There are trade signals within your audio data

Anthony Tassone

July 25, 2019

Identification and interpretation of pace, volume, pitch, and cadence are part of the human brain’s processing ability to identify sentiment, based on years of training.

Natural Language Processing (NLP) technology refers to leveraging computers and statistics to process and make sense of human language in a systematic way. It can be used to transcribe voice and to extract keywords, intents and sentiment at scale.

A firm might use GK to understand the sentiment across a sales trading desk for a particular instrument type, or for markets in general over a period of months. It could be used more specifically to monitor real-time sentiment scanning for a change in an influential customer’s voice, just as a seasoned trader would pick up on.

We view the combination of these things: the financial products discussed (product mentions), and how they are discussed (sentiment) as a relatively untapped source of alpha.

As firms go in search of new so called “alternative data sources” it’s important not to overlook the highly valuable unstructured data that may already be sitting on their servers – such as historical audio files, or being  produced in real time via chat or telephony systems.

Example:

“signals”:

“dateTime”: “07/01/2019 @ 5:44:000pm (UTC)”,

“signalType”: “price”,

“signalDetails”: {

“priceList”: [{

“value”: 105.6,

“category”: “bid”,

“type”: “notional”

 

“dateTime”: “07/01/2019 @ 5:44:000pm (UTC)”,

“signalType”: “sentiment”,

“signalDetails”: {

“velocity”: 125,

“cadence”: 140,

“tone”: {

“signalRange”: 3,

“clipping”: 20,

“meanFrequency”: 100

 

“urgency”: {

“value”: 5,

“label”: “low”

 

“polarity”: {

“value”: 0.7,

“label”: “positive”

 

“dateTime”: “07/01/2019 @ 5:44pm (UTC)”,

“signalType”: “keyWord”,

“signalDetails”: {

“keyWordList”: [

“firm”

Originally published on LinkedIn.