Using letter frequencies scraped from Wikipedia, this tool conducts a Chi Squared test against inputted text and all the available language frequencies, and returns the name of the language with the highest P-Score.
I was recently looking through some old University assignments and noticed one on cracking a ceaser cypher using letter frequency analysis, I had forgotten about this project and thought it was pretty cool. Using this as inspiration, I thought it would be interesting to see if I could make a more complex project using this statistical technique, so here's the random idea of estimating languages using letter frequency analysis.
I have no idea how beneficial this is. I assume whoever made the "Detect Language" feature on Google Translate has figured out a much more accurate way to detect the language you are typing.
With the above said, this is pretty fast, and whilst it occassionally gives a wonky result, I've found it to be pretty accurate even on short sentences.
Most results, regardless of the text size will be available in under ~1ms. Generally the first result takes around ~12ms to compute, from some brief testing I believe this is because the Accord.Statistics package I use for the Chi-Squared tests is caching values based on the degrees of freedom provided in the first run, to be used in subsequent runs - although I haven't 100% verified this.