Creating a workspace

About this task

For each model that you want to build and use, you create a single workspace that contains the artifacts and resources needed to build the model. You then train the model to produce a custom model that can be deployed to an external service for use.

Before creating a workspace, answer these questions:

• What type of model do you want to create?

  • Machine learning model: Uses statistical approach to finding entities and relationships in documents. This type of model can adapt as the amount of data grows.
  • Rule-based model: Uses a declarative approach to finding entities in documents. This type of model is more predictable, and is easier to comprehend and maintain. However, it does not learn from new data. It can only find patterns it has been taught to look for.
  • Advanced Rules model: offers deeper customization for text analysis than rule-based models. See Creating an advanced rules model for instructions.

  You can also create one workspace that contains both one rule-based model and one machine learning model.

• What services will use the model?

  See Watson services integration for information about the other Watson services that custom models can be used with.

Procedure

To create a workspace, complete the following steps:

1. Log in as a Knowledge Studio administrator, and click Create Workspace.

   People with the project manager role can perform almost all tasks except creating a workspace. An administrator must create the workspace initially and assign project managers to it.

2. Give the workspace a name. Choose a short name that reflects your domain content or the purpose of the model. If you need to, you can change the workspace name later.

3. Identify the language of the documents in your workspace. The documents that you add to the workspace, and the dictionaries that you create or upload, must be in the language that you specify.

4. Optional: If you want to change the tokenizer that is used by the application from the default machine learning-based tokenizer, then you can expand the Advanced Options section, and choose Dictionary-based tokenizer.

   The default tokenizer is more advanced than the dictionary-based tokenizer; it uses machine learning to identify the tokens in the source documents based on the statistical learning it has done in the language of the source documents. It identifies tokens with more precision because it understands the more natural and nuanced patterns of language. The dictionary-based tokenizer identifies tokens based on language rules. See Tokenizers for more details.

5. Optional: If you want to add project managers to the workspace, then expand the Advanced Options section, and select the names of people you want to add as project managers from the list. The administrator can add or remove project managers later by editing the workspace.

   Only the names of people that you assigned to the project manager role from the User Account Management page for the instance are displayed. See Assembling a team for more information about adding users.

   If you have a Lite plan subscription, skip this step. You cannot add other users, so you cannot assign anyone to the project manager role. You do not need a separate project manager. As an administrator, you can perform all the tasks that a project manager would typically perform.

6. Click Create.

What to do next

After the workspace is created, you can start configuring the workspace resources.

To change the workspace description or workspace name, or to add or remove project managers later, an administrator can edit the workspace. From the Knowledge Studio home page, click the Show menu icon on the workspace tile, and choose the Edit menu option.

Related concepts:

Uploading resources from another workspace

Related reference:

Language support

Tokenizers

A tokenizer groups characters into tokens, and tokens into sentences. A token is loosely equivalent to a word.

The actions that a tokenizer must take to identify a document's tokens differ depending on the language of the document. In English, tokens are often equated to words as delimited by white spaces in a sentence. However, they do not always match one-to-one with words; other textual elements are considered tokens in some situations. For example, punctuation at the end of a sentence is considered a token, and contractions are often expanded into two tokens. In languages that do not use white spaces, such as Chinese, more complicated statistical algorithms are used to identify the tokens.

The tokenization process is important because it determines the groups of characters that users can highlight for annotation in the ground truth editor. Annotations of entity and relation mentions are generally aligned with token boundaries, and must be labeled within a sentence; they cannot span sentence boundaries.

Summary of inputs, outputs, and limitations

Different stages of model development require different inputs and produce different outputs.

For each stage of the model development process, this table summarizes the typical activities that you perform, the supported input file formats, the outputs that can be produced, and any sizing limits or other requirements.