A few quick web app ideas, to get the ball rolling.. Some previous ones.
1. Your English is so 1890.
There was a popular widget some time ago, in which you paste some text that you wrote (or it processes your blog) and through a statistical analysis of word frequencies it tells you whether you write like Hemingway or more like Virginia Woolf or like Edgar Allan Poe or whoever. Then you get a badge you can put on your blog or on your wall in your favorite social network. It was quite nice and popular, although I can’t find it anymore (EDIT: here it is).
Mine is a tweak on this idea. Google has released a data set with the frequencies of words in books throughout the years (ngrams.googlelabs.com). Say, you might find that “computer” appeared in 5% of the books in 1970 and in 15% of those in 1985 (I am making this up). So, given any year, you can easily find out how likely a certain word is to appear in a randomly picked text. With Bayes rule and some reasonable assumptions we could reverse this and for a given snippet of text, compute the probability that the text was written in 1800 or in 1850 or in 1900, etc. We would pretend that a priori the text is just as likely to come from any year. This is not exactly true, since we know the text comes from the present, or at least we know that the amount of text available from any year grows as we get closer to the present, but let’s ignore all that for a moment. It would be more fun than exact science anyway. However, it could still generate interesting comparisons between the style, choice of words, favorite topics of different people, or the styles of different authors from the past. For example we might discover that some authors had a language that anticipated (or influenced) the future, while others were stuck using the language of the past.
2. World map with radios.
Online radios are a nice tool in learning any language, but often it is hard to find suitable ones. For learning or practicing a language the topic is perhaps less important as long as it is at least mildly interesting and as long as there is sufficient amount of talking (instead of just music), with a clear pronunciation that is representative for a given language/dialect. So what I would like is a world map with a curated list of online radio stations, placed on the map with pins, and for each station it would be given exactly what language/dialect they use, approximately how difficult are the topics discussed, what is the amount of slang, etc.
3. World map with unofficial regions.
Finding a political map with the official borders for countries/states/counties/etc. is easy. There exist however, many other types of regions:
- historical ones (those that in the past might have formed an administrative unit but they don’t anymore)
- linguistic regions (the region where a language/dialect is spoken)
- geographic regions not necessarily corresponding to administrative units (Silicon Valley, Gangetic Plain, etc.)
- regions of scientific interest: (the range of the Neanderthal man, the region of influence of classical Greek culture, the range of the Bengal tiger in 1900)
It would be nice to have a tool for annotating maps in this way (possibly based on Google Maps) just as easily as we can add locations or paths between different points. Such regions, once drawn could be saved, embedded, printed, used in presentations, news articles, lectures, etc. or shared similarly to how users can share photos attached to places on Google Maps. Then, instead of the usual political/geographic view of the world, we could choose other filters/overlays created by others.
Another type of overlay would not have strict borders like regions, but would be a mapping of values onto geographical coordinates. The function value could be shown in colors or using contour curves or any other visualization, and the values could come from uploaded datasets or from some specified function. These types of visualizations exist of course, but I don’t know of a simple interface where users could create such maps.
4. World map live postcard.
A simple service that shows an empty world map and generates a unique url on demand. If someone visits the url, their geographic location appears on the map. This can be done of course with analytics software, but the key here is that a) this is real-time b) accessible to everyone who has the url and c) starting a new map is just one click with no registration and then the url can simply be copy/pasted and sent to others.
Example usage: I am regularly visiting a forum, I am curious where the other visitors are from, so I put the link in the message and watch the map in real time.
Example 2: I am chatting with a group of friends and I hear that a common friend of ours is ill. I create a new map, and send the link to everyone, so when they click it, they appear on the map. They can also watch as more and more people click the link. They can also leave one liner messages that will be displayed on the map next to their location. As our friend checks the link, he/she can see “get well soon” messages popping up from all over the world. This is all a bit similar to Wikipediavision, but more user-generated and with throw-away urls.
5. Centralized phpBB spam filter.
PhpBB is a robust, well-known and loved dicussion board (forum) software that has been around for a long time. The main aspect in which I find it inferior to centralized services like Disqus is spam. Spam filtering is something that is simply easier to do in a centralized way than for each individual board separately. This centralized filter would be a subscription-based API that I can call whenever I get a message on my board and it returns a spam/not spam flag. This can be integrated into my phpBB installation and the necessary action is taken automatically. This way, I wouldn’t have to bother users with captcha’s, registration, senseless restrictions, and I could still have them use the familiar interface of phpBB. If such a service works for WordPress comments and other blogging software, surely it can be retrofitted for phpBB (?).
6. Name face search.
When picking a baby name, many parents would like to find out as much as possible about it, including the history and etymology of the name, which historical persons have had that name, current popularity of the name, etc.
In this service you type in a first name and it gives you the photos of 100 random people having that name. Looking at the age of people on the photos, you can guess whether the name was more popular one or two generations ago than it is today. Also useful if you hear a name in a foreign language and you are not sure whether it is a male or female name. Images of faces can be found on Facebook or on Google Images, but there are still some technical difficulties: a) all 100 should be different persons b) the images should be filtered so that they only contain photos of people and not other random images.