Skip to content

Language evolves over time. We assume that it is possible for a machine learning model to pick up on these small changes, given a large enough dataset.

Notifications You must be signed in to change notification settings

NeXTormer/LTM-LanguageTimeMachine

Repository files navigation

LTM: language time machine

AIR - WS2023 - Group 11

Felix Holz David Wildauer Leopold Magurano
Data (pre-)processing ML model  Visualization and evaluation

Language evolves over time. We assume that it is possible for a machine learning model to pick up on these small changes, given a large enough dataset.

In this project, we aim to employ various document representation techniques, such as analyzing the word frequencies or using doc2vec (an extension of word2vec), to create embeddings of the documents. These embeddings are used to train a machine learning model, which is able to predict the time period (year) in which a text snippet was published. 

The download of the raw unprocessed dataset can be found here

For our training, testing, and evaluation dataset we will be using text snippets from a subset of the publicly available Project Gutenberg eBooks. The text snippets with the corresponding publishing dates are parsed from the eBooks extracted from the .zim file[1] and written to a simple database to give everyone in our team easy access to it. We aim to implement and test different embeddings of our text snippets, which we then use for our machine learning model to predict the time period, or more accurately the year, in which this text snippet was published.

About

Language evolves over time. We assume that it is possible for a machine learning model to pick up on these small changes, given a large enough dataset.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages