There is much talk and writing about machine learning and artificial intelligence. The importance of data is usually emphasised whenever machine learning is discussed, and no wonder: to put it bluntly, any AI or machine learning model is only as good as the data used to train it.
Too often we think of data as just numbers, when any information can be data. Machine learning is already used to process huge amounts of text. An AI system determines whether an incoming email is spam or not, and a search engine asks if you meant something else, and so on.
Ten years ago, machine translation was laughed at, but now highly efficient language models can translate many texts so well that you wouldn’t even know it was done by AI.
Text data and AI go very well together. Machine learning can make text processing systems smarter and help serve the end user better.
Solutions to data management challenges
For example, text search and semantic search capabilities can be added to essential business systems, enabling search to understand the user’s intent better and provide more complete and relevant search results.
Suppose important information, such as suggested actions, gets lost in the shuffle. In that case, a natural language understanding model can be taught to identify recommended actions in reports, and an interface can be developed for finding, browsing, and highlighting them in documents.
A language model can also be taught to identify feedback in emails that needs to be addressed immediately or to direct messages, tickets, or written bug reports to the right teams based on their content.
Language models can summarise long reports, automatically extracting what is relevant.
The generation of reports can also be automated using text generation models. When a problem arises, a language model could be asked to summarise thousands of bug reports and correction reports on a given topic, and AI could tell what to do – or not to do – in the situation.
AI could also analyse these reports and find similarities or cause-and-effect relationships that might not otherwise have been noticed over time.
ATR’s own AI, AIMO, helps us check work-time records. By making sure that the hours worked by our employees are recorded correctly, we can improve resourcing and better estimate the number of hours required for different tasks.
If you are interested in the potential of machine learning in your organisation, you can find out more about it in the AI section of our website. You can also book a sparring session in our CEO Teemu’s calendar. He did his first AI project back at the turn of the 2000s.
The blog’s author, Iidaliisa Pardalin, is studying Digital Linguistics at the University of Turku and is currently doing an internship at ATR as a Data Scientist Trainee. In her spare time, Iidaliisa knits socks and practices survival skills in the post-apocalyptic world of The Long Dark.
