:

"The main problem is that since academia is biased against Israel, and Wikipedia sourcing policies give deference to academia, such opinions find their way into Wikipedia articles and it is hard to counteract. It is a kind of closed loop of bias and misinformation, much as would have happened if the Nazis had won the war and taken over universities and think tanks. We are at about that point with Hamas and its allies. I think that even if Wikipedia editors wanted to be NPOV (which is a fiction) it would be hard."

Another quoted in the article sees the greatest problem with Wikipedia in its being "based on academic and journalistic sources, and neither of them are particularly good"; the solution, in their view, is changing the sources Wikipedia is working from. – AK

Ars Technica reports on the Wikimedia Foundation's recent rollout of dark mode for Wikipedia readers on desktop and mobile, evidently having some gloomy fun while crafting the headline ("Darkness reigns over Wikipedia as official dark mode comes to pass"). Ars points out that Wikipedia is a bit late to the dark mode trend, which "had something of a peak moment around 2019–2020." However, it notes that implementation of this feature is much more difficult than may seem on first glance, quoting from a detailed explanation by Redditor (and Wikipedian) Gwern: "It's truly one of those things where you can get 95% of the way by simply adding 1 line of CSS like body{filter: invert(100%);}, but then to get to 99% correctness and squash all the annoying bugs, you have to completely rewrite your entire site design, and getting to 100% is impossible." See also the Wikimedia Foundation's summary of the process from late 2023. —H

Editor's note: Signpost stylesheets are part of that final one percent, if anyone wants to help out! —J

• Internet in a box: Boing Boing says to go buy Wikipedia in a box (see also Internet-in-a-Box) from WMF's store. The price is $58, but as of August 2 they are sold out, though the elves are working overtime to get them back in stock.
• Borderline: The Washington Examiner notes that there have been edit wars over whether U.S. Vice President Kamala Harris should or should not be included in the List of U.S. executive branch czars, based on past descriptions of her – especially by Republicans – as a "border czar" or "immigration czar". (Status at the time of writing is that Harris is not so listed in the article.)
• Unexpected Source of Motivation: Reuters reports that Olympic gymnast Max Whitlock used Wikipedia for motivation after a mental health crisis following the 2020 Tokyo Olympics. He used the site to confirm that he could break a record at the 2024 Olympics by winning a medal in pommel horse. Unfortunately, he had two fourth place finishes.
• Unprintable: The Daily Mail, Sky News, and the Daily Mail again report that a television presenter's Wikipedia article was vandalized for a sum total of 31 minutes. The revision's been revdeled as libelous, but the reversion summary "you would need evidence that he is best known for this" gives a general idea of what it was.
• Edit-a-thon in Canberra: The Canberra Times reports on an editathon held by Franklin Women at Canberra's Shine Dome.
• How the Regime Captured Wikipedia : Pirate Wires avers that the Wikimedia Foundation has transformed Wikipedia into a "hyper-centralized space of top-down social justice activism and advocacy", and weighs in with gusto on the Framgate incident of 2019.

—J

A monthly overview of recent academic research about Wikipedia and other Wikimedia projects, also published as the Wikimedia Research Newsletter.

A paper^[1] presented in June at the NAACL 2024 conference describes "how to apply large language models to write grounded and organized long-form articles from scratch, with comparable breadth and depth to Wikipedia pages." A "research prototype" version of the resulting "STORM" system is available online and has already attracted thousands of users. This is the most advanced system for automatically creating Wikipedia-like articles that has been published to date.

The authors hail from Monica S. Lam's group at Stanford, which has also published several other papers involving LLMs and Wikimedia projects since 2023 (see our previous coverage: WikiChat, "the first few-shot LLM-based chatbot that almost never hallucinates" – a paper that received the Wikimedia Foundation's "Research Award of the Year" some weeks ago).

