GraphGPT

Natural Language → Knowledge Graph

If you want to use knowledge graphs in your project, check out GPT Index.

GraphGPT converts unstructured natural language into a knowledge graph. Pass in the synopsis of your favorite movie, a passage from a confusing Wikipedia page, or transcript from a video to generate a graph visualization of entities and their relationships.

Successive queries can update the existing state of the graph or create an entirely new structure. For example, updating the current state could involve injecting new information through nodes and edges or changing the color of certain nodes.

The current few-shot prompt guides GPT-3 in accurately understanding the JSON formatting GraphGPT requires for proper rendering. You can see the entire prompt in public/prompts/xxx.prompt. A major issue at the moment is latency. Due to the nature of OpenAI API calls, it takes up to 20 seconds to receive a response.

Expected Functions

• Take csv files as input
• Generate a graph from the csv files
• Allow users to query the graph
• Allow users to update the graph
• Allow users to change the color of nodes
• Allow users to change the color of edges
• Allow users to export the graph as a triple class

Possoible Issues

• When the input is empty and users click "Generate", the default or random graph will be displayed
• The Graph zone is not scrollable
• The graph is not responsive to the size of the window.
• Stateless and Stateful components require different formatting due to stop symbol.

How to install GraphGPT

Download or Fork

Run web app

1. Run npm install to download required dependencies (currently just react-graph-vis).
2. Make sure you have an OpenAI API key. You will enter this into the web app when running queries.
3. Run npm run start. GraphGPT should open up in a new browser tab.

Find a example file

Prompts are located in the public/prompts folder.

Test the prompt example

1. text input

2. change the color of the node

3. remove duplicate nodes

Modify the prompt to make GraphGTP works better!

Data

Obtain all csv files from AccreditationExplorer

File Name	Content	Structure
activity.csv	Unit Activities	unitCode: string, to uniquely identify a unit unit: string, name of unit activity: string, to match different activity according to unit
advisable_prior_study.csv	Unit Prior Studies Prerequisites	unitCode: string, uniquely identify a unit that one wants to study apUnitCode: string, uniquely identify a unit advised to have studied before
AQF.csv	Unit Outcomes and AQF ID	unitCode: string, uniquely identify a unit aqfId: string, identify corresponding AQF Id outcomeId: integer, corresponding to unit outcome Id
cbok_end.csv	End Level CBoK (E.g: abstraction)	End: End levels CBoK Sub: Sub levels CBoK, map to End levels
cbok_sub.csv	Sub Level CBoK (E.g: problem solving)	Sub: Sub levels CBoK Top: Top levels CBoK, map to Sub levels
cbok_top.csv	Top Level CBoK (E.g: General, Essential)	Two lines string containing CBoK top level knowledge areas
course.csv	Program Course Information	courseCode: string, uniquely identify course title: string abbreviation: string, abbreviation of the course title CRICOSCode: string degree: string, it is either PG or UG
MIT_AQF_Outcome_cat.csv	AQF categories for MIT	Program: string, abbreviated program name (MIT) ID: integer, used to identify AQF category Category: string, AQF category name Description: string, full description
MIT_AQF_Outcomes .csv	AQF outcomes for MIT	aqfId: string, used to identify aqf outcome Description: string, full outcome description Area_ key : integer, used to link to aqf category
incompatibilities.csv	Units Incompatibilites	unitCode: string, uniquely identify a unit iUnitCode: string, uniquely identify a unit that is incompatible with unitCode.
outcome.csv	Unit Outcomes	unitCode: string, uniquely identify a unit outcome: string, to match different outcome according to unit outcomeId: integer, label the outcome in the unit with a serial number level: string describe: string, to explain the level means
prer_course.csv	Course Prerequisites for Units	unitCode: string, uniquely identify a unit courseCode: string, uniquely identify a course
prer_program.csv	Programming Units Prerequisites	unitCode: string, uniquely identify a unit programmingPoint: integer, points of programming-based units required by specific unit
prer_unit.csv	Unit to Unit Prerequisites	unitCode: string, uniquely identify a unit preUnitCode: string, uniquely identify a unit studied
role.csv	Units Role in Program Course	unitCode: string, uniquely identify a unit courseCode: string, uniquely identify a course role: string, Conversion, Option or Conversion
unit.csv	Units Information	unitCode: string, unit code title: string credit: integer, each unit has 6 points programmingBased: integer, 0 or 1, 0 for non-programming-based, 1 for programming-based availabilities: string
unit_activity_cbok.csv	Unit-Activities-CBoK Mappings	unitCode: uniquely identify a unit unit: string, name of unit activity: string, activity title of a unit knowledge: string, CBoK knowledge area taxonomy: string, Bloom’s taxonomy

Demo

First prompt

unitCode,courseCode\nCITS4009,62510\nCITS4009,62530

Second prompt

unitCode,title,credit,programmingBased,availabilities\nCITS4009,"Computational Data Analysis",6,1,"Semester 2 2021, Crawley (Face to face); Semester 2 2021, Crawley (Online-TT) [Contact hours: n/a];"

Third prompt

courseCode,title,abbreviation,CRICOSCode,degree\n62510,Master of Information Technology,MIT,083866G,PG\n62530,Master of Data Science,MDSc,093310E,PG