EQT is proud to announce that an academic paper authored by the data science team at EQT Motherbrain has been accepted for publication by the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP), acknowledged as one of the best Natural Language Processing (NLP) academic conferences in the world of AI. The paper originated from the work of developing one of the many advanced deep learning models used by the Motherbrain platform.

Motherbrain is EQT’s proprietary investment platform driven by diversified big data and cutting-edge algorithms. EQT uses Motherbrain across the EQT platform to source deals and to help investment teams make better informed investment decisions. One of many analytical scenarios that Motherbrain helps tackle is defining the similarity between companies, which can be useful in tasks such as competitor mapping.

To solve this problem, the Motherbrain team proposed a novel method – PAUSE: Positive and Annealed Unlabeled Sentence Embedding, which generates numerical representations from company descriptions enabling a measurement of closeness between any two companies. The model turns out to be best-in-class at defining similarity without relying heavily on pre-existing annotations by investment professionals, as other high performing models do.

Victor Engelsson, Partner within EQT Growth’s Advisory Team, commented, “EQT Growth invests in companies at the scale-up stage and uses Motherbrain and underlying novel algorithms like PAUSE to assess and make faster and more substantiated decisions. Being data driven and having access to the best market intelligence is truly one of EQT’s competitive advantages.”

High performing deep learning models have not been easily applicable in the private capital domain as they require a large amount of annotated data. The Motherbrain team is constantly moving the needle when it comes to applying advanced analytics in the private capital sector, and PAUSE will facilitate that adoption by mitigating the reliance on heavily annotated data. With this work, the Motherbrain team is also proud to contribute to the academic community with a creative method born out of a real industrial use case.