Tumblelog by Soup.io
Newer posts are loading.
You are at the newest post.
Click here to check if anything new just came in.

July 01 2011

16:00

The revolution will be translated: Global Voices’ citizen-powered site experiments with English-second

When Ethan Zuckerman and Rebecca MacKinnon created Global Voices in 2004, English was the language of the blogosphere.

“A lot of the people who were using weblogs were writing in English even if it wasn’t their first language,” Zuckerman said. “You would see top Arabic bloggers writing first and foremost in English because they wanted that global audience.”

So Global Voices — a news site about places where English isn’t spoken — was built on English. The idea is that it’s an efficient “bridge language,” reaching a larger potential audience: More translators can do English to Swahili than, say, Tagalog to Swahili.

The site was opened up to translators in 2007 under Project Lingua — a movement that sprang up from the community — but GV still required that all original content be composed in English.

The problem with that workflow? Most of GV’s bloggers don’t speak English as a first language. “People started saying, ‘Look, I work for GV Français, and I write much better in French than I do in English. Why should I have to write in English first and then in French?’” Zuckerman said.

So Global Voices is experimenting with a decentralized, English-second workflow for the first time in its history. Paula Góes, the site’s multilingual editor, is leading the transition.

“The Internet has made the world much, much smaller, but language is still probably the only barrier that really makes it difficult for people to understand each other,” said Góes.

Góes is helping Global Voices build a (virtual) multilingual newsroom, with bloggers and editors assigned to regions and languages. If breaking news happens in South America, a blogger on the ground can choose to write in Spanish first. “It’s obviously much easier for them to write in their own language,” Góes said. “It takes less time for them, too. We’re able to get their stories out there quicker.” The goal is stories that are richer, more nuanced, more genuine. And it opens Global Voices to much wider pool of would-be volunteers.

Paula Góes, Global Voices Multilingual Editor

It all sounds kind of obvious, Zuckerman said — why not let people write in their first language? — but the translations pose a lot of challenges for the organization.

“At the end of the day, everything that ends up on Global Voices in any language is the responsibility of our managing editor, Solana Larsen. And Solana speaks five languages, but she doesn’t speak 30,” Zuckerman told me. “The question became, if we start writing in Chinese first, which Solana does not speak, how can she be responsible for what comes out?”

There was a lot of resistance to decentralization, at least at first. While GV is seen as a pillar of open, citizen-powered media, Zuckerman noted, it’s hardly lawless. “We always have to remind people that we have boatloads of editors. We are a heavily, heavily edited platform,” he said.

“We don’t want to do this in a way that people say, ‘Oh yeah, that Spanish Global Voices, that’s much further to the left than GV is’…. That would be a sign that we’re doing it wrong.” Under the new model, each language site is trusted to enforce its own editorial standards.

So far, the experiment has paid off richly. For example: “Our francophone and Africa coverage had been pretty poor. It was not our strongest section,” Zuckerman said. “It’s gotten better by leaps and bounds since we’ve done this. The francophone West Africans who are part of our community are just much more comfortable writing in French. They write more and they write better.”

To continue the improvement, Góes’ job is to find efficiencies in translation, study metrics, help define best practices, figure out what works. The ultimate goal is to reach more readers in more countries — and English still plays an essential role. All stories are translated to English first, as a rule, but that can take half an hour, a day, two days. “It depends on the urgency or the resources we have,” Góes said.

One helpful thing: Translators for individual language sites can volunteer to take on a story. “We don’t really tell people to translate anything,” Góes said. “It’s completely up to the community. We trust that they will know what posts will be more interesting to their own readers or more important to show in their countries.”

Translators are every bit as much journalists as the writers, because good translation requires an appreciation for context. How do you translate an article about female genital mutilation into Malagasy, for example, when the concept is foreign to an audience in Madagascar? And then there are links, which point to resources outside of the site’s control — resources that will most likely be in a language that’s different than the one spoken by a story’s intended audience. Translators at Global Voices follow each link to try to find relevant alternatives. Google Translate can’t do all that. (Nor does it cover all the languages GV does.)

“If you really want to understand a culture, have a deep understanding about culture, and you don’t speak the language, you cannot really rely on Google Translate,” Góes said. “How would you be able to understand the situation in Syria through Google Translate? I would’t trust Google to let me know about the world.”

