eBay (ShopBot) is a Neo4j powered ML chatbot. AirBnb also builds knowledge graphs with Neo4j.
Video from GraphConnect today talking about knowledge graphs: https://youtu.be/dqrlotzdUlo?t=3175
Transparency: I'm an employee at Neo4j.
- Disclosure: I work at Grakn Labs. That being said, I am convinced that it is one of the most innovative solutions out there, and we have a great community working on really neat projects.
It makes an internal knowledge graph as one uses the product (stored in postgres, runs fast). It builds an object model on the fly as a side-effect of using the product, using relationships, numbers, etc as knowledge at an atomic level where words are secondary. The best info organizer (for my style at least) that I know of, though (so far) less feature-rich than many products. I hope the About page at that link explains the present and future well.
VoCol https://github.com/vocol/vocol
ROBOT https://github.com/ontodev/robot
Karma http://usc-isi-i2.github.io/karma/
for data:
Blazegraph https://www.blazegraph.com/
for querying:
SPARQL kernel for Jupyter https://github.com/paulovn/sparql-kernel
1. Operates as a database for both OLTP (traditional query transactions) and OLAP (distributed graph analytics as a language)
2. Has an ontology as a flexible object model (i.e. schema) with types, subtypes, rules and instances
3. Guarantees logical integrity of data with regards to the ontology (i.e. schema constraint, but on a much more expressive data model)
4. Reasoning query language, to retrieve explicitly stored data and implicitly derived information (i.e. infers types, relations, context, and hierarchies of rules, in real time OLTP).
GRAKN.AI has the logical integrity of SQL, which NoSQL and Graph databases lack. It has the data relationships like Graph databases, which SQL and NoSQL do not have. And it scales horizontally like NoSQL, which SQL and Neo4j could not do.
And here's a more detailed differentiator table with granular points: http://links.grakn.ai/362529/10476081
GRAKN.AI is made of 2 core components:
1. Grakn: the storage (i.e. knowledge base) where you store data
2. Graql: the language to retrieve the data Both Grakn and Graql is opensource and will always be opensource, forever. Just like MySQL, Hadoop, Spark, etc.
GRAKN.AI Enterprise is a commercial distribution (which will be released in 3 months), which comes with:
1. Cluster manageent: monitoring and provisioning
2. Security: authenticaion and custom user access right (granular separation of access for users based on different portions of the data model)
3. Natural Language search: both fuzzy string matching and NLP search
4. Knowledge IDE: and IDE for UI-driven knowledge modeling, and IDE to develop the model, and all kinds of modeling and analysis tool to help you manage your knowledge base.
All 4 features above are not available in the opensource distribution. To get them, you need to purchase GRAKN.AI Enterprise. The user can decide to purchase them when they need them.
I hope that helps? Let me know if that makes sense. :)
Grakn sits a layer above this in that is a knowledge graph. You can think of a knowledge graph as a property graph consolidated by an ontology or schema which enables it to encapsulate domain specific information in a structured manner. This in turn enables more advance features such as the automatic resolution of data based on pre defined rules.
You could indeed build a knowledge graph using Blazegraph (or any other property graph) but you would have to go through all the pains of coming up with an integrated and flexible schema as well as a resolution mechanism. Grakn comes with these things out of the box.
So Grakn is not competing with Blazegraph but rather builds on the core principals used by Blazegraph, TitanDB, JanusGraph, and other property graph systems. It builds on it to provide a structured yet flexible graph as well as a built in resolution system.
Shameless plug: we are incorporating both into products and will be offering support/services around both. If anybody is looking for help with this stuff, give us a shout.