Introduction

Developing a competitive edge in M&A deal-flow is something that very few companies have been able to do effectively. Banks and corporate advisors are constantly on the look out for ways to detect and gain access to the latest corporate M&A in real time.

Using market signals and company reports as prediction forecasters is another painstaking challenge, but having an early indication on merger activity can be highly valuable to anyone whose business relies on this sort of market activity.

The Solution

We set out to measure the co-occurrence of company entities in public news in combination with a module that analyses up-to-date company strategy, resulting in a  score that is able to indicate possible future M&A activity. 

The Perlon solution required considerable data collection, cleaning and reprocessing to allow for the Models and Functions to create actionable data. NLP components including LLMs, embedding, and Named Entity Recognition (NER) models were all used to achieve the final output.  

The core components consist of:

Strategy Module 

Initial indication of suitability based on known strategic direction of selected companies. Uses company profiles through an embedding model and an LLM to create a pairwise M&A score and an M&A distance score. 

Co-occurrence Module 

NER used on company profiles + scraped & preprocessed news data. Entity and contextual vector embedding models + TF-IDF are then used to create a co-occurrence score. 

Result

A highly intuitive system that is able to provide insights and market signal which would otherwise be impossible to detect.