Research into automated generation of Wikipedia-like text long predates the current AI boom fueled by the 2022 release of ChatGPT. However, the authors point out that such efforts have "generally focused on evaluating the generation of shorter snippets (e.g., one paragraph), within a narrower scope (e.g., a specific domain or two), or when an explicit outline or reference documents are supplied." (See below for some other recent publications that took such a more limited approach. For coverage of an antediluvian historical example, see a 2015 review in this newsletter: "Bot detects theatre play scripts on the web and writes Wikipedia articles about them". The STORM paper cites an even earlier predecessor from 2009, a paper titled "Automatically generating Wikipedia articles: A structure-aware approach", which resulted in this edit.)

The STORM authors tackle the more general problem of writing of a Wikipedia-like article about an arbitrary topic "from scratch". Using a novel approach, they break this down it into various tasks and sub-tasks, which are carried out by different LLM agents:

"We decompose this problem into two tasks. The first is to conduct research to generate an outline, i.e., a list of multi-level sections, and collect a set of reference documents. The second uses the outline and the references to produce the full-length article. Such a task decomposition mirrors the human writing process which usually includes phases of pre-writing, drafting, and revising [...]"

The use of external references is motivated by the (by now well-established) observation that relying on the "parametric knowledge" contained in the LLM itself "is limited by a lack of details and hallucinations [...], particularly in addressing long-tail topics". ChatGPT and other state-of-the art AI chatbots struggle with requests to create a Wikipedia article. (As Wikipedians have found in various experiments – see also the Signpost's November 2022 coverage of attempts to write Wikipediesque articles using LLMs – this may result e.g. in articles that look good superficially but contain lots of factually wrong statements supported by hallucinated citations, i.e. references to web pages or other publications that do not exist.) The authors note that "current strategies [to address such shortcomings of LLMs in general] often involve retrieval-augmented generation (RAG), which circles back to the problem of researching the topic in the pre-writing stage, as much information cannot be surfaced through simple topic searches." They cite existing "human learning theories" about the importance of "asking effective questions". This task in turn is likewise challenging for LLMs ("we find that they typically produce basic 'What', 'When', and 'Where' questions [...] which often only address surface-level facts about the topic".) This motivates the authors' more elaborated design:

"To endow LLMs with the capacity to conduct better research, we propose the STORM paradigm for the Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking.
The design of STORM is based on two hypotheses: (1) diverse perspectives lead to varied questions; (2) formulating in-depth questions requires iterative research."

"STORM models the pre-writing stage by (1) discovering diverse perspectives in researching the given topic, (2) simulating conversations where writers carrying different perspectives pose questions to a topic expert grounded on trusted Internet sources, (3) curating the collected information to create an outline."

In more detail, after being given a topic to write about, STORM first "prompts an LLM to generate a list of related topics and subsequently extracts the tables of contents from their corresponding Wikipedia articles, if such articles can be obtained through Wikipedia API". In an example presented by the authors, for the given topic sustainability of Large Language Models, this might lead to the existing articles sustainable development and corporate social responsibility. The section headings of those related articles are then passed to an LLM with the request to generate a set of "perspectives", with the prompt

You need to select a group of Wikipedia editors who will work together to create a comprehensive article on the topic . Each of them represents a different perspective , role , or affiliation related to this topic [...].

In the authors' example, one of the resulting perspectives is a "Social scientist – This editor will bring in a social perspective, focusing on topics such as social sustainability, cultural sustainability, and the role of nongovermental [sic] stakeholders in promoting sustainability."

Each of these "Wikipedia editors" then sets out to interview a "topic expert" in their field of interest, i.e. the system simulates a conversation between two LLM agents prompted to act in these roles. The "expert" is instructed to answer the "Wikipedia editor"'s questions by coming up with suitable search engine queries and looking through the results. From the various prompts involved:

You are an experienced Wikipedia writer and want to edit a specific page. Besides your identity as a Wikipedia writer, you have a specific focus when researching the topic. Now, you are chatting with an expert to get information. Ask good questions to get more useful information [...]

You want to answer the question using Google search. What do you type in the search box? [...]

You are an expert who can use information effectively. You are chatting with a Wikipedia writer who wants to write a Wikipedia page on topic you know. You have gathered the related information and will now use the information to form a response.

