CONVEX is an unsupervised method that can answer incomplete questions over knowledge graphs (Wikidata in our case) by maintaining conversation context using entities and predicates seen so far and automatically inferring missing or ambiguous pieces for follow-up questions. The core of our method is a graph exploration algorithm that judiciously expands a frontier to find candidate answers for the current question. For details, please refer to the paper.
The website of our work (including a demo) is available here: https://convex.mpi-inf.mpg.de/ A preprint of our corresponding CIKM'19 paper can be found here: https://arxiv.org/abs/1910.03262
A demo of CONVEX is available on our website https://convex.mpi-inf.mpg.de. Have a look at the sample conversations to get a feel for the input, and try it out!
pip install spacy requests hdt networkx
python -m spacy download en_vectors_web_lg
(If there are any issues installing hdt, please check out issue #5).
(with new train-dev-test split)
Adjust the settings file of the project.
Download the data folder and initialize the project.
bash initialize.sh
Run CONVEX on the ConvQuestions benchmark. The results will be printed into a results.txt file.
nohup python convex.py &
Adjust the settings file of the project.
Download the data folder and initialize the project.
bash initialize.sh
Run CONVEX on the given benchmark. The results will be printed into a results.txt file.
nohup python convex.py &
Please cite our CIKM 2019 paper if you use CONVEX in your work:
@inproceedings{christmann2019look,
author = {Christmann, Philipp and Saha Roy, Rishiraj and Abujabal, Abdalghani and Singh, Jyotsna and Weikum, Gerhard},
title = {Look Before You Hop\&\#58; Conversational Question Answering over Knowledge Graphs Using Judicious Context Expansion},
booktitle = {Proceedings of the 28th ACM International Conference on Information and Knowledge Management},
series = {CIKM '19},
year = {2019},
isbn = {978-1-4503-6976-3},
location = {Beijing, China},
pages = {729--738},
numpages = {10},
url = {http://doi.acm.org/10.1145/3357384.3358016},
doi = {10.1145/3357384.3358016},
acmid = {3358016},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {conversations, knowledge graphs, question answering},
}
The CONVEX project by Philipp Christmann, Rishiraj Saha Roy and Gerhard Weikum is licensed under MIT license.