Introduction

This beginner-level tutorial shows how you can build a Knowledge Graph based on input data from a comma-separated value file (.csv), an excel file (.xlsx) or a database table (jdbc).

The complete tutorial is available as a project file. You can import this project

  • by using the web interface (Create → Project → Import project file) or
  • by using the command line interface

    cmemc -c my-cmem project import tutorial-csv.project.zip tutorial-csv
    BASH

This step is optional and makes some of the following steps of the tutorial superfluous.


The documentation consists of the following steps, which are described in detail below:

  1. Registration of the target vocabulary
  2. Uploading of the data (file)/Connect to JDBC endpoint
  3. (Re-)View your data table

  4. Creation of a (target) graph
  5. Creation of the transformation rules
  6. Evaluation of the results of the transformation rules
  7. Execution of the transformation to populate the target graph

Sample Material

The following material is used in this tutorial, you should download the files and have them at hand throughout the tutorial:

  • Sample vocabulary which describes the data in the CSV files: products_vocabulary.nt
  • Sample CSV file: services.csv

    ServiceIDServiceNameProductsProductManagerPrice
    Y704-9764759Product AnalysisO491-3823912, I965-1821441, Z655-3173353, ...Lambert.Faust@company.org748,40 EUR
    I241-8776317Component ConfabulationZ249-1364492, L557-1467804, C721-7900144, ...Corinna.Ludwig@company.org

    1082,00 EUR

    ...



  • Sample Excel file: products.xlsx

    ProductIDProductNameHeightWidthDepthWeigthProductManagerPrice
    I241-8776317Strain Compensator1268158Baldwin.Dirksen@company.org0,50 EUR
    D215-3449390Gauge Crystal77581915Wanja.Hoffmann@company.org2,00 EUR
    ...






Register the vocabulary

The vocabulary contains the classes and properties needed to map the data into the new structure in the Knowledge Graph.

  1. In Corporate Memory, click Vocabularies under EXPLORE in the navigation on the left side of the page.

  2. Click Register new vocabulary on the top right.
  3. Define a Name, a Graph URI and a Description of the vocabulary.
    In this example we will use:
    • Name: Product Vocabulary
    • Graph URI: http://ld.company.org/prod-vocab/
    • Description: Example vocabulary modeled to describe relations between products and services.
    • Vocabulary File: Browse in your filesystem for the product_vocabulary.nt file and select it to be uploaded.
cmemc vocabulary import products_vocabulary.nt
BASH

Upload the data file / Connect to the JDBC endpoint

  1. In Corporate Memory, click Projects under BUILD in the navigation on the left side of the page.
    image1_Build_Projects.png

  2. Click Create at the top of the page.  
  3. In Create new item window, select Project and click Add. 
    The Create new item of type Project window appears.  
  4. Fill in the required details such as Title and Description.  
    Alternatively, import the existing project by clicking Import Project File and selecting the file from your system.  
  5. Click Create. Your project is created.
  6. Within your project, click Create or Create item
  7. In the Create new item dialog, select CSV.
  8. Fill out a label and upload the services.csv sample file.

  9. Click CreateLeave all other parameters at their default values.
  10. Create a second dataset. Choose Excel and upload the products.xlsx file.

Instead of uploading the services.csv sample file into Corporate Memory, you can also load it into a SQL database and access it from Corporate Memory using the JDBC protocol.

