Crystallizing Financing – the Language of Decision-Making

Decision-making is a verbal activity throughout the decision’s life-cycle. The accountable needs to write down their proposal for others to digest. The pros and cons are debated and when giving a stance one might wish to express his/her reasoning behind the vote. The words convey the information and, if recorded, allow us to investigate past decisions and the thoughts expressed at the time.

But natural language is not merely a tool of information exchange, like Python is a tool of programming. The realm of person-to-person communication is much more complex and multi-faceted. Words are not defined with mathematical precision and May at once represent multiple objects or abstract concepts at once. The rich vocabulary of natural languages allows the speaker to choose from a multitude of possible options to convey his/her message. Thus, it can be useful to also tap into the world of subtle meanings. There is a difference when a decision fails and when it flops.

Fingertip collects and stores reams of text data. This rich store of information can produce true gems of insight when analysed with modern text analytics methods. Next, I will present a small glimpse of the possibilities. The data comes from Decision descriptions written in English at Fingertip. Descriptions are a small briefing on what the Decision concerns that brief the people involved about the subject matter. Thus, they are in essence a summary of the Decision.

I use modern NLP toolbox udpipe to infer a word’s basic wordform, i.e.. lemma and it’s part of speech tag. For security reasons I limit myself only to verbs. I investigate verb co-occurrence in same Decision descriptions. For example if words ”click” and ”refresh” both individually appear 10 times in the whole set of data and 5 times of those are together, they are likely semantically close words used in similar contexts. I create a network of these co-occurrences, limiting data to only words that appear a minimum of 5 times together and have a minimum correlation 0.3.

From the network we quickly see few distinct modes of talking. On bottom-right corner we have words such as funding, crystallize, receive, and finance that clearly used in Decisions that have something to do with economics. Then we have a trio of words ”thank”, ”summarize” and ”invite” that relate to human interactions, such as networking events like Fingertip Breakfast. In the largest group of words see a multitude of IT related words bungled together, such as ”navigate”, ”log”, ”click” and ”download”. These naturally relate to software development.

This quick network already reveals what activities relate to each other at Fingertip. A whole lot more information could be had when adjectives and substantives would be involved. Text analytics also allows numerous other methods for mining text data. Text is an untapped gold mine and Fingertip the natural way to collect it relating to all aspects of organization!