- add links to base KBs
- [] literature review on AXI
- [] report on the state-of-the-art
- AAAI 2018 paper on Explicit Reasoning over End-to-End Neural Architectures for Visual Question Answering, which used the PSL engine from Hinge-Loss Markov Random Fields and Probabilistic Soft Logic, with code
Inference-based explanations for science questions, etc. See http://cognitiveai.org/explanationbank/ The dataset COLING 2016 is a good one for understanding the details in the paper Paper
| Conference | CFP | Remarks |
|---|---|---|
| AAAI/ACM Conference on AI, Ethics, and Society | http://www.aies-conference.com | Submission deadline: October 31st, 2017 |
| KB Name | Link | Remarks |
|---|---|---|
| YAGO | https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/research/yago-naga/yago/ | Manual verification, high accracy. Yago3 is open source now (https://github.com/yago-naga/yago3). 17m entities and 150m facts. |
| :BaseKB | http://basekb.com/ | Converted from Freebase with 1.3 billion facts about 40 million subjects |
| Freebase | https://developers.google.com/freebase | Project shut down, but data dump available (1.9 billion triples) |
| ConceptNet5 | http://conceptnet.io/ | Multilingual. 8m nodes, 21m edges, with 1.5m nodes in English. Paper on word embeddings using ConceptNet (https://arxiv.org/pdf/1612.03975.pdf) |
| DBpedia | http://wiki.dbpedia.org/downloads-2016-04 | Multilingual. 38.3m subjects in all languages, 4.5m subjects in English. |
| Wikidata | https://www.wikidata.org | Multilingual. 34m items. |
| Webchild | https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/research/yago-naga/webchild/ | |
| Babelnet | http://babelnet.org/ | Multilingual. 14m entries. Wordnet based. |
| Google KG | https://developers.google.com/knowledge-graph/ | Not great coverage/hierarchy/expressivity, but easy to access |
Note on the deprecated Freebase: some (not all) of the Freebase entities are linked to Wikidata entities, and they can identified using SPARQL queries on Wikidata as long as the Freebase identifiers are known, e.g.,
SELECT ?ent WHERE {
?ent wdt:P646 "/m/0bgvcp"
}
| KB Name | Link | Remarks |
|---|---|---|
| The OBO Foundry | http://www.obofoundry.org/ | A collection of domain ontologies, incl. GO, BFO and many others. |
| Oxford ontology library | http://www.cs.ox.ac.uk/isg/ontologies/lib/ | |
| Bio-ontologies | http://bioportal.bioontology.org/ | |
| Protege ontology repo | https://protegewiki.stanford.edu/wiki/Protege_Ontology_Library#OWL_ontologies | Protege is the most popular ontology editor developed in Stanford |
| Dumontier's repo | http://dumontierlab.com/ontologies.php | Prof. Dumontier's collection of Biomedical KBs |
Hinton's recent paper in NIPS 2017 on "Capsules". See also his old paper on capsules
There are several workshops under main AI conferences dedicated for explanations and interpretability in machine learning. The links given are for the latest series only.