Global Voices is like the Red Cross in that the leadership team is paid, but most of the staff volunteers. That means the quality of a language site depends on the time, talent, and interest of unpaid people. (Góes reminded me that her staff is always looking for volunteers. She recently put up an FAQ page for would-be translators.)

I had to ask, what does motivate people to do all this work for free?

“Two things,” Góes said. “One is learning, because when you translate about any other country in the world, you learn so much about it. You have to do research. It’s really, really exciting. I think it’s quite easy to get hooked to.

The other thing? “People think it’s important to bring perspectives into their languages, present them to their friends who can’t speak English in a way that’s not biased.”

Photo of Paula Góes by David Sasaki used under a Creative Commons license

June 25 2010

14:00

The Wikipedia of news translation: Yeeyan.org’s volunteer community

BEIJING — Yeeyan.org has 150,000 registered users, who collectively translate 50 to 100 news articles every day from English to Chinese. Since its inception in 2006, the site has grown into a key gateway for Chinese speakers who want to follow international news. It has been so successful that it has attracted the attention of major news sources like The Guardian and ReadWriteWeb — and also the Chinese government, which abruptly shut Yeeyan down last year for several months.

But this is not a story about China. I believe that Yeeyan is pioneering cost-effective solutions to a major global problem: the ghettoization of information by language. This is a change with potentially far-reaching implications for journalism. I met Kitty Wang, the vice general manager, and Walter Wang, Yeeyan’s community manager (no relation), in a Beijing cafe and asked them to explain to me how Yeeyan works, from technological, social, and business perspectives.

The name Yeeyan derives from the Chinese characters 译 (yi) and 言 (yan), which together mean something like “translate the information,” and Kitty and Walter told me that the site’s primary aim is to increase the flow of information between cultures. Yeeyan.org looks like a news site, with headlining photos and editor-selected hot stories on the front page. (English readers can check out the Google translation.) Stories are arranged into typical sections such as business, sports, technology, and life. The difference is that all of the Chinese-language material on the site has been translated from English sources by members of the Yeeyan community, almost always for free.

The success of the site in producing a continual stream of translations — over 60,000 so far — is the result of careful community management and well-designed social features. And it’s a model that seems like it could be replicated for other languages.

Putting the community to work

Aside from reading stories, users can perform two basic actions: recommend a story or a URL for translation, or translate a recommended story. All visitors to the site are readers, many are recommenders, and only a few thousand — a couple percent — actually create translations. That turns out to be enough, but Yeeyan’s existence depends on getting people to translate.

The site’s design encourages participation in a number of different ways. The front page prominently displays a staff-curated selection of recommended but as-yet-untranslated articles. Users can create “projects,” collections of articles around a specific topic, such as “foreign affairs,” “film lovers,” or “Toyota recall,” and active topics are featured on the front page. Each user has a profile which shows a history of their recommendations and completed translations, and a number of typical social networking features are supported, such as comments on articles and messages between users.

Yeeyan has also recently adopted a badge system, to encourage both participation and quality. There are automatically awarded badges for things like “most translations this week” and “most comments this week,” as well as a series of overall “levels” that users can attain by translating and commenting. Kitty says participation has shot up since the introduction of these incentives.

“Amazing ah?” says Kitty. “Even this little thing can intrigue passion.” As Napoleon once said, a soldier will fight long and hard for a bit of colored ribbon.

But clever software can never replace the involvement of human community managers. Yeeyan’s staff must read each translation before it is posted to ensure that it does not violate government taboos on reporting. (Since reopening in January, Yeeyan has dropped its “current events” category and now avoids all overtly political news, including stories from erstwhile partner The Guardian.) All websites in China are required to self-censor in this manner, but Yeeyan also takes this opportunity to interact with its translators.

“We are the first readers, so we comment first, we encourage users first, we proofread first,” says Kitty. “Those are all important to build up [the] community phenomenon.”

Participation over quality

Kitty told me that there had been much early discussion over whether the site should publish only “good” translations, but in the end they decided that “the gate should be opened to everyone.” Part of their strategy is to encourage readers to become translators. Beginning translators tend to produce rough texts and make many mistakes, says Kitty, but “it is cruel if we don’t even provide a chance.” The policy occasionally drives good translators away from the site, but the Yeeyan team sees translator training as an important part of their social mission.

Nonetheless, Yeeyan has recently debuted a proofreading feature. The original text and a user translation are displayed side by side, and the proofreader can comment on each paragraph. Participation is encouraged by awarding badges for proofreading.

