What is Watson Knowledge Studio?

IBM Watson Knowledge Studio is a cloud-based application that enables developers and domain experts to collaborate on the creation of custom annotator components that can be used to identify mentions and relations in unstructured text.

Watson Knowledge Studio is:

  • Intuitive: Use a guided experience to teach Watson nuances of natural language without writing a single line of code

  • Collaborative: SMEs work together to infuse domain knowledge in cognitive applications

Why you need Watson Knowledge Studio

Systems like IBM Watson provide natural language processing power to "understand" unstructured data. While AlchemyLanguage understands generic entities and relations "out of the box", some organizations need to mine unstructured data for specific information that is unique to their industry or business needs. As a result, organizations must customize the NLP model in order to realize the full value/benefit of mining the unstructured data.

For instance, imagine you are an auto manufacturer; you're looking to proactively identify safety defects using traffic incident reports. To achieve this, you would need to create a NLP model that, through context, understands manufacturer, make, model, type of incident, and date of incident.

Visualization of custom model extraction

How does it work?

IBM Watson Knowledge Studio uses a guided experience that enables subject matter experts to teach Watson the nuances of natural language. Rather than writing code, experts annotate example documents from which Watson can learn.

With Watson Knowledge Studio, you can train custom models to:

  • Identify custom entity types that aren't present in generic models

  • Identify unique user-defined relations between entities

  • Learn natural language cues that are unique to your domain in order to identify entities and relations

In addition, Watson Knowledge Studio enables experts to collborate to infuse domain knowledge in cognitive applications and provides mechanisms to resolve differences in interpretations, which ultimately improves the quality of the model.

Finally, Watson Knowledge Studio provides a full solution for managing the entire model lifecycle -- and, it's integrated with AlchemyLanguage to provide straight-forward deployment.

Example request

The following AlchemyLanguage request uses a custom model trained on traffic incident reports to analyze the following sentence: "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer."

curl "https://gateway-a.watsonplatform.net/calls/text/TextGetTypedRelations?apikey=API_KEY&model=en-us-tir&outputMode=json&text=The%20vehicle%20rotated%20out%20from%20the%20initial%20wall%20impact%20and%20was%20subsequently%20struck%20by%20a%202013%20BYD%20Qin%20pulling%20a%20single%20trailer."

Output:

{
  "status": "OK",
  "usage": "By accessing AlchemyAPI or using information generated by AlchemyAPI, you are agreeing to be bound by the AlchemyAPI Terms of Use: http://www.alchemyapi.com/company/terms.html",
  "totalTransactions": "1",
  "language": "english",
  "typedRelations": [
    {
      "arguments": [
        {
          "argnum": "1",
          "entities": [
            {
              "id": "-E5",
              "text": "rotated out",
              "type": "IMPACT"
            }
          ],
          "text": "rotated out"
        },
        {
          "argnum": "2",
          "entities": [
            {
              "id": "-E2",
              "text": "vehicle",
              "type": "Vehicle"
            }
          ],
          "text": "vehicle"
        }
      ],
      "score": "0.889433",
      "sentence": "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer.",
      "type": "impactPoint"
    },
    {
      "arguments": [
        {
          "argnum": "1",
          "entities": [
            {
              "id": "-E3",
              "text": "wall impact",
              "type": "IMPACT"
            }
          ],
          "text": "wall impact"
        },
        {
          "argnum": "2",
          "entities": [
            {
              "id": "-E2",
              "text": "vehicle",
              "type": "Vehicle"
            }
          ],
          "text": "vehicle"
        }
      ],
      "score": "0.562738",
      "sentence": "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer.",
      "type": "impactPoint"
    },
    {
      "arguments": [
        {
          "argnum": "1",
          "entities": [
            {
              "id": "-E6",
              "text": "struck",
              "type": "IMPACT"
            }
          ],
          "text": "struck"
        },
        {
          "argnum": "2",
          "entities": [
            {
              "id": "-E1",
              "text": "Qin",
              "type": "MODEL"
            }
          ],
          "text": "Qin"
        }
      ],
      "score": "0.808518",
      "sentence": "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer.",
      "type": "impactPoint"
    },
    {
      "arguments": [
        {
          "argnum": "1",
          "entities": [
            {
              "id": "-E1",
              "text": "Qin",
              "type": "MODEL"
            }
          ],
          "text": "Qin"
        },
        {
          "argnum": "2",
          "entities": [
            {
              "id": "-E7",
              "text": "2013",
              "type": "MODEL_YEAR"
            }
          ],
          "text": "2013"
        }
      ],
      "score": "0.953107",
      "sentence": "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer.",
      "type": "hasProperty"
    },
    {
      "arguments": [
        {
          "argnum": "1",
          "entities": [
            {
              "id": "-E1",
              "text": "Qin",
              "type": "MODEL"
            }
          ],
          "text": "Qin"
        },
        {
          "argnum": "2",
          "entities": [
            {
              "id": "-E0",
              "text": "BYD",
              "type": "MANUFACTURER"
            }
          ],
          "text": "BYD"
        }
      ],
      "score": "0.994090",
      "sentence": "The vehicle rotated out from the initial wall impact and was subsequently struck by a 2013 BYD Qin pulling a single trailer.",
      "type": "hasProperty"
    }
  ]
}

Pricing

To create and use a custom model, you will need to purchase access to Watson Knowledge Studio (purchased separately for the duration of your training). Additionally, after you publish your custom model, you'll need to pay for usage of the AlchemyAPI service, which is available through Bluemix. The API pricing is available in the Advanced plan and consists of a model instance fee as well a metered fee based on API usage.

How to create a custom model