Models for word prediction and information retrieval

Abstract

In the past, many techniques have been explored to predict words in a sentence or extract information from textual documents based on a query. The purpose of this thesis is to compare and contrast different word-prediction models in the Natural Language Processing area of N-grams which can ultimately be used to facilitate the process of writing texts which follow a certain structure or pattern. Moreover, an Information Retrieval System will be created using the word2vec framework which can be used to obtain a pre-specified number of documents based on a given query. The application section involves applying these models on data acquired from the recruitment industry which can be split as job descriptions and CVs.