Copyright and the business

Under international law, permission from the copyright holder is generally required to create or publish a translation. By publishing user-supplied translations of arbitrary news material, Yeeyan creates a public good in a legally dubious fashion. But it’s worth remembering that many of the vital information services we now take for granted began on similarly vague principles. The web search engine could not exist without wholesale duplication of the entire web onto local servers, a move which was by no means obviously legal when the first commercial search engines appeared — and which some news organizations still aren’t sure about. The legality of Google scanning books is similarly being challenged.

Even so, Yeeyan is actively seeking agreements with copyright holders to create and publish translations of their work. “We do not want to use content for business illegally, but how to get authorization is a big problem,” Kitty said. “That’s why we are trying to talk to [copyright holders] to have win-win-win business model.”

The three parties in “win-win-win” are the content producer, Yeeyan, and the translator. Yeeyan has just such an agreement with ReadWriteWeb. All RWW articles are translated by a paid freelancer and posted on rwwchina.com, with the ad revenue split between Yeeyan and RWW.

Yeeyan is reluctant to put too much advertising on the main site, both because of the legal questions raised by commercial use of translations and for fear of alienating its all-volunteer community. But there’s money to be made offline if you have access to a huge pool of translation talent, and connections to publishers on both sides of the language divide. Yeeyan hopes to make its money out of brokering translations for foreign firms eager to enter the Chinese market, both online and offline. The company already handles the Chinese language versions of Men’s Health and several other magazines and has brokered more than 20 book deals. Translators are drawn from the best of Yeeyan’s volunteer talent pool. As an incentive to reach professional proficiency, translators who have earned the “Level 4″ badge can apply to be Yeeyan partners. If approved, these skilled translators get the “Partner” badge, plus 3 RMB for every 1,000 views of their translated articles — and possibly a translation job offer later.

Journalism in an era of cheap translation

Yeeyan’s success raises broader questions for journalists and journalism. First, could the model be replicated? Could, say, the Associated Press cultivate a community that actively translated their reporting into other languages? I don’t see why not, though any organization that tried this would need a deep understanding of “community” and everything that implies — and deliver such an obvious public good that thousands of people would be willing to volunteer their time. The business model might also be different, but I can think of a number of ways to monetize a pool of translators and an audience eager for foreign-language news.

But suppose that a news organization was able to deliver a substantial amount of content to foreign-language audiences for very little cost, through communities like Yeeyan, or machine translation, or a combination of the two as in the hybrid World Wide Lexicon project. Such translations would not be up to professional quality initially — if ever — and publishers may be hesitant to endorse error-prone representations of their work. But asking about absolute accuracy and brand dilution misses the point — it’s like critiquing Wikipedia for its (improving) accuracy without discussing the net benefit to humanity. How would cheap translation change foreign reporting, and the very concept of international news? It’s a question which will soon be forced upon the profession by rising technological tides.

For the curious, and because I have repeatedly advocated that reporters make available their full source material, here’s a transcript of my followup IM chats with Kitty. In it we discuss further details of the site’s origin and operations, and their experience with the Chinese censors.

March 15 2010

16:00

The Google/China hacking case: How did the story flow through Chinese-language media?

HONG KONG — A few weeks ago, Jonathan Stray looked at how news is reported and repeated in the new news ecosystem by tracking a single international story — the revelation that last year’s hacking of Google and other companies had been traced to two schools in China. His finding: 121 distinct versions of the story, but only 13 of which included any original reporting.

But Stray’s analysis only looked at English-language media. I wanted to compare his findings with how their Chinese news ecosystem reported the story. So I applied the same research methods to the Chinese-language reporting of this story; I went through every version of the story listed on the China, Hong Kong, and Taiwan editions of Google News to quantitatively examine the coverage on the Chinese-language Internet.

Here’s what I found: Although the total number of versions of the story in Chinese (151) was similar to the number of versions in English (121), the Chinese web pages were almost entirely verbatim reposts of only six pieces of copy, of which four included original reporting.

When Chinese news organizations follow an important and sensitive event, their coverage reflects state media policies. The coverage of the so-called “hacker-training schools” in China offers several clues as to how the Chinese media system, not known for its press freedom, actually works. Independent web sites are not allowed to gather news, and the vast majority didn’t. It’s also much safer to repeat official reports than write original copy when covering politically sensitive topics — and this was certainly a sensitive story for the Chinese government, which has been not-quite-accused by Google of state-sponsored hacking.

