Inspiration
We were inspired by how inaccurate Google Translate can be when translating the same phrase through multiple languages.
What it does
The input is a sentence or text. The sentence is translated into a randomly selected language, and the output from that translation is translated into a second randomly selected language. The process continues with a total of five translations. The final text is translated back to English, and the original and the generated sentence after all the translations are compared. A score is assigned based on the similarity of the sentences. Then, another score is assigned to the original input based on the complexity of the sentence. Both scores are weighted and added to generate a cumulative score.
How we built it
We used NLP frameworks and Google Translator API to compare translations. The programming language was Python.
Challenges we ran into
Finding the best algorithm for comparing the meaning of two texts and generating a score. Calculating the complexity of a linguistic phrase.
Accomplishments that we're proud of
The effective implementation of Natural Language Processing.
What we learned
Word embeddings, linguistic analysis, Google Translate API implementation, Tkinter, Sklearn
What's next for Translatr
Improving the algorithm for calculating the cumulative score.
Log in or sign up for Devpost to join the conversation.