What is an annotation? Explanation from its meaning to its relationship with AI and machine learning.

1. What is Data Annotation? When is it used?

1-1. Data annotation

Data annotation is the process of adding tags or metadata to each piece of data, such as text, audio, images, and videos, in the field of IT. The word "annotation" is derived from the English words "note" or "comment".

1-2. Scenes where you see the term "data annotation"

This word is also used in the YouTube settings options. Recommended videos and subscription buttons that appear on the screen at the end of the video are also annotations. Here, information is added to the area on the screen.
When you type "annotation" on Google, the word "java" is also suggested. This is because when engineers write code, they write annotations to share with others what the code is used for.

1-3. Why Data Annotation is Getting Attention

Annotation and tagging of data are necessary for the effective utilization and management of vast amounts of data, known as big data, in recent years. In addition, annotation is essential for creating training data for AI machine learning, which is said to improve business efficiency.
In recent years, there has been a shortage of labor in a wide range of industries due to the declining population caused by aging. Therefore, the use of big data and AI has been attracting attention.

1-4. Means of effectively utilizing vast amounts of big data

Big data refers to a huge amount of data that is difficult for humans to grasp, without a clear definition. These data sets include social data written on SNS and behavioral logs collected from websites. When utilizing these data sets for business, detailed data analysis is necessary. By annotating and tagging the data, it becomes easier to analyze and classify, allowing for efficient utilization in business.

1-5. Create the necessary training data for AI machine learning

AI (Artificial Intelligence) can use data collected from big data to predict future data. In order to do so, it is necessary to use a large amount of training data for machine learning. In annotation, by adding information tags (metadata) that indicate "what kind of data this is", it is possible to create the necessary training data for AI to learn correctly.

2. Positioning of Data Annotation in AI Development

2-1. Data Annotation for Creating Training Data

In the process of AI development, the process of adding information to data is called annotation. The annotated data is called training data and is used for machine learning in AI. Annotation in AI development refers to the process of creating training data.
The positioning of annotation can be shown in a diagram as follows.

 

Let's organize the terminology here for now.

AI: Refers to artificial intelligence itself.
Machine learning: Training to improve the accuracy of AI.
Training data: Data used for machine learning.
Data annotation: The process of creating training data.

2-2. Mechanism of Machine Learning Using Teacher Data

For example, humans teach AI both the question "What is this?" and the answer "This is onigiri" by showing a photo of onigiri. They show the same type of photo multiple times. As a result, the accuracy of the answers "This is onigiri" and "This is not onigiri" increases when AI is shown a photo and asked "What is this?"

In this example, the process of adding information about the question "What is this?" and the answer "This is onigiri" to image data one by one is called data annotation. This process is done manually. Once the information has been added, the data becomes training data for machine learning.
Just like humans, the accuracy of AI increases as it learns. To improve the accuracy of AI, a large amount of training data is necessary.

Please also refer to this article for the meaning and creation of teacher data.
>>What is Teacher Data? Explanation from the relationship with AI, machine learning, and annotation to the creation method.

3. Types of Data Annotation

 

There are different types of data annotation depending on the purpose. Here, we will explain three types: "image", "audio", and "text".

3-1. Image Data Annotation

Image annotation can be broadly classified into three categories.

・Object Detection
Find objects from the image and add meaningful tags such as "onigiri", "human", "car", etc. depending on the target.

・Region Extraction (Semantic Segmentation)
Select regions within an image and annotate them with tags. Identify the meaning of the selected regions, such as "nori" for this region, "clothing" for this region, and "door" for this region.

・Image Classification
Tagging attributes to images. Adds information such as "salmon or cod", "polka dots or stripes", "open or closed".

3-2. Data Annotation for Audio

There are two types of tagging: one for volume and type of sound, and one for the meaning of human speech. In the latter case, the usual procedure is to transcribe the speech into text and tag each word individually. This is mainly used in the fields of speech recognition and intention extraction.

3-3. Text Data Annotation

Through data annotation, specific text can be extracted from a large amount of documents, and necessary text and phrases can be aggregated from scattered data according to desired information. Tagging can be done according to pre-defined rules, and document identification and content analysis can also be performed. It is also used to remove inappropriate content.
In addition, by using pre-defined labels, sentences can be classified semantically. Among text annotation, it is also called text classification annotation, and for example, categorization of articles such as "economy" and "politics" on news sites can also utilize such annotation.

4. Demand for Data Annotation in Conjunction with the Advancement of AI

4-1. Data annotation is essential for the development of AI technology.

In order to have AI perform intelligent tasks, it is necessary to educate AI using annotated data. The background behind the increasing capabilities of AI always involves the process of annotation.

4-2. Increasing Demand for Data Annotation in the Future

As the use of AI continues to expand in various fields such as voice recognition and intent extraction used in smartphones and AI speakers, "Hey Siri" and "OK, Google", as well as automatic driving, suggest functions used in e-commerce, and utilization in the medical and construction industries, the market for data annotation is also growing.

5. Data Annotation Outsourcing Service by Human Science Co., Ltd.

5-1. Rich track record of creating 48 million pieces of teacher data

At Human Science, we are involved in AI development projects in various industries such as natural language processing, medical support, automotive, IT, manufacturing, and construction. Through direct transactions with many companies including GAFAM, we have provided over 48 million high-quality training data. We handle various annotation projects regardless of industry, from small-scale projects to large-scale projects with 150 annotators. If your company is interested in introducing AI but unsure where to start, please consult with us.

5-2. Resource Management without Using Crowdsourcing

At Human Science, we do not use crowdsourcing and instead directly contract with workers to manage projects. We carefully assess each member's practical experience and evaluations from previous projects to form a team that can perform to the best of their abilities.

5-3. Utilizing the Latest Data Annotation Tools

One of the annotation tools introduced by Human Science, AnnoFab, allows customers to check progress and provide feedback on the cloud even during project execution. By not allowing work data to be saved on local machines, we also consider security.

5-4. Equipped with a security room within the company

At Human Science, we have a security room that meets the ISMS standards in our Shinjuku office. We can handle highly confidential projects on-site. We consider ensuring confidentiality to be extremely important for all projects. We continuously provide security education to our staff and pay close attention to handling information and data, even for remote projects.