  1. In the project, Click Create and select the JDBC endpoint type.
  2. Define a Label for the dataset, specify the JDBC Driver connection URL, the table name and the user and password to connect to the database.
    In this example we will use:

    • Name: Services_ServiceDB
    • JDBC Driver Connection URL: jdbc:mysql://mysql:3306/ServicesDB
    • table: Services
    • username: root
    • password: ****

    The general form of the JDBC connection string is:

    jdbc:<vendor>://<hostname>:<portNumber>/<databaseName> 
    CODE

    Default JDBC connection strings for popular Relational Database Management Systems:

    VendorDefault JDBC Connection StringDefault Port
    Microsoft SQL Serverjdbc:sqlserver:<hostname>:1433/<databaseName>1433
    PostgreSQLjdbc:postgresql:<hostname>:5432/<databaseName>5432
    MySQLjdbc:mysql:<hostname>:3306/<databaseName>3306
    MariaDBjdbc:mariadb:<hostname>:3306/<databaseName>3306
    IBM DB2*jdbc:db2:<hostname>:50000/<databaseName>50000
    Oracle*jdbc:oracle:thin:<hostname>:1521/<databaseName>1521

    *IBM DB2 and Oracle JDBC drivers are not by default part of Corporate Memory, but can be added.

    Instead of selecting a table you can also specify a custom SQL query in the source query field.

(Re-)View your Data Table

To validate that the input data is correct, you can preview the data table in Corporate Memory.

  1. On the dataset page, press the Load preview button
  2. Once the preview is loaded, you can view a couple of rows to check that your data is accessible.
  3. Optionally, you can click start profiling and explore statistics about the dataset.

Create a Knowledge Graph

  1. Click Create at the top of the page.  
  2. In Create new item window, select Knowledge Graph and click Add.
  3. The Create new item of type Knowledge Graph window appears.  
  4. Define a Label for the Knowledge Graph and provide a graph uri. Leave all the other parameters at the default values. 
    In this example we will use:
  5. Click Create.

Create a Transformation

The transformation defines how an input dataset (e.g. CSV) will be transformed into an output dataset (e.g. Knowledge Graph).

  1. Click Create in your project.  
  2. On the Create New Item window, select Transform and click Add to create a new transformation.
  3. Fill out the the details leaving the target vocabularies field at its default value all installed vocabularies, which will enable us to create a transformation to the previously installed products vocabulary.
    In this example we will use:
    • Name: Lift Service Database
    • In the section INPUT TASK in the field Dataset, select the previously created dataset: Services (Input Dataset).
    • Select the previously created dataset as the Output Dataset: Service Knowledge Graph


  4. In the main area you will find the Mapping editor.
  5. Click Mapping in the main area to expand its menu.
  6. Click Edit to create a base mapping.
  7. Define the Target entity type from the vocabulary, the URI pattern and a label for the mapping.
    In this example we will use:

    • Target entity type: Service

    • URI pattern:

      • Click Create custom pattern
      • Insert http://ld.company.org/prod-inst/{ServiceID}
      • where http://ld.company.org/prod-inst/ is a common prefix for the instances in this use case,
      • and {ServiceID} is a placeholder that will resolve to the column of that name
    • An optional Label: Service

  8. Click Save
    Example RDF triple in our Knowledge Graph based on the mapping definition:

    <http://ld.company.org/prod-inst/Y704-9764759> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://ld.company.org/prod-vocab/Service> 
    TEXT
  9. Evaluate your mapping by clicking the Expand button in the Examples of target data property to see at most three generated base URIs.

    We have now created the Service entities in the Knowledge Graph. As a next step, we will add the name of the Service entity.
  10.  Press the circular Blue + button on the lower right and select Add value mapping.
  11. Define the Target property, the Data type, the Value path (column name) and a Label for your value mapping.
    In this example we will use:
    • Target Property: name

    • Data type: StringValueType
    • Value path: ServiceName
      • which corresponds to the column of that name
    • An optional Label: service name
  12. Click Save.

Evaluate a Transformation

Go the Transform evaluation tab of your transformation to view a list of generated entities. By clicking one of the generated entities, more details are provided.

Execute a Transformation to build a Knowledge Graph

  1. Go into the mapping and visit the Transform execution tab.
  2. Press the button and validate the results. In this example, 9x Service triples were created in our Knowledge Graph based on the mapping.
  3. Finally you can use the DataManager Knowledge Graphs module to (re-)view of the created Knowledge Graph: http://ld.company.org/prod-instances/