About this project
EmoDex is a project that we (Louis, Oxana, Karsten) have developed in our studies for our research on emotional text classification. In our studies we have developed a powerful and fast tool that needs less than 100k sentences for accurate results (compared to the products of our competition). With this website we offer our model for classification for free.
EmoDex is available on RapidAPI! We offer different plans for easier classification via the RapidAPI service.
Need an API key for classifying even larger batches? Write us about your project and project needs. You should state how much data you need to process, how often you need to request classifications and the overall business case. Please also state whether you use our services for research, commercial or other purposes.
Write us at: firstname.lastname@example.org
Your data is instantly deleted after being rated by our model. We do not log your inputs. The model is not trained on any users data.
⨯ Machine classification
For our emotion classification we use a hybrid approach of machine- and rule-based classification. We use rule-based semantic analysis tools to broaden our feature set. The foundation of our machine learning model is based on word embedding, namely GloVe. We preprocess our texts but keep the syntactic sugar of special characters such as emoticons and emojis. Our model is trained on less than 100k of classified sentences. To build our corpora we used existing corpora as well as crawling for additional data by means of distant supervision. As classifier we use a Random Forest classifier provided by SciKit Learn. Read more about our approach in our research paper.
Your language is not supported? No worries. In our research we have found that most translation tools have a very high quality in preserving a texts emotion keywords. Simply translate your text to english and classify the outcome with our tool. Make sure however to keep the emoticons and emojis untouched, as those carry significant meaning.> Back