Analysing Twitter comments can be challenging because of its short content and frequent sarcastic expressions. In response to that we have created special engine which analyses the text based on our pre-trained model, giving you over 80% accuracy out of the box.

In order to ensure such a high accuracy level, we have trained our models on billions of Twitter comments that were manually verified for accuracy.

You can test our API by going to live demo page and selecting the "Twitter content" checkbox.

Creating your own Twitter analysis models

You can also train your own models based on your specific training data. To do that, you can go to API models in administration panel. This will use custom categorisation engine, so after your custom model is trained, you will get normal document score plus the categories defined as you have set it up (eg. neutral, negative and positive).

Your training set could look like this:

Negative @lning I don't know. Somebody is spamming with my name and URL I don't know why, I don't know who...
Negative suddenly my jolly mood is saddened by memories of a person i used to really like..
Negative In not looking forward to the Graphic Design Essay which is due before the end of term, your bloody kiddin' me!
Negative @samfjade Its just so not funny thats its funny... off to dry my hair...
Positive Entrepreneurial Wisdom From Sir Richard Bransons Private Island - - Please Share
Positive @AJMIX969 you are going to let one chick do that to you??? that sucks.. im sure there are chicks out there that would never stand you up
Neutral @ChakaDeirvvdre Booty tiem!
Neutral @kana always non negative
Positive has sooo much fun @ bfoxwvoods last night
Positive gotta go work on my stupid english project. ugh. be on tomorrow. byeee
Positive @tatashaskywalker I hope you had a good time tonight!

By using pre-trained models we are able to easily detect irony in the comments

Register free account to try it.

See quick video

social media