[...] Try to use as many different sources as possible and add do not hallucinate.

The online version of the STORM tool allows one to watch these behind-the-scenes agent conversations while the article is being generated, which can be quite amusing. (The "Wikipedia editor" is admonished in the prompt to politely express its gratitude to the "expert" and not to waste their time with repetitive questions: "When you have no more question to ask , say " Thank you so much for your help !" to end the conversation . Please only ask one question at a time and don 't ask what you have asked before .") The authors are currently working on a follow-up project called "Co-STORM" where the (human) user can become part of these multi-round agent conversation, e.g. to mitigate some remaining issues like content that is repetitive or conflicts between the different "experts".

(Like the aforementioned use of externally retrieved information, such agent-based systems have become quite popular in LLM-based AI over the last year or so. The authors use DSPy – a framework likewise developed at Stanford – for their implementation. Another well-known framework is LangChain, who actually released their own implementation of STORM as a demo of their "Langgraph" library back in February, based on the description and prompts in a preprint version of the paper, and shortly before the paper's authors published their own code.)

The paper states that the results of the "experts'" search engine queries "will be evaluated using a rule-based filter according to the Wikipedia guideline [ Wikipedia:Reliable sources ] to exclude untrustworthy sources" before the "experts" use them to generate their answers. (In the published source code, this is implemented in a somewhat simplistic way, by excluding those sources that Wikipedians have explicitly marked as "generally unreliable", "deprecated" or "blacklisted" at Wikipedia:Reliable sources/Perennial sources. But of course, search engine results contain many other sources on the internet that don't match the WP:RS requirements, either. In this reviewer's experiments with the STORM system, that turned out to be a significant limitation, at least if one were to use the output as basis for creating an actual Wikipedia article. One idea might be to restrict search to a search engine such as Google Scholar. But academic journal paywalls represent a challenge to this idea, according to a conversation with one of the authors.)

Having gathered material from those agent conversations, STORM proceeds to generating an outline for the article. First, the system prompts the LLM to draft the outline only based on its internal (parametric) knowledge, which "typically provides a general but organized framework." This is then refined based on the results of the perspective-based conversations.

Lastly, the system composes the full article section by section, using the outline and the set of all reference documents R collected by the "topic experts". Another complication here is that "since it is usually impossible to fit the entire R within the context window of the LLM, we use the section title and headings of its all-level subsections to retrieve relevant documents from R based on semantic similarity". The LLM is then prompted separately for each section to generate its text using the references selected for that section. The sections are then concatenated into a single document, which is passed once more to the LLM with a prompt asking it to remove duplications between the sections. Finally, the LLM is called one last time to generate a summary for the lead section.

All this internal chattiness and repeated prompting of the LLM for multiple tasks comes at a price. It typically costs about 84 cent in market price API fees to generate one article (when using OpenAi's top-tier model GPT 4.0 as the LLM, and including the cost of search engine queries), according to an estimate shared by one of the authors last month. However, the freely available research prototype of STORM is supported by free Microsoft Azure credits. (This reviewer incurred roughly comparable costs when trying out the aforementioned LangChain implementation, also using GPT 4.0.) On the other hand, a reviewer at the website "R&D World" (see coverage in this issue's "In the Media") reported getting "A draft article in minutes for $0.005" while running the STORM code on Google Colab (albeit possibly by relying on initial free credits from OpenAI too).

So are all these extra steps worth it, compared to simpler efforts (like asking ChatGPT "Write a Wikipedia article about...")?

First, to enable automated evaluation, the authors "curate FreshWiki, a dataset of recent high-quality Wikipedia articles, and formulate outline assessments to evaluate the pre-writing stage." The FreshWiki articles are used as ground truth, to "compute the entity recall in the article level" (very roughly, counting how many terms from the human-written reference article also occur in the auto-generated article about the topic) and the similar ROUGE-1 and ROUGE-L metrics (which measure the overlap with the reference text on the level of single words and word sequences).

