Thesis--University of Florida. / Description based on print version record. Typescript. Includes bibliographical references (leaves 55-58).

Psycholinguistic evidence suggests that people predict upcoming words during language comprehension. While many studies have addressed what information people predict, less is known about the role of factors that potentially mediate predictive processing. This thesis examines predictions of semantic information and word form information. It investigates whether predictive processing is mediated by availability of cognitive resources and time to generate predictions, and compares predictive processing in native (L1) speakers and non-native (L2) speakers. This thesis presents two major lines of work. Two eye-tracking studies investigate prediction of semantic and word form information using a visual world paradigm. In further two ERP studies, we address the interplay of semantic and word form information in a paradigm which combines both possibilities. Experiments 1 and 2 were an eye-tracking study conducted on L1 and L2 speakers of English. The study has demonstrated that L1 and L2 speakers predict semantic information, but their predictive eye movements are delayed when they are under a cognitive load. The effects of cognitive load on predictive eye movements suggest a role of cognitive resources in language prediction in both L1 and L2 speakers. Experiments 3 and 4 were another eye-tracking study conducted on L1 and L2 speakers. The study has shown that L1 speakers predict word form information, but L2 speakers do not. Experiments 5 and 6 were an ERP study, which investigated the interplay of prediction of semantic and word form information in L1 English speakers. Consistent with the two sets of eye-tracking experiments, L1 speakers predicted both semantic and word form information, but word form was only predicted when sentences were presented at a slower rate, while semantic information was predicted at standard and slow presentation rates. Experiments 7 and 8 used the same method as Experiments 5 and 6, conducted on L2 English speakers. L2 speakers comprehended sentences incrementally, but there was no clear evidence that they predicted semantic information or word form information. Experiments 5 – 8 suggest that prediction of word form information is mediated both by nativeness of the target language and by reading rates. To conclude, both L1 and L2 speakers make predictions, but prediction of semantic information occurs only when there are enough cognitive resources available. Prediction of word form can occur in L1 speakers, but it occurs only when there is enough time available. There is no evidence that L2 speakers predict word form, suggesting a role of nativeness of the target language. The findings are consistent with the production-based prediction model of language prediction, in that prediction of word form is less likely to occur compared to prediction of semantic information. Furthermore, the findings are also consistent with the claim that not everyone makes predictions, and predictions do not always occur. The thesis concludes that prediction is additional processing for the comprehension system, and is not always implicated in the comprehension system.

Language modeling is a very broad field and has been used for various purposes for a long period of time to make the lives of people easier. Language modeling is also used for text prediction for mobile keyboards to make the user experience smooth. Tobii has been working since 2001 for users who are suffering from ALS (Amyotrophic Lateral Sclerosis). In this disease, users are unable to talk, walk or chew due to the weakening of voluntary muscles and this gets worse day by day. Tobii has designed an Eye Tracker solution for people suffering from ALS to do their tasks more conveniently. They also developed a keyboard for talking which is controlled by an Eye Tracker device. Users can write sentences using the keyboard and then convey them to other people by conversion of this keyboard written text to speech. Therefore, the thesis is related to predicting the text on the initial input of the keyboard to make the user experience fast, easy and less hectic. This thesis project was conducted at Tobii Dynavox with the objective to build a language model which is an automatic, fast, and efficient approach to predict the text for the given input of text. It explores the way to predict sentences by using deep learning models on the initial text input from users and predict the text by taking into consideration user-specific writing style. The model developed in the thesis could be used by Tobii Dynavox for the end-users to predict the text. Part of the objective is also to find out which is the better approach for the implementation of the language models. The results show that federated learning is performing better than centralized machine learning. After analysing the results, it can also be said that Gated Recurrent Units (GRU) will be a good choice for our models because these models show better results for accuracy and take less training and response times.