How Roam Kyoto is made

Where the facts come from

Every fact on Roam Kyoto comes from an official or own source. That means the place's, venue's or organiser's own website, the company that runs it, or a public body such as a city office, a park or a museum. Review sites, blogs, aggregators and Google Maps are never a source.

Listing sites can point to an event, but they're only leads. An event goes up once the organiser's own page confirms it, and anything that can't be confirmed there is left out.

A ticket page counts only when it's the primary seller's page for that event, it names the organiser, the site's robots.txt allows reading it and it doesn't put up a bot check. Resale sites never count.

Every site's robots.txt is honoured, and a page behind a login, a captcha or a bot check is never read. Each listing links to the page it came from, so you can check the details yourself before you go.

How events and gigs are found

Events come from the pages of Kyoto's organisers and official bodies: festivals, museums, parks, markets, venues and city offices. The search for them runs in Japanese as well as English.

Gigs come from each venue's own schedule. A venue shows on the site only while its upcoming gigs can be read.

Many sources have a reader of their own, a small program written for that site's calendar, or a feed such as RSS or a sitemap. The readers run twice a week, on Mondays and Thursdays. Sources without a reader are read once a week by AI sessions that work to the same rules and link every row to its page.

The rule for every run is that a missing event is fine and a wrong date isn't. A date that can't be confirmed on the source doesn't go up.

What gets checked twice

When an AI session researches a fact you'd act on, like a price, opening hours and closing days, a date, a closure or an address, a second session checks it again on its own. Descriptions and guide text get one pass and a spot check.

A daily fact check also re-reads the official pages of the places and music venues on the map, a batch at a time, so every entry comes round again within about a month. When a fact has changed, only that fact is edited. A place that has closed for good comes off the map, and its page says it's closed.

What AI writes

AI writes the guides, the place descriptions and the neighbourhood guides, from facts already checked on official pages. The sessions that find events and gigs and read the sources are AI too, working to the rules on this page.

What AI doesn't do

It doesn't invent facts. If an opening time, a price or a date can't be confirmed on an official page, the field stays empty and the page shows nothing for it.

It doesn't copy anyone's writing either. The site stores facts only, never an organiser's blurb, a flyer or an artist's photo.

Photos

Place photos are Creative Commons photos (CC BY, CC BY-SA, CC0 or the public domain mark) that show the actual place, credited to the photographer with the licence. Stock photos are never used.

The licence is checked before a photo goes up, and a photo that doesn't recognisably show the place is taken down. A place without a checked photo has none. No photo has a stranger you could recognise as its main subject.

How the site makes money

Hotel links and hotel maps come from Stay22. If you book through one, Roam Kyoto may earn a commission at no extra cost to you. The booking links are marked as sponsored, and nobody pays to be on the map or in a guide. The privacy and disclosure page has the details.

Daniel's role

Daniel founded the Roam sites. He splits his time between Tokyo and Montreal, and he sets the rules on this page.

There's more about him and the site on the About page.

Spotted a mistake?

Email hello@roamcitymaps.com with the page and what's wrong, like a closed venue, a changed price or a wrong date. The fix is checked against the official page and goes up within the week.