![](https://legal.fronteo.com/wp-content/uploads/2020/09/dmitry-ratushny-wpi3sDUrSEk-unsplash-scaled.jpg)
Glossary
2020/ 3/ 5![](https://legal.fronteo.com/wp-content/uploads/2020/09/pixta_27629574_L-1024x614-1.jpg)
Turn text data into knowledge
2020/ 4/ 1![](https://legal.fronteo.com/wp-content/uploads/2020/09/pixta_33391009_XL-1024x683-1.jpg)
It's been about five years since the term "third artificial intelligence (AI) boom" came into the media. There are many services that use AI in the world, and it is in the news every day, but some people still do not know how artificial intelligence is used and how they should use it. Let's do it.In addition, there are many stories that the company said, "Do something with artificial intelligence," but got stuck.
FRONTEO started with international litigation support and fraud investigation, and developed its own artificial intelligence engine "KIBIT" that performs natural language processing in 2012, and has been using it in-house for nearly 7 years and outside for more than 4 years. We provide AI solutions for.Here, I would like you to know the process of analyzing familiar "text data" from the starting point of utilizing artificial intelligence, and to get a chance to move from knowledge to practice.
To utilize artificial intelligence, first "data"
One of the strengths of artificial intelligence is "speed of processing."It is said that the third artificial intelligence boom occurred due to the development of "big data" and "computing power", but the present age is exactly the "century of data".Each person has a smartphone with the processing power of a large computer in the past, and the data is collected in a large-scale system on the cloud, enabling analysis that was unthinkable a long time ago. I am. (However, it is not good if there is a lot of data. I will explain later.) Without such data, the meaning of moving artificial intelligence and the strength of "speed of processing" will not be exhibited.In addition, there is no need to use artificial intelligence as long as the content and amount are small enough for humans to visually judge. How to handle "a large amount of data that people cannot easily see" is the beginning of the use of artificial intelligence.
Figure 1. Data handled by artificial intelligence
There are various types of data handled by artificial intelligence (Fig. 1).Let's focus on text data here.Even if you don't have "big data collected from your smartphone on the cloud", the familiar things you use in your daily work will also be the data you give to artificial intelligence. FRONTEO achieved the No. 2016 share in the domestic language analysis market using artificial intelligence from FY2018 to FY1※ 1However, the data handled in this "language analysis" is mainly text data, and most of them are familiar.There are various text data <Fig. 2> such as emails exchanged daily at the company, documents written daily, telephone conversations converted into text, interview records, writing to Internet services, etc., and FRONTEO The artificial intelligence "KIBIT" originally developed by KIBIT is analyzing.In recent years, "digital transformation (DX)" has been widely advocated, but it digitizes familiar actions and efforts that we are in contact with every day, and uses the power of ICT, IoT, and AI to live and live. It is a movement to change the structure and structure of industry and society.
Figure 2. Various text data
What is the purpose and "result" of using artificial intelligence?
The next important thing is the purpose of utilizing artificial intelligence and what kind of outcomes you want.Even if you collect data, artificial intelligence cannot be used unless you have decided what and how to use it. Most of the cases where "the purpose is to use artificial intelligence" or "I was able to show a tendency by PoC (proof of concept) but I do not know what to do next" are the "results" I aim for. The setting of is often ambiguous. FRONTEO's artificial intelligence KIBIT's daily language analysis is currently being used by more than 3 companies in total with the aim of achieving the results shown in <Fig. 200>.
Figure 3. A wide range of “results”
Once the “data” and the desired “results” have been determined, the next step is to select the optimal machine learning method (Fig. 4).The third artificial intelligence boom is said to have been brought about by “deep learning.” The development of "big data" and "computing power" matched the characteristics of deep learning, bringing unprecedented breakthroughs.On the other hand, it is said that deep learning is not all-purpose, and that it has strengths and weaknesses depending on the type of data and the desired results.When choosing a machine learning method, we also consider the time required for introduction and operation, running cost, learning process, tuning for accuracy, and various other factors in total, and design with an eye on operation and implementation. must be done. The artificial intelligence engine "KIBIT" developed in-house by FRONTEO uses a unique algorithm.It specializes in text data, can be analyzed with a small amount of training data, can be introduced in a short time, does not require large-scale computing resources, and is characterized by light processing.
Figure 4. Selection of optimal machine learning is the key to utilization
What are the specific outcomes of introducing artificial intelligence with these preparations and choices?I think you can think of it. FRONTEO has been providing companies with AI solutions using artificial intelligence KIBIT since 2014, and the results shown in <Fig. 5> have been obtained according to the purpose and desired results.
The elements explained so far and the areas of artificial intelligence utilization provided by FRONTEO are shown in <Fig. 6>.
Figure 5. Results of AI solutions using KIBIT
Figure 6. Artificial intelligence utilization area provided by FRONTEO
"Natural language processing" is a technology that uses a computer to process the language that humans normally use. FRONTEO is developing a technology to analyze freely written natural language based on the structure of sentences.We have natural language processing-related technology that can handle a variety of valuable sentences that exist in society.
KIBIT is a Japanese artificial intelligence engine independently developed by FRONTEO.Combining the subtleties of the human mind (KIBI) and the smallest unit of information, the "bit" (BIT), it means "artificial intelligence that can learn the subtleties of humans".By combining integrated and systematized behavioral informatics through artificial intelligence-related technologies that are responsible for learning and evaluation and data analysis sites, it is possible to reproduce judgments and information selection on behalf of humans.