Fintech: Language algos may struggle to spot the BS
A great amount of effort from the buy and sell sides goes into analysis of the statements of executives on conference calls, but can’t machines do that for us, these days?
Sure enough, in April and May, UBS put its Transcriptlytics tool to work on 250 earnings calls to extract a measure of management confidence (net sentiment) as well as of language complexity.
The tool then traces sector patterns for rising and falling confidence relative to nine years’ worth of past earning-call transcripts and focuses on individual company stories where the conference-call language indicates heightened complexity.
So, confidence is apparently rising in the commercial services sector and in healthcare, but falling in consumer services and in real estate, for example.
Worryingly, UBS points out that semiconductors are often a leading indicator of macro trends for the rest of the market and confidence is falling sharply there as managements discuss trade wars and inventory levels.
My assumption has always been that everything before the ‘but’ is bullshit - Banks analyst
For our favourite sector, banks, Transcriptlytics finds top quintile positive moves in management net sentiment at Banco BPM, Banco de Sabadell, Bankia, Danske Bank, HSBC and Intesa Sanpaolo. So, that sounds like terrific news for banking.
Or is it? When Euromoney speaks to the head of investor relations at a bank not on that list and who always complains that the stock market does not appreciate the positives of his bank, Euromoney suggests he think about tools such as Transcriptlytics.