Yes, Datasaur's LLM (Large Language Models) suite is currently free for all users. It's part of our effort to provide an effective tool for fine-tuning the intelligence of large language models. However, if you need access to the LLM Evaluation Project or LLM Ranking Project features and they are not yet available on your account, you can send a request to support@datasaur.ai.

Datasaur supports integration with multiple cloud storage providers such as Amazon S3, Google Cloud Storage, and Azure Blob Storage. Here are the detailed guide links for the setup:

1. Amazon S3: https://datasaurai.gitbook.io/datasaur/integrations/external-object-storage/aws-s3

2. Google Cloud Storage: https://datasaurai.gitbook.io/datasaur/integrations/external-object-storage/google-cloud-storage

3. Azure Blob Storage: https://datasaurai.gitbook.io/datasaur/integrations/external-object-storage/azure-blob-storage

By integrating these services, you can create projects using your preferred cloud storage provider. This also allows you to maintain control over access from Datasaur to your object storage.

Datasaur complies with stringent security and privacy standards, including SOC 2 Type II and HIPAA.

SOC 2 Type II is a component of the American Institute of CPAs (AICPA)'s Service Organization Control reporting platform. Its goal is to assure the security, availability, processing integrity, confidentiality, and privacy of customer data. The certification demonstrates that an organization has strict internal controls regarding the handling and protection of their clients' data.

HIPAA, or the Health Insurance Portability and Accountability Act, is a US legislation that provides data privacy and security provisions for safeguarding medical information. Being HIPAA compliant means that any personally identifiable information in medical records, conversations between health care professionals, billing information, or any health care paperwork are suitably protected.

In Datasaur, you can use the Mask PII (Personally Identifiable Information) setting to anonymize selected information attributes in your project. Here are the steps:
1. Turn on the Mask PII setting, this can be found under project settings.
2. After enabling the Mask PII setting, click on Data Masking Settings to define the masking method and select the specific attributes to mask.
3. In the PII fields table, you can check the boxes next to the information attributes you want to mask.
4. You can also set a regular expression rule to define what entities to mask.
5. There are two ways to mask the information attributes:

Random Character: This replaces the personal information with random characters.

Asterisk: This replaces the personal information with asterisks.

Remember, this masked data will be visible to all team members in the project.

The documentation is available at:https://datasaurai.gitbook.io/datasaur/workspace-management/workspace#mask-personally-identifiable-information-pii.

Yes, Datasaur has a feature called Datasaur Dinamic which allows data labeling to train a model natively. It creates a ready-to-use model directly from Datasaur. You can import your labeled or unlabeled dataset into Datasaur, review it, and then use one of the supported providers for training. Currently, Datasaur supports integration with Hugging Face and Amazon Sagemaker.

Here are the steps to create a model:

1. Import your dataset into a Datasaur project and start labeling, or upload a pre-labeled dataset.

2. Review and correct any possible mistakes in the labeled data.

3. Go to the 'Assisted Labeling' dropdown menu and select 'Predictive Labeling' then 'Train'.

4. Choose the supported service provider that you prefer (Hugging Face or Amazon Sagemaker).

5. Follow the on-screen instructions related to the chosen provider.

Once the model training is completed, predictions will be displayed and you can accept or reject the labels based on your needs. With Datasaur Dinamic, you can optimize prediction performance and reduce the consumption of annotation time.

For detailed instructions, please refer to the documentation:
https://datasaurai.gitbook.io/datasaur/advanced/extensions/datasaur-dinamic.

Yes, that's correct. Datasaur supports using your own model via a custom API for machine learning assisted labeling. You will need to provide your API name, API URL, and any necessary credentials. Once it's set up, this model can be used to automatically generate labels for your datasets.

To access the LLM Lab feature, you need to reach out to Datasaur's support team via support@datasaur.ai. This feature is not available to all users by default, and needs to be enabled for your usage. Once your request is received and approved, Datasaur will enable it for your account.

Span labeling in Datasaur refers to the process of applying labels to specific portions or spans of text in a document. This could be individual tokens (like words or punctuation) or even multi-token entities. This type of labeling is particularly useful for tasks such as Named Entity Recognition (NER) and Part of Speech tagging (POS).

In deep, when users work on a span-based project, they highlight the tokens in a document and assign them labels based on their characteristics or the role they play in the context. The labels are predefined and are part of a label set which is decided before beginning the project.