There were 151 items on the topic in the Google News story cluster when I gathered them. I went through each, tracking the original source of the copy and the source of the information, among other things, and gathered the results in an spreadsheet. These are the major findings:

There were only six distinct written stories. Four newspapers, a website, and a wire service offered distinct versions of the story of tracing Google’s recent attackers to two schools in China. These media were, in the order that Google News ranked them, Elite Reference (Beijing), China Times (Beijing), China News Service (nationwide), Dazhong Web (Shandong), Qilu Evening News (Shandong), and Information Times (Guangzhou). These six stories were widely reposted by both commercial websites and local newspaper websites. The story from the China News Service was reposted 68 times, while the least repeated story was reposted seven times.

Four of these six stories were based on significant original reporting: China Times, China News Service, Dazhong Web, and Qilu Evening News. The other two stories (from Elite reference and Information Times) rearranged the facts from other media, adding a few comments from news conferences or netizens.

Out of the 151 web pages, 76 (50 percent) were the online outlets of traditional media. Among them, 56 (37 percent) were primarily newspapers, while the others are the websites of TV or radio stations. Private companies or individuals are not permitted to run a newspaper or broadcasting station independent of government oversight in China, so these figures mean that half of the websites following the Google hacking news are effectively state-run media.

Xinhua, the primary state-run Chinese news agency, did not contribute any stories in Chinese, but offered a report in English.

Essentially every web site that was not affiliated with a news agency reposted one of the six stories verbatim. This differed from the practice of English-language web sites, which mostly rewrote the story (without additional reporting). Depending on how you look at it, this is either blatant plagiarism — or extremely efficient.

Again, this is the result of state policy. In China, independent websites are not allowed to conduct interviews or do original reporting. “By maintaining the strictest control over the right to issue, review and revoke press accreditation, the government can exercise control over the media — and potentially over individuals who dare to practice ‘journalism’ outside the system,” according to Qian Gang and David Bandurski of the China Media Project at the University of Hong Kong. Authorized news sites are fed by licensed traditional media. For all other sites, reposting the official stories protects them politically, both from violations of reporting restrictions and from off-message coverage of sensitive topics.

One website, Huameiwang, wrote a summarized story with hyperlinks to relevant articles reposted on its own domain. Due to the journalistic restrictions in force, this form of aggregation is common for online news organizations when covering important news events.

Linking to sources was rare on the Chinese internet. Overall, 118 websites (81 percent of the 145 total reposts) did not link back to the source of their text, and 15 sites did not mention any source at all.

The four pieces offered by the Hong Kong edition of Google News were reposted or rewritten from mainland media and the NYT, as were the stories in the Taiwan edition. For whatever reason, the Chinese-language media in these much less restricted regions did not do original reporting on this story.

Google News missed at least one original story, from the Chinese version of Global Times. There were also different versions of the copy that were not listed, including some less-known local media and bloggers, such as the Xiaoxiang Morning Post in Hunan province.

To summarize: newspapers still played a dominant role in reporting this story, and websites reposted newspaper content repeatedly both for economic and political considerations. Chinese websites rarely did independent reporting, because it isn’t necessary for the online outlets of existing news agencies, and isn’t allowed for all other sites.

The distinct versions of the story are listed in the following table. More information on each of the 151 items (whether or not linked to source, country of publication, primary medium, etc.) is available in the full spreadsheet.

Outlet Sources Dateline Times Reposted Elite Reference NYT, AFP, Xinhua 13 China Times Original, NYT Beijing 29 China News Service Original, NYT Jinan 68 Dazhong Web Original Jinan 7 Qilu Evening News Original 10 Information Times Shanghai Evening Post, Shanghai Morning Post, Qilu Evening News 11

 
Thanks to Jonathan Stray and Yuen-Ying Chan for their contributions.

February 24 2010

15:00

The Google/China hacking case: How many news outlets do the original reporting on a big story?

We often talk about the new news ecosystem — the network of traditional outlets, new startups, nonprofits, and individuals who are creating and filtering the news. But how is the work of reporting divvied up among the members of that ecosystem?