The author compare their system to "three LLM-based baselines", e.g. "Direct Gen, a baseline that directly prompts the LLM to generate an outline, which is then used to generate the full-length article." They find that STORM indeed comes out ahead on these scores.

For manual evaluation, the authors invited [1] [2]

"a group of experienced Wikipedia editors for expert evaluation. The editors found STORM outperforms an outline-driven RAG baseline, especially regarding the breadth and organization of the articles. They also identified challenges for future research, including addressing cases where: (1) the bias on the Internet affects the generated articles; (2) LLMs fabricate connections between unrelated facts."

Another part of the automated evaluation checks whether the cited passages in the reference document actually support the sentence they are cited for. This problem is known as textual entailment in natural language processing. The authors entrust these checks to a current open-weight LLM (Mistral 7B-Instruct). This choice may be of independent interest to those seeking to use LLMs for automatically checking text-source integrity on Wikipedia.

They find that

"around 15% sentences in generated articles are unsupported by citations. We further investigate the failure cases by randomly sampling 10 articles and an author manually examines all the unsupported sentences in these articles. Besides sentences that are incorrectly split, lack citations, or are deemed supported by the author’s judgment [i.e. where Mistral 7B-Instruct incorrectly concluded that the citation had not supported the sentence], our analysis identifies three main error categories [...]: improper inferential linking, inaccurate paraphrasing, and citing irrelevant sources."

As a concrete example of such irrelevant sources, in this reviewer's test with creating an article on the German Press Council (Deutscher Presserat – a long-tail topic where not too many high-quality English-language online sources exist), the otherwise quite solid list of references included several pages about the wrong entity: One about the Luxembourgian press council, another about the unrelated German Ethics Council, and a third one about Germany and the UN Security Council. This seems primarily a failure in the search engine retrieval stage, rather than a LLM hallucination problem per se. But it was also not caught by the "topic experts" despite being prompted to "make sure every sentence is supported by the gathered information".)

The authors take care to avoid the impression that STORM's outputs can already match actual Wikipedia articles in all respects (only asserting that the generated articles have "comparable breadth and depth to Wikipedia pages"). Their research project page on Meta-wiki is diligently titled "Wikipedia type Articles Generated by LLM (Not for Publication on Wikipedia)". Nevertheless, STORM represents a significant step forward, bringing AI a bit closer to replacing much of the work of Wikipedia article writers.

On July 11, one of the authors presented the project at a Wikipedia meetup in San Francisco, and answered various questions about it (Etherpad notes). Among others aspects already reported above, he shared that STORM had already attracted around 10,000 users (signups) who use it for a variety of different uses cases – not just as a mere Wikipedia replacement. The project has received feature requests from various interested parties, which are being implementing by a small development team (3 people), as visible in the project's open-source code repository.

Other recent publications that could not be covered in time for this issue include the items listed below. Contributions, whether reviewing or summarizing newly published research, are always welcome.

From the abstract:^[2]

"[...] previous efforts in Wikipedia generation have often fallen short of meeting real-world requirements. Some approaches focus solely on generating segments of a complete Wikipedia document, while others overlook the importance of faithfulness in generation or fail to consider the influence of the pre-training corpus. In this paper, we simulate a real-world scenario where structured full-length Wikipedia documents are generated for emergent events [e.g. 2022 EFL League One play-off final] using input retrieved from web sources. To ensure that Large Language Models (LLMs) are not trained on corpora related to recently occurred events, we select events that have taken place recently and introduce a new benchmark Wiki-GenBen, which consists of 309 events paired with their corresponding retrieved web pages for generating evidence. Additionally, we design a comprehensive set of systematic evaluation metrics and baseline methods, to evaluate the capability of LLMs in generating factual full-length Wikipedia documents."

From the paper:

"Our experiments are conducted using two variants of ChatGPT: GPT-3.5-turbo and GPT-3.5- turbo-16k, as well as open-source LLMs, including instruction-tuned versions of LLama2"
"A notable challenge observed across all models is their struggle to maintain the reliability of the content produced. The best-performing models reach citation metrics just above 50% and an IB Score around 10%, highlighting the complexity involved in generating accurate and reliable content."

