Graph Database
Imagine if your document or experience can be converted into graph database. And this graph database will have all entities and relationships interconnected in graphical visualize pattern. You can also query your document based on the relationships between such entities.
The main reason of getting popularity of such graph database is the ease of data retrieval of unstructured data. It will be very fast and efficient for querying entities in a deeper level where traditional RDBMS are getting slow.
Neo4j is the most popular graph database which we are going to talk about and use. We can use LangChain framework and OpenAI like LLM to retrieve the entities and relationships in between for the given document. Then such data can be imported into graph database for later meaningful and insightful retrieval.
Explanation via Example
Let’s say we have very basic sample text document as below.
text = """ There are two friends named Ankit and Gaurang. Ankit is an employee of Tata Communication. Gaurang is also an employee of Tata Communication. Ankit is good at VOIP and AI. Gaurang is good at VOIP. Ankit is higher in experience then Gaurang. Ankit and Gaurang are living in Ahmedabad. The person having higher experience loves eating tasty food and the other one loves travelling. Ankit is married and Gaurang is unmarried. The married person has one daughter whose name is Prahi. His daughter is studying in Senious KG in school named APSI. """
Above text data we can convert into a meaningful graph database. And to this graph database we can query and retrieve insightful data in graphical format. Once we import data into graph database we can have below insights form it.
Schema of database
So, here we can see the different nodes like Person, Location, Skill, Organization and different relationships like LIVES_IN, GOOD_AT, STUDIES_IN, EMPLOYEE, HAS_CHILD, FRIENDS, MORE_EXPERIENCE etc. This is overall schema in which our data is imported into graph database.
QUERY: Getting persons who are living in locations
Let’s say we need to know about how many persons we have and they are living in some locations. For that we can efficiently query the database and can have below graph data available.
QUERY: Getting person having child and studying in some institute
Let’s say we need to know about a person who has child and child is studying in some institute. This detail relationship we can easily retrieve from our graph database as below. Here you can notice in our original text document we haven’t defined explicitly about Ankit has child, even though it is correctly recognizing the Person Ankit has a child and who is studying in APSI school.
QUERY: Getting persons having some particular skill
Let’s say we need some persons having some particular skill for our project. Then we can retrieve those details also as shown below. Here, we are searching candidates having skill of ‘Voip’
Like these we can get number of relations between unstructured data as per our requirement. Also think about if we have this graph database with live data updated automatically, then we can have very good and valuable insights.