To try to build a datapoint on that question, I chose a single big story and read every single version listed on Google News to see who was doing the work. Out of the 121 distinct versions of last week’s story about tracing Google’s recent attackers to two schools in China, 13 (11 percent) included at least some original reporting. And just seven organizations (six percent) really got the full story independently.

But as usual, things are a little more subtle than that. I chose the Google-China story because it’s complex, international, sensitive, and important. It’s the sort of big story that requires substantial investigative effort, perhaps including inside sources and foreign-language reporting. Call it a stress test for our reporting infrastructure, a real-life worst case.

The New York Times broke the story last Thursday, writing that unnamed sources involved in the investigation of last year’s hacking of a number of American companies had traced the attacks to a prestigious technical university and a vocational college in mainland China. The article included comment from representatives of the schools and, while it had a San Francisco dateline, credited contributions from Shanghai staff. Immediately, the story was everywhere. Just about every major American newspaper and all the wires covered it.

When I started investigating the issue on Monday morning, Google News showed 800 different reports. But how many of these reports actually brought new information to light? By default, Google does not display duplicate copies of syndicated (or stolen) content, bringing the total down to more than 100 unique pieces of copy. I read each one, and several hours later, I had a spreadsheet recording the sourcing for each story. I also recorded the country of publication, the dateline or contributor location if noted, and the primary publishing medium of each outlet (paper, online, radio, etc.) An excerpt of this data is reproduced in the table below.

Here’s what I found:

Out of 121 unique stories, 13 (11 percent) contained some amount of original reporting. I counted a story as containing original reporting if it included at least an original quote. From there, things get fuzzy. Several reports, especially the more technical ones, also brought in information from obscure blogs. In some sense they didn’t publish anything new, but I can’t help feeling that these outlets were doing something worthwhile even so. Meanwhile, many newsrooms diligently called up the Chinese schools to hear exactly the same denial, which may not be adding much value.

Only seven stories (six percent) were primarily based on original reporting. These were produced by The New York Times, The Washington Post, the Wall Street Journal, The Guardian, Tech News World, Bloomberg, Xinhua (China), and the Global Times (China).

Of the 13 stories with original reporting, eight were produced by outlets that primarily publish on paper,  four were produced by wire services, and one was produced by a primarily online outlet. For this story, the news really does come from newspapers.

14 reports (12 percent) were produced by Chinese outlets, had a China dateline, or mentioned the assistance of staff in China. For a story about China, that seems awfully low to me. Perhaps this has to do with cutbacks of foreign correspondents?

Nine reports (7 percent) mentioned no source at all. Five more were partially unsourced. Given the ease of hyperlinks, this frightens me.

Google News tended to rank solid original stories fairly high in its list. Google says they rank stories based on criteria such as the reputation of a source, number of references by other articles, and the headline clickthrough rate — though they won’t reveal exactly how it’s done. The spreadsheet and table below list stories in the order that Google News ranked them.

Google’s story-clustering algorithm included three unrelated stories and missed at least one original report. The three extraneous stories were about Google and China, but not about the recent trace. The exclusion of the Financial Times’ excellent piece is a disappointment — perhaps this has something to do with their paywall? Maybe I’m biased because, as a computer scientist, I appreciate the difficulty of the problem — but I actually think this means that Google News works remarkably well, for a completely unsupervised algorithm that crawls billions of pages to find millions of stories in dozens of languages.

What were those other 100 reporters doing? When I think of how much human effort when into re-writing those hundred other unique stories that contained no original reporting, I cringe. That’s a huge amount of journalistic effort that could have gone into reporting other deserving stories. Why are we doing this? What are the legal, technical, economic and cultural barriers to simply linking to the best version of each story and moving on?

The punchline is that no English-language outlet picked up the original reporting of Chinese-language Qilu Evening News, which was even helpfully translated by Hong Kong blogger Roland Soong. A Chinese reporter visited one of the schools in question and advanced the story by clarifying that serious hackers were unlikely to have been trained in the vocational computer classes offered there. Soong told me that Lanxiang Vocational School is well known in China for their cheesy late-night commercials and low-quality schooling — more of an educational chop shop for cooks and mechanics than the training ground for military hackers than the Times claims.

Tracing one story doesn’t prove anything conclusive beyond that one story, of course. And using Google News as a filter doesn’t truly represent the new news ecosystem: It excludes lots of smaller blogs and other outlets. Soong said Google News told him that his site is not eligible for inclusion in their results because they don’t include small blogs written by a single author. This seems like an arbitrary distinction, but it’s hard to imagine what defensible choice Google could make in an era where the definition of a news source is so up for grabs.