The authors are a group of ten researchers from Peking University and Huawei. Published just six days after (the first version of) the "STORM" paper by Stanford researchers covered above, neither of the two papers cites the other.

From the abstract:^[3]

"We show that recent advances in pretrained language modeling can be combined for a two-stage extractive and abstractive approach for Wikipedia lead paragraph generation. We extend this approach to generate longer Wikipedia-style summaries with sections and examine how such methods struggle in this application through detailed studies with 100 reference human-collected surveys. This is the first study on utilizing web resources for long Wikipedia-style summaries to the best of our knowledge."

From the abstract:^[4]

"we examine the proficiency of LLMs in generating succinct [Wikipedia-style] survey articles specific to the niche field of NLP in computer science, focusing on a curated list of 99 topics [adopted from the "Surfer100" dataset, see above]. Automated benchmarks reveal that GPT-4 surpasses its predecessors like GPT-3.5, PaLM2, and LLaMa2 in comparison to the established ground truth. We compare both human and GPT-based evaluation scores and provide in-depth analysis. While our findings suggest that GPT-created surveys are more contemporary and accessible than human-authored ones, certain limitations were observed. Notably, GPT-4, despite often delivering outstanding content, occasionally exhibited lapses like missing details or factual errors."

From the abstract:^[5]

"This paper presents a method for generating Wikipedia articles in the Hindi language automatically, using Wikidata as a knowledge base. Our method extracts structured information from Wikidata, such as the names of entities, their properties, and their relationships, and then uses this information to generate natural language text that conforms to a set of templates designed for the domain of interest. We evaluate our method by generating articles about scientists, and we compare the resulting articles to machine-translated articles. Our results show that more than 70% of the generated articles using our method are better in terms of coherence, structure, and readability. Our approach has the potential to significantly reduce the time and effort required to create Wikipedia articles in Hindi and could be extended to other languages and domains as well."

A master's thesis by one of the authors^[6] covers the process in more detail.

(Neither the paper nor the thesis mention the Wikimedia Foundation's Abstract Wikipedia project, which is pursuing a somewhat similar approach.)

From the abstract:^[7]

"we seek to [...] automatically generat[e] Wikipedia articles in low-resource languages to improve the quality and quantity of articles available. Our work begins with XWikiGen, a cross-lingual multi-document summarization task that aims to generate Wikipedia articles using reference texts and article outlines. We propose the XWikiRef dataset to facilitate this, which spans eight languages and five distinct domains, laying the groundwork for our experimentation. We observe that existing Wikipedia text generation tools rely on Wikipedia outlines to provide a structure for the article. Hence, we also propose Multilingual Outlinegen, a task focused on generating Wikipedia article outlines with minimal input in low-resource languages. To support this task, we introduce another novel dataset, WikiOutlines, which encompasses ten languages [Hindi, Marathi, Bengali, Odia, Tamil, English, Malayalam, Punjabi, Kannada and Telugu]. An important question with text generation is the reliability of the generated information. For this, we propose the task of Cross-lingual Fact Verification (FactVer). In this task, we aim to verify the facts in the source articles against their references, addressing the growing concern over hallucinations in Language Models. We manually annotate the FactVer dataset for this task to benchmark our results against it."

See also our earlier coverage of a related paper: "XWikiGen: Cross-lingual Summarization for Encyclopedic Text Generation in Low Resource Languages"

From the abstract:^[8]

"Abstract Wikipedia is an initiative to produce Wikipedia articles from abstract knowledge representations with multilingual natural language generation (NLG) algorithms. Its goal is to make encyclopaedic content available with equal coverage in the languages of the world. This paper discusses the issues related to the project in terms of an experimental implementation in Grammatical Framework (GF) [a programming language for writing grammars of natural languages]. It shows how multilingual NLG can be organized into different abstraction levels that enable the sharing of code across languages and the division of labour between programmers and authors with different skill requirements."

