Some finance companies have contacted Fast Data Science with a need for a very customised named entity recognition solution. Clients prepare lists of investments which could be funds or companies, and request a check on those companies.
The problem is that company names and financial instrument names are not standardised worldwide, and people often refer to companies leaving off the legal suffixes such as Ltd or Holding. Furthermore, the list of suffixes is an open set when we include different languages, and transcription errors and spelling mistakes are common.
So how do you reliably resolve Microsoft to Microsoft Corp and Mueller to Müller AG?
Natural language processing
We have found that a custom Elasticsearch index is able to retrieve a shortlist of companies, but we get better results if we combine Elasticsearch with machine learning libraries in Python. Elasticsearch also allows a customised ranking metric, but we found it was an easier solution to combine Elasticsearch with a machine learning model trained in Python, which could re-rank and identify the most likely company given an input text, using both linguistic features and other known information about the companies such as domicile.
We were also able to output a measure of our confidence of a given match. So our model could be 89% confident that an input should be resolved to Müller AG. This allowed the client to flag items which needed manual review.
If you have a need for a custom financial AI system or financial named entity recognition solution, please let us know. You may also be interested in our drug named entity recognition and country name recogniser.
Dive into the world of Natural Language Processing! Explore cutting-edge NLP roles that match your skills and passions.
Explore NLP Jobs
We are excited to introduce the new Harmony Meta platform, which we have developed over the past year. Harmony Meta connects many of the existing study catalogues and registers.
Guest post by Jay Dugad Artificial intelligence has become one of the most talked-about forces shaping modern healthcare. Machines detecting disease, systems predicting patient deterioration, and algorithms recommending personalised treatments all once sounded like science fiction but now sit inside hospitals, research labs, and GP practices across the world.
If you are developing an application that needs to interpret free-text medical notes, you might be interested in getting the best possible performance by using OpenAI, Gemini, Claude, or another large language model. But to do that, you would need to send sensitive data, such as personal healthcare data, into the third party LLM. Is this allowed?
What we can do for you