The table below is an extract from the data I collected, with original reporting highlighted. The full spreadsheet also includes country of publication, primary medium for each organization, and lists whether or not each story hyperlinked to its sources.

Article Sources Dateline Calgary Herald Xinhua, NYT (via AFP) ABC AP, Xinhua Shanghai Xinhua original Shanghai MarketWatch NYT, Xinhua San Francisco Reuters Xinhua, NYT Shanghai OneIndia China Daily, NYT (via ANI) Bejing Economic Times ? Washington PC Magazine Blogs NYT Washington Post original, NYT Bejing Times Online NYT Washington Information Week NYT, original FOX News NYT (via AP) The Canadian Press NYT (via AP) Taipei Times (via NYT) San Francisco The Register NYT, Guardian UK, blog The Inquirer AP MarketWatch NYT San Francisco ComputerWorld NYT, blog Telegraph UK NYT PC World NYT, Xinhua Telegraph UK NYT Los Angeles Wall Street Journal original, Xinhua, NYT The Guardian NYT, original Business Week (Bloomberg) Washington AFP NYT New York Reuters NYT New York New York Times original San Francisco, Shanghai Daily Contributor PC World CCTV China Daily, NYT, original Australia Network News Xinhua, NYT After Dawn ?, NYT Top News NYT Daily Latest News ? Press Trust of India China Daily, NYT Bejing UPI NYT New York Security Pro News ? Gizmodo NYT Tom’s Guide NYT Digital Media Wire NYT Mountain View Tech News World original, NYT Global Times original, “agencies” io9 NYT, Guardian ZD Net NYT Benzinga NYT Fox Business NYT CrunchGear NYT AOL News NYT, Guardian, WSJ Tech Blorge NYT KLIV NYT Silicon Valley eWeek NYT TMCnet NYT News.am NYT Chattabox NYT Datamation NYT The New New Internet NYT IT Pro Portal Business Week, Telegraph, PC World The Hill NYT Grab Geek Points NYT DBTechno NYT Boston IT Chuiko NYT All Things Digital NYT Before It’s News NYT V3 ? San Jose Business Journal NYT Help Net Security NYT Channel Web NYT Marketing Pilgrim NYT The Money Times NYT TG Daily NYT, Guardian ABH News NYT, ? Top News NYT, ? PCR NYT Top News NYT Daily Finance NYT, Hacker Journals Shuttervoice ? Thinq NYT Top News NYT New York Magazine NYT Venture Beat NYT Fast Company NYT Gather News NYT Newser NYT NASDAQ NYT (via Dow Jones Newswire) Reuters Xinhua Shanghai PC World NYT, Xinhua Herald Sun NYT, Xnhua (via AFP) Bejing The Hindu ? The Times of India ? Daily Mail NYT PC World NYT, blogs ComputerWorld NYT (via IDG) News.com.au NYT The Globe and Mail NYT, original (via Reuters) 9News NYT Redmond Pie NYT,? Red Orbit NYT New Public NYT Sydney Morning Herald NYT (via AP) Gulf Times NYT MyNews Xinhua, NYT (via Indo Asian News) Zeenews (India) NYT, Xinhua (via PTI) The Tech Herald NYT, Guardian Bejing Web Pro News Financial Tines, NYT Business Insider NYT The Financial Express original, NYT (via Bloomberg) Tech Eye NYT, ? CIO NYT, WSJ (via IDG) Tech Blorge NYT, Xinhua CNET NYT, Xinhua ZD Net NYT, Washington Post China Daily NYT, original Bejing News ? What’s on Xiamen NYT, Xinhua NPR NYT San Francisco Chronicle NYT, Xinhua (via AP) Shanghai The Cap Times NYT, AP, Computer World Little About NYT, Xinhua (via Indo Asian News) Jinan Little About NYT, original (via Asian News Intl) Bejing San Francisco Chronicle NYT (via AP) San Francisco Portfolio.com NYT World Market Media ?
Older posts are this way If this message doesn't go away, click anywhere on the page to continue loading posts.
Could not load more posts
Maybe Soup is currently being updated? I'll try again automatically in a few seconds...
Just a second, loading more posts...
You've reached the end.
Get rid of the ads (sfw)

Don't be the product, buy the product!

Schweinderl