Skip to content
This repository has been archived by the owner on Jan 11, 2024. It is now read-only.
/ textent Public archive

Representation Learning of Entities and Documents from Knowledge Base Descriptions

Notifications You must be signed in to change notification settings

studio-ousia/textent

Repository files navigation

TextEnt

TextEnt is a software to obtain vector representations of entities and documents from Wikipedia.

Installation

You can install this software by:

% git clone https://github.com/studio-ousia/textent.git
% cd textent
% pip install http://download.pytorch.org/whl/cu80/torch-0.1.12.post2-cp27-none-linux_x86_64.whl
% pip install -r requirements.txt

Currently, the code is based on Python 2.

Reproducing Experiments

The experimental results presented in the paper can be reproduced by running the commands described below. We conducted experiments using Python 2.7, Ubuntu Linux 16.04, and NVIDIA GTX GPUs.

Resources

Please unarchive these files before proceeding to the next step.

Entity Typing experiment:

% python cli.py evaluate_entity_typing textent_300d entity_db_20160601.pkl

The dataset used in this experiment is contained in the dataset/entity_typing directory.

Text classification expeiment on the 20 newsgroups dataset:

% python cli.py evaluate_text_classification textent_300d entity_db_20160601.pkl --tagme-cache=tagme_cache.pkl -t 20ng

Text classification experiment on the R8 dataset:

% python cli.py evaluate_text_classification textent_300d entity_db_20160601.pkl --tagme-cache=tagme_cache.pkl -t r8

About

Representation Learning of Entities and Documents from Knowledge Base Descriptions

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages