There are several recommendation system (RS) models that use knowledge graphs (KGs) either as the main data source or as an additional data source for explainability and/or increase in performance.
Since DBpedia is an aggregation KG it can be used directly as additional data for RSs.
Therefore the aim of this project is to develop a benchmark that combines RS datasets and models with additional data from DBpedia into a collection that can be referred to when looking to implement recommendation into search engines, question answering systems or others.
Read some knowledge aware RS surveys:
First and foremost it’s important to do community gathering. Therefore, make a comment in this post introducing yourself, your skills and why would you like to work with this theme.
Clone the repository GitHub - AKSW/natuke, explore and run the experiments to get used to extracting features from KGs and how to use ranking metrics to evaluate machine learning models.
Now that you have an understanding of the possibilities to use KGs for machine learning, more specifically RS, I invite you to bring your own ideas and solutions.
How can the knowledge be extracted from DBpedia to use in a recommendation system pipeline?
How will the final recommendation be presented?
How can the model be evaluated?
Examples of excelent proposals:
- Project size: according to your project scope (from 175 to 300)
- Difficulty: medium
@pauloricardo, @edgardmarx, @tiwarisanju18
Feel free to contact us for more information. We eagerly look forward to working with you and contributing towards making data accessible to all.
Hii @pauloricardo @emarx glad to get a chance to see such a theme, building a recommendation system from such a large DBpedia KG system. I’m Ranjit Patro, a Maths major and CS minor, BS-MS research scholar, from the Indian Institute of Science Education and Research, Berhampur, India. Maybe I am a little late, I am currently at the first step of warm-up tasks. I previously worked in an internship about building Knowledge graphs from Different Datasets (from DBpedia as well). There I also studied and applied the different tasks of Knowledge Graphs which are Knowledge Graph Embedding, Knowledge Graph completion, Knowledge graph alignment, and Continual Entity Alignment for Growing Knowledge Graphs. A knowledge graph-aware Recommendation system, I believe will be an extension of these tasks more specifically inclined toward knowledge Graph completion tasks.
I would love to study and work on building a knowledge Graph aware recommendation system, to extend my knowledge further and discover something new. Love to hear more from others, and how they find this topic interesting. My research interest includes Natural language processing, Knowledge Graph, Cryptology, and Mathematics. I would like to know what should my next step (after completing warmup steps and papers) be to go ahead with this theme of DBpedia in GSOC 2023 and possibly more engagements on this topic from mentors.
Here is my GitHub profile: Ranjit246 (Ranjit Patro) · GitHub (I am interested in working on open-source stuff!!)
Thanks & Regards,
Hello Ranjit, we’d be happy to get your ideas as a proposal in GSoC 2023. I can discuss the theme and give more guidance. Are you able to join DBpedia’s slack?
@pauloricardo Yes, I am in the DBpedia’s slack channel. Please let me know, how can i connect with you there.
Hi mentors, @pauloricardo @emarx,
I’m Alvaro Lopes, a Computer Science undergraduate from the University of São Paulo - Brazil. My research interests are Machine Learning applied to Graphs, Natural Language Processing, and Explainable AI.
I’m currently working as an undergraduate researcher on Machine Learning applied to Graphs (GNNs) and anomaly detection in heterogeneous networks for fraud analysis on invoices. We are currently working on a new Node Embedding technique that uses Masked Language Modeling (BERT). I believe this project would be a great opportunity to apply what I’m learning in my research and to improve my knowledge of such an interesting task of recommendation systems.
I would like to work on this project because I’m interested in contributing to the open-source and ML community. I think there is a lot of potential on DBpedia that could help make ML better by providing quality data and by helping ML models be more explainable and accessible. I believe this project is perfect to explore this potential.
I have already finished the warm-up tasks, and now I’m searching for some RS datasets to get a better understanding of how those datasets can be enriched with additional data from DBpedia. This way, I’ll be able to be more comfortable with RS and DBpedia, so I can make my proposal with your help.
My Github, Linkedin
Hello @pauloricardo and @emarx
My name is Ishaan Watts, and I am a final year student at IIT Delhi majoring in Engineering Physics and minoring in Computer Science with a CGPA of 9.11. I would like to express my interest in being a part of this project.
I have experience working with graph neural techniques having previously worked as a Data Scientist intern at Udaan where I developed an embedding generation framework. My approach involved using an unsupervised graph auto-encoder architecture having heterogeneous nodes and multi-relational edges to depict different relations between them.
I believe this approach could be used in recommendation systems.
Thank you for considering my application. I look forward to hearing from you soon. Here is the link to my CV and LinkedIn
Hello community! I’m Mehrzad, an engineering PhD candidate at ETS university in Montreal, Canada. I fell in love with semantic web stack and formal ontologies during my masters, and to date, I’ve gained a lot of knowledge on OWL ontologies, RDF, JSON-LD, and of course SPARQL. I also have hands-on experience with python libraries such as RDFLib.
During my studies, I’ve been exploring the intersection between knowledge graphs, deep learning and NLP. I investigated RDF-driven KG embeddings in the “smart buildings” domain. Here I shared some notebooks for two papers in which we explored RDF2Vec; I’ve been also experimenting with OWL2Vec. Later, these experiments motivated me to work on Question-Answering from KGs.
Last month, I finished a 1-year paid internship at a software company in Montreal, where we explored question-answering from building graphs. Here you can find a short video or try the demo wep app I developed (demo link provided in video description). During this internship I gained profound knowledge on semantic/neural search using SOTA deep learning techniques, and tons of hands-on experience with TensorFlow and HuggingFace.
My key motivation here is to give back to the community as I’ve been using tons of open-source modules in my works! I found this project particularly appealing as it’s perfectly aligned with my research interests and background.
Dear @pauloricardo and @emarx, please let me know if you have any questions; looking forward to hearing from you!