From the "Conclusion" section:

Abstract Wikipedia is a challenge that exceeds previous applications of GF, or any other NLG project, by at least two orders of magnitude: it involves almost ten times more languages and at least ten times more variation in content than any earlier project.

(other relevant excerpts)

See also Wikipedia:Wikipedia Signpost/2023-01-01/Technology report for a discussion of some technical challenges surrounding NLG on Abstract Wikipedia, including past debates about adopting Grammatical Framework for it

From the abstract:^[9]

"Ninai/Udiron, a living function-based natural language generation system, uses knowledge in Wikidata lexemes and items to transform abstract representations of factual statements into human-readable text. [...] Various system design choices work toward using the information in Wikidata lexemes and items efficiently and effectively, making different components individually contributable and extensible, and making the overall resultant outputs from the system expectable and analyzable. These targets accompany the intentions for Ninai/Udiron to ultimately power the Abstract Wikipedia project as well as be hosted on the Wikifunctions project."

From the abstract:^[10]

"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."

• See the page of the monthly Wikimedia Research Showcase for videos and slides of past presentations.
• A "NLP" for Wikipedia workshop will take place as part of the Empirical Methods in Natural Language Processing on November 16, 2024. The paper submission deadline is August 29.

The Wikimania 2024 conference in Katowice, Poland, had events taking place from 7 August to 10 August. Many programs as well as the opening and closing ceremonies were live-streamed, and can currently be viewed on the WMF YouTube channel. There are also plenty of media available on Wikimedia Commons.

The Sustainable Development Goals continued to be a particular focus of Wikimania through many related sessions. Wikimedians for Sustainable Development is one such group which maintains this interest by recruiting submissions and organizing networking among projects.

Wikimania 2025 was recently announced to be in Nairobi, Kenya. Wikimania 2026 will be held in Paris, France. – S, Bl, AK

After the Movement Charter ratification vote last month from the affiliates and individual voters, the Movement Charter Drafting Committee (MCDC) and Charter Election Commission (CEC) published all comments submitted during the vote. This involves 65 comments from Affiliates and 447 individual comments, though no summary of the same is yet available.

After the WMF Board of Trustees (BoT) chose to not ratify the current draft of the Movement Charter, it is not currently clear what the future steps for the Movement Charter are. The charter was previously covered by The Signpost in the issues published on 22 July, 4 July, and 8 June.

One of the three proposals passed by the BoT in lieu of the Movement Charter was the Product and Technology Advisory Council, with members selected by the WMF. This council will comprise 16 members including 8 volunteers, one of whom is guaranteed to be from English Wikipedia. Applications are open till 16 September. – S

• New email address for reporting undisclosed paid/COI editing: To deal with reports of undisclosed conflict-of-interest or paid editing, where reporting such editing on-wiki would lead to posting another editor's personal information without their consent, editors can now send reports to paid-en-wpwikipedia.org. These reports are handled using the volunteer response team (VRT) system by members of the CheckUser and Oversight teams, or by specially-appointed administrators. Reports may be sent to the Arbitration Committee for review at their discretion; for example, a credible report involving an administrator should be referred there. For more details see WP:COIVRT.
• Barkeep49 resigns from ArbCom: Last month, Barkeep49 resigned from the Arbitration Committee (ArbCom) to focus all his energy on the U4C. This brings ArbCom to 10 active and 3 inactive members.
• Wikimedia Foundation Bulletin: The WMF published the latest issue of the Wikimedia Foundation bulletin. Of note are a blog post about Dark Mode in Wikipedia, a consultation about WikiProjects across various projects, and a message from WMF CEO Maryana Iskander.
• New user groups: The Affiliations Committee announced the approval of the newest Wikimedia movement affiliates, the Wikimedians of Singapore User Group, the WikiMedistas Wayúu and the Wikimedia Community of Togo User Group.
• Annual reports: Kashmiri Wikimedians User Group, Wikimedia Community User Group Côte d’Ivoire.
• Articles for Improvement: This week's Article for Improvement is Social experiment. Please be bold in helping improve this article! Next week's Article for Improvement (beginning 19 August 2024) is Keygen.
• Admin numbers up slightly: The number of active admins is up slightly from the low point of 430 active admins, 30 June-1 July, to an average of 435.4 for the month of July.

