Automated assistance for writing in a foreign language

Abstract

In this project we present a system with the purpose of assisting users writing in a foreign language. When people write in a language that they are not uent in they usually fall back to thinking in their own language. This results in a lot of grammatical mistakes. People tend to rely on translation systems such as Google Translate, where the users are not shown enough information on the translation made. The users have to blindly accept and trust that the translation is correct. Main aim of this project is to develop a tool which while the users are typing in the foreign language they are given a lot of information about what everything means. The rst objective was to develop an algorithm which generates a list of words based on the two languages. We used language models to generate the list and rank the words based on their probabilities. The results were quite good except that the suggestions were heavily e ected by the topic of the corpus used. The second objective was to create an algorithm which takes in the sentence that the users have written in the foreign language and provide a list of grammatically correct sentences that x any errors the users have made. To create this algorithm we used tag probabilities and also language models to sort the words in the given sentence. When evaluated we achieved a 0.78 score with the METEOR metric. We analyzed all the parts of the method which could have made errors and also proposed suggestions on how this technique could be improved. We also proposed an evaluation technique which could be used to evaluate how well the whole application works when it comes to teaching a user the foreign language.