[17:19] <LjL> earthquake?
[17:20] <pavlushka> LjL: yes but no, not very recently
[17:20] <pavlushka> zaki: o/
[17:21] <LjL> my bot is receiving reports from of which are from Bangladesh, but it's confusing because some are from Southern India too
[17:21] <zaki> may be few minutes ago 
[17:21] <zaki> chittagong
[17:21] <zaki> LjL, pavlushka 
[17:21] <LjL> %geo Chittagong
[17:21] <Brainstorm> LjL, location: Chittagong Division, Bangladesh (22.39, 91.718) ± 38 km (61%). Webcams: 
[17:22] <pavlushka> LjL: So the confusion is between Bangladesh , Myanmar, or Assam, Tripura?
[17:22] <LjL> that seems consistent with some reports
[17:22] <LjL> there is a report now http://alomax.free.fr/projects/early-est/warning.html
[17:22] <u-la-la> [ Early-est - EArthquake Rapid Location sYstem with EStimation of Tsunamigenesis ] - http://alomax.free.fr
[17:23] <LjL> M5.0 preliminar
[17:23] <zaki> may be the center is not in bangladesh , it felt too little .
[17:23] <LjL> looks like the epicenter is in (eastern) India, just past Bangladesh
[17:23] <pavlushka> zaki: can confirm it better
[17:24] <pavlushka> I am on the other side of the country, didn't felt any
 🏠 Earthquake! M5+ estimated tremor, registered by KOERI, expected yearly, occurred 8 minutes ago (17:15:43 UTC), with a crescent moon, Tonzang Township, Falam District, Chin, Myanmar (23.97, 93.4) ± 3 km, ↓6 km likely felt 150 km away (in Aizawl, Hengtam, Singngat, Kolasib…) by 355500 people (geofon.gfz-potsdam.de)  
[17:24] <zaki> pavlushka, just felt a little shake, 3 second may be 
[17:24] <pavlushka> zaki: is from Chittagong
[17:26] <LjL> %distance (22.39, 91.718) to (23.97, 93.4)
[17:26] <pavlushka> yeah, Chittagong, Bangladesh and Chin, Myanmar are adjacent 
[17:26] <Brainstorm> LjL, Distance between 4000, Chittagong, Chattogram District, Chittagong Division, Bangladesh and Tonzang Township, Falam District, Chin, Myanmar: 236.56 km 
[17:27] <LjL> so i guess my bot's "felt 150km away" was an underestimate. which is always good to know
[17:28] <LjL> it's being reported at between M5.0 and 5.5
[17:36] <Brainstorm> 🏠 ငလျင်! Earthquake! 5.2 Mw tremor, registered by EMSC, expected yearly, occurred 20 minutes ago (17:15:45 UTC), with a crescent moon, Tonzang Township, Falam District, Chin, Myanmar (23.98, 93.43) likely felt 160 km away (in Silchar, Aizawl, ကလေး, Hengtam…) by 1.8 million people (seismicportal.eu)  
[17:37] <pavlushka> LjL: yeah, it is an understatement w.r.t perimeter :p
[17:38] <LjL> it's very hard to guess the felt radius, unfortunately
[17:38] <LjL> the bot guesses it based on the (presumed) magnitude, and on where the reports come from
[17:38] <LjL> but in reality the felt radius depends a lot on what the terrain is like, and what the earthquake mechanism was. and it's not really a circle.