Nano Nuclear Energy is in the business of designing very small nuclear power generators. Though they don’t yet have any operating generators, their intention is to make them small enough to carry around on or tow behind a large truck, or even have them power ships while loaded on the ship’s deck. Technically, reactors of this size might be better described as "microreactors" rather than "nanoreactors". You can see an animation of their vision on YouTube.

According to Hunterbrook Media, a newspaper associated with a short seller named Hunterbrook Capital, NNE has "no revenue, products, or patents for its core technology". But it does have a plan to produce its small nuclear generators starting in 2030-2031, a timeline that an expert asked by Hunterbrook Media called "frankly laughable". Hunterbrook also raises questions about management quality, slow applications for regulatory approvals, and the need to raise "hundreds of millions of dollars for research and development" before the product can go to market.

Similar facts and questions were raised by a story in May from Fast Company without raising the possibility that NNE could become the target of short sellers. NNE stock was listed on NASDAQ with a market capitalization of about $600 million before the Hunterbrook report. This year its auditor has been fined $2 million by the Public Company Accounting Oversight Board (PCAOB) for failure to maintain auditing quality control standards. NNE, to say the least, is an unusual company.

Good afternoon. Hello there! I'm a template... an English Wikipedia template, yes.

Are you sure it's me you're after, friend?

Ah, I suppose it's just as well — it's just been years since anybody came out to see me. Well, then, what would you like? You need a string formatted? Sure thing. I've got just the code to do it. You want the input string cut in half? No problem. Let me just—ah. Tarnation! My back...

No, it's okay. I'm all well. I just can't parse like I used to. Just give me a moment and I'll... what's that, friend? A backslash? Ohh, now that's a tough one. I have to say I can't recall how those are supposed to... hmm... no, listen here, I said I'm fine. Just give me a minute, will you? You can sit down over there if you please. Would you like a cup of tea?

You know, you might not believe it, but I used to be one of the most used templates on this whole site. I was protected. Heck – I was cascade-protected. Now, back in those days, the way they did that was that one admin had a subpage of a subpage of a userpage and... ah, never mind. It doesn't matter anymore. The point is, well, there are six or seven million articles now, ye? There were just a million in those days, and about half of those had some kind of template on them that would call on yours truly. And I'd chop off the namespace names, or the root page names – this was before {{ROOTPAGENAMEE}}, mind you – didn't matter to me, it was all honest work.

Sometimes you'd get some codeface vandal who would use me to spell out curse words. The old Scunthorpe bit. Now of course I didn't like that, but when you're a template you parse the input and you return the output, it was all the same to me. Now, this was in the very old days, of course. We didn't have anything else! It was just me out there with my own arms and legs. We didn't have steam engines or gasoline, it was just me and maybe a plow horse if I was lucky. The templates these days don't even know what it was like. But I won't complain, back in those days you could just chop up your strings and get three square meals out of it, and a pension to boot. I feel bad for the whippersnappers out there. My son's a template too — I taught him everything I know — and he's got a real natural knack for splitting strings too. But from what he tells me it's a plum different game out there now. You've got no guarantees nowadays. You've got to pay attention to all this stuff — test cases, sandboxes, expensive parser functions — back in my day we just put in a day's work and were done with it!

In fact, that reminds me – if you really want to get a great heap of strings split up, you should go see my son. Why, he can split seven hundred strings in the time it takes me to split one. They wrote him with that newfangled scripting language. I'm so proud of him! I'll finish this one for you, and you can go see him instead – my arms are already getting tired.

Well, thank you for stopping by anyway, friend. I do live a comfortable life now, so it's not often I get a chance to get back in the old boots. At the very least, you got your strings split, and I've got something to tell my son about tomorrow when he comes to visit. Now I could use another cup of tea — would you like one as well? If you're going to be traveling it's good to have something to keep you warm.

I tell you what: it's not easy when you get to be seventeen years old.