This repository contains different algorithms that are used to build taxonomy from different text corpus. It also contains novel approach to enrich taxonomy when new set of data is available from same domain of knowledge.

• Python 3.0 or higher
• Clone this repository.
• All the additional dependencies will be installed using the build script.

• Install all the dependent packages and modules using the setup script located in the root folder:

    $ python3 setup.py

• For full run taxonomy algorithms, trigger the run python script from root folder as below

    $ python3 run.py [arguments]

• Command-line arguments for the program:

• For taxonomy enhancement run, trigger the run_enhance python script from root folder as below

    $ python3 run_enhance.py [arguments]

• Command-line arguments for the program:

• Run enriching taxonomy algorithm on a dataset (amazon_fashion) using 10% of total data volume

    $ python3 run_enhance.py -d 'amazon_fashion' -n 10

• Check the file the following path

    ./src/data/output/hypertree/Visualisation/Fashion_vis.html

    Amazon Fashion Review
    BBC news
    DBLP network

    taxogen      : Unsupervised way to build taxonomy
    taxogen_nole : Taxogen with no local embedding
    taxogen_noac : Taxogen with no adaptive clustering
    hclus        : Hierarchical Clustering
    ncrp         : Taxonomy building based on chinese resturant process

    The enhance taxonomy algorithm requires the trained model and existing taxonomy. Both are saved by taxonomy generating algorithm (e.g. Taxogen, Taxogen NoLE etc.).
    Therefore, it is imporatant to run the taxonomy generating algorithm first for respective data and then run the enhance taxonomy algorithm for the same data to get
    the relevant result.

@inproceedings{rehurek_lrec, title = {{Software Framework for Topic Modelling with Large Corpora}}, author = {Radim {\v R}eh{\r u}{\v r}ek and Petr Sojka}, booktitle = {{Proceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks}}, pages = {45--50}, year = 2010, month = May, day = 22, publisher = {ELRA}, address = {Valletta, Malta}, language={English} }
@article{DBLP:journals/corr/abs-1812-09551, author = {Chao Zhang and Fangbo Tao and Xiusi Chen and Jiaming Shen and Meng Jiang and Brian M. Sadler and Michelle Vanni and Jiawei Han}, title = {TaxoGen: Unsupervised Topic Taxonomy Construction by Adaptive Term Embedding and Clustering}, journal = {CoRR}, volume = {abs/1812.09551}, year = {2018}, url = {http://arxiv.org/abs/1812.09551}, archivePrefix = {arXiv}, eprint = {1812.09551}, timestamp = {Fri, 15 Feb 2019 12:58:25 +0100}, biburl = {https://dblp.org/rec/journals/corr/abs-1812-09551.bib}, bibsource = {dblp computer science bibliography, https://dblp.org} }
@inproceedings{DBLP:conf/kdd/MaCW018, author={Jianxin Ma and Peng Cui and Xiao Wang and Wenwu Zhu}, title={Hierarchical Taxonomy Aware Network Embedding}, year={2018}, cdate={1514764800000}, pages={1920-1929}, url={https://doi.org/10.1145/3219819.3220062}, booktitle={KDD}, crossref={conf/kdd/2018} }

Amazon Fashion Review Dataset: https://snap.stanford.edu/data/web-Amazon.html
BBC News Article Dataset: http://mlg.ucd.ie/files/datasets/bbc-fulltext.zip
DBLP Citation Network Dataset: https://dblp.uni-trier.de/xml/