We design a language model based on a dataset provided by Swiftkey, which is now part of Microsoft (https://www.microsoft.com/en-us/swiftkey?activetab=pivot_1%3aprimaryr2). The dataset provided has sub-data in four languages (Finnish, Russian, German, and English). We use the English data here. After a detailed analysis, we decided to build a model using N-Grams of one, two, three, four, and five words using all of the data provided, and supplemented the language model with 6-grams from the blogs subset of the data. We built a shiny app based on this language model. The test version is located at: https://rubiera.shinyapps.io/capstone_test/

We analyze here a dataset provided by Swiftkey, which is now part of Microsoft (https://www.microsoft.com/en-us/swiftkey?activetab=pivot_1%3aprimaryr2). The dataset provided has sub-data in four languages (Finnish, Russian, German, and English). We analyze the English data here. Here are the steps in this analysis: Task 2: Exploratory Data Analysis of the Swift English Dataset as an N-Gram Model Explore 1-grams, 2-grams, and 3-grams in a 70 percent training sample of the dataset to: - Understand the distribution of words and relationship between the words in the corpora. - Understand frequencies of words and word pairs. - Assess how many unique words are needed in a frequency sorted dictionary to cover (50%,90%) of all word instances in the English dataset. Use wordnet (https://wordnet.princeton.edu/) to: - Evaluate if a word is in the English language, which, by inference, can be used to establish that it is from another language. - Explore ways to increase coverage using synonyms. Task 3: Draft "Next Word" Prediction Model - Explore predictions based on the (N-1) gram to compare use of back-off to the (N-1) gram and/or the use of multiple lower order N-Grams. - Explore techniques to handle unseen N-Grams. - Explore N-grams in the testing dataset to predict the next word knowing the previous 1, 2, or 3 words. - Explore how big an N is needed in our N-Gram model to maximize correct predictions while minimizing response time to user and storage requirements.

The Securities and Exchange Commission requires U.S. stock-issuing, or publicly listed companies to file a large number of reports. These reports contain financial data annotated with text of varying lengths. In this shiny app, we have collected a small sample of recent annotations contained in the financial reports of three companies with different types of operations, and different styles of text annotation.

In this analysis, we evaluate the data used by Velloso et al. in their paper "Qualitative Activity Recognition of Weight Lifting Exercises" [ACM SIGCHI 2013]. The data was collected for six young men performing weight-lifting exercises with a light dumbbell (1.25 KG) using five pre-determined sequences (named as the variable 'classe' in the dataset). Sequence A is the correct sequence, and sequences B, C, D, and E are variations of the men performing the weight-lifting exercise incorrectly. Sequences B, C, D, and E are specific ways to perform the weight-lifting exercise incorrectly, and this means that the four wrong sequences should be just as separable from each other as they are from A.