Can you break contextual frame analysis down into laymen's terms please? How is your PoC coming along for this? How is contextual frame analysis important to the long term sustained success of the investment platform?
Hey, thanks for your question. So, contextual frame analysis (CFA), is one of the more hypothetical elements of the platform. It is not necessary for its operation, but we believe it would serve to further reduce risk and increase returns by improving accuracy as compared to the more standard sentiment analysis approaches - of which there are many.
In the simplest terms, CFA is an attempt to capture how language use (contextual frames) drives decision making. The seminal example given is the 'Virus vs Beast' study by Thibodeau and Boroditsky (2011), in which an identical article about a crime wave - except a one word difference - is shown to two groups of people. One article used the word 'beast' and the other used 'virus'. The consensus of the group that saw 'beast' was that punitive action should be taken against the criminals, however, the group that saw 'virus' leaned strongly toward rehabilitation. When the people in these groups were asked to justify their decision, nobody mentioned the virus/beast word usage, and quoted data/statistics from the article. (See pp. 28-29 in our white paper)
That is, people came to
opposing decisions while justifying them with the
same data, driven by the
contextual frame in which they saw the data. We are working with Mieke, a linguistic & expert in frame analysis, to develop a strategy that enables us to determine whether these 'contextual frames' are more likely to drive an investor toward 'bullish' or 'long' vs 'bearish' or 'short' when consuming news media and other data sources from which they make their investment decision. Underpinning this theory is the concept that decision making is, to a significant extent, emotional - even when investors insist their actions are data driven, they are ultimately slaves to the cognitive decision making processes we all evolved with.
At the moment we are developing a manual CFA process based on 'traditional' linguistic frame analysis as a proof-of-concept on a relatively small number of news articles, as this is a novel idea in both investment and linguistics. Once the relationship between frames and responses is established in a few key examples, we will quickly scale it to operate on a large scale using natural language processing techniques within artificial intelligence and machine learning, such that our algorithms can learn to optimally identify contextual frames based on historical data. We aim to have this operational along with the rest of the platform - however, this is a scientific endeavour, and as such may be subject to further development before we release it in the wild.
It is probably the most exciting of the linguistic analysis approaches we're taking, but it is not mission critical to operation of the fund - it will simply serve to enhance it.
Let me know if you have any more questions. If you're interested in the linguistics side of things, Mieke hangs out in the Telegram group.
James