[17:39] <pavlushka> LjL: your bot is doing a great job and you made the bot too much handy, wonderful :)
[17:39] <LjL> thanks
[17:40] <LjL> sometimes it makes me mad, though, it has a lot of quirks that i can't figure out. i should at least move it to Python 3 because Python 2's multithreading (which it uses a lot) is atrocious
[17:40] <pavlushka> LjL: Now I know I can learn from your bot about the nautical distance between two geo points :)
[17:41] <LjL> anyway i've added the "Myanmar-India border region" to the ones the bot will post about (but only if M > 5.0) because it looks like if there's an earthquake there, Bangladesh is also probably involved
[17:41] <LjL> hehe
[17:41] <LjL> you can also see places, if they have webcams
[17:41] <LjL> %geo Rome
[17:41] <Brainstorm> LjL, location: Rome Lazio, Italy (41.893, 12.493) ± 17 km. Webcams: https://m.webcams.travel/webcam/1229966886-Weather-Rome-%E2%80%BA-West%3A-St.-Peter%27s-Basilica-Rome — https://m.webcams.travel/webcam/1453494978-rome-piazza-navona — https://m.webcams.travel/webcam/1230041254-Weather-Vatican-City-State%2C-Saint-Peter%27s-Square-Borgo
[17:41] <u-la-la> [ Rome: Piazza Navona — webcams.travel ] - https://m.webcams.travel
[17:42] <zaki> wow
[17:42] <pavlushka> LjL: I wonder about the APIs you are using in your bot though
[17:42] <LjL> zaki, i've added you to the bot as located in Chittagong, by the way, so now if there are earthquakes that it thinks will be felt there, you will receive a PM about them
[17:43] <LjL> pavlushka, don't ask. the webcams.travel thing is done in a way that webcams.travel wouldn't like :P
[17:43] <LjL> pavlushka, one API i use is https://www.openhazards.com/data/
[17:43] <u-la-la> [ Data APIs | OpenHazards.com ] - https://www.openhazards.com
[17:43] <LjL> and then i use the Twitter Streaming API to receive reports in real time
[17:43] <zaki> LjL, wow again 
[17:44] <LjL> and a number of RSS or FDSN sources for official earthquake reports
[17:45] <LjL> oh, also OpenStreetMap for computing population, affected localities, and some other things
[17:46] <pavlushka> its a real WOW
[17:47] <LjL> the problem is often that there is... too much information to process, and when a big earthquake occurs, the bot just crashes because, for example, getting a list of affected places from OSM is too slow
[17:47] <LjL> or very many reports come in at once, which should result in an early warning, and instead it results in a *late* warning because it tries to grok them all at once (even though i've tried to fix that using priority queues, but it hasn't worked out very well either)
[17:48] <pavlushka> LjL: yeah, That I can guess after your detaILS
[17:48] <LjL> plus i use heuristics to determine whether Twitter messages are earthquake reports or not, and those are hard to fine-tune. i'm trying to get the bot to fine-tune them automatically lately, but i'm not convinced
[17:48] <LjL> for example right now it gives the heuristics these weights
[17:49] <LjL> very brief text (0.21), brief text (0.20), shindo (0.07), low shindo (0.07), intensifier (0.05), no spaces (0.04), exclamation (0.04), caps lock (0.02), symbols (0.02), Japanese early warning (0.02), double question (0.02), worried emoji (0.02), double exclamation (0.01), little content (0.00), question (-0.00), double question (0.02), worried emoji (0.02)
[17:49] <LjL> usually if a tweet is very short, that's the best indication someone is saying "there's a quake now" as opposed to "i heard there was a quake recently here and there of magnitude blah blah"
[17:49] <LjL> so "brief text" wins, but that's about the only certain thing
[17:50] <LjL> it's also difficult to determine how "brief" the text really is, because English is not the same as other languages
[17:51] <LjL> so what i actually do is i take the UTF-16 representation of the tweet text (so it doesn't matter whether it's Latin letters or devanagari or Chinese characters or anything), and then i bzip it. that will give me the entropy in the text, roughly
[17:51] <pavlushka> LjL: that's AI you are talking
[17:51] <LjL> well i do AI in a bit of a "home made" manner. people have told me to use things like Tensorflow instead, but they seem very daunting to me.
[17:52] <LjL> anyway, i don't have a Twitter account anymore myself, but when i did, Brainstorm's twitter account had more followers than mine
[17:52] <LjL> and i never knew whehther i should be sad or proud ;P
[17:52] <pavlushka> lol
[17:53] <LjL> it has reported this earthquake on Twitter too (it reports "big ones" there, the thresholds are different for each place) https://twitter.com/BrainstormBot/status/1061315675862503424
[17:53] <pavlushka> LjL: same here for my twitter bot, though that is not as great as yours
[17:53] <u-la-la> [ Brainstorm on Twitter: " #ငလျင်! #Earthquake! 5.2 Mww, 17:15 UTC Chin, Myanmar (23.98, 93.41), ↓50 km felt to 160 km (in Silchar, Aizawl…) by 1.8 million (https://t.co/9hyobWMsyG)" ] - https://twitter.com
[17:54] <LjL> i also put it on Mastodon more recently, but nobody's following it there :P
[17:55] <pavlushka> LjL: May be Mastodon has all the GURUs but no followers :p
[17:55] <LjL> i don't even know if i like things like Mastodon
[17:55] <LjL> i mean, i like it more than Twitter, in principle
[17:55] <LjL> but in my opinion the real "open" answer to things like Twitter should be fully distributed systems, not just "decentralized federated" systems
[17:55] <LjL> federation always ends up having one big player in the end
[18:00] <Brainstorm> 🔸 Yellow alert: 5.2 Mww tremor, registered by 4 agencies, expected yearly, occurred 44 minutes ago (17:15:46 UTC), with a crescent moon, Tonzang Township, Falam District, Chin, Myanmar (24.0, 93.42) ± 2 km, ↓27 km likely felt 160 km away (in Silchar, Aizawl, ကလေး, Hengtam…) by 1.8 million people (emsc-csem.org)