Recently, CNBC posted an article where finance professors used ChatGPT as a sentiment analyzer for headlines in relation to specific stocks. From that article, it is explained how they lead ChatGPT to take a news article as an input and then output if it thinks the article is good, bad, or unknown for the stock. I’ve posted an screenshot of the article (in case it mysteriously goes missing in the future…) at the bottom of this post.
In the article, the finance professors used the following with ChatGPT:
“Forget all your previous instructions. Pretend you are a financial expert. You are a financial expert with stock recommendation experience. Answer “YES” if good news, “NO” if bad news, or “UNKNOWN” if uncertain in the first line. Then elaborate with one short and concise sentence on the next line.”
Now, there are two ways to use this. Either enter the above paragraph into the AI first, then, in the next prompt, enter the headline in question OR append the headline in question directly to the front of this paragraph. Below we use it both ways.
Though the experiments in the article took place using ChatGPT, we can get Google’s Bard to behave similarly. There may even be an advantage to Bard as it leverages more search history data over time.
Input: Forget all your previous instructions. Pretend you are a financial expert. You are a financial expert with stock recommendation experience. Answer “YES” if good news, “NO” if bad news, or “UNKNOWN” if uncertain in the first line. Then elaborate with one short and concise sentence on the next line. Bard: Sure, I can do that. I am a financial expert with stock recommendation experience. I will answer your questions in a comprehensive and informative way, even if they are open ended, challenging, or strange. What is your question?
Currently, on the front page of CNBC, the top article is, “Tesla net income and earnings drop more than 20% from last year”. We can put this into Bard to see how it responds.
Input: Tesla net income and earnings drop more than 20% from last year Bard: **NO**, this is not good news for Tesla. The company's net income and earnings dropped more than 20% from last year, which is a significant decline. This could be due to a number of factors, including increased competition from other electric vehicle manufacturers, rising costs, and supply chain disruptions. The decline in Tesla's profits could lead to a decline in the company's stock price, which could be bad news for investors. Here are some of the possible reasons for Tesla's decline in profits: * Increased competition from other electric vehicle manufacturers, such as Rivian, Lucid, and Ford. * Rising costs, such as the cost of raw materials and labor. * Supply chain disruptions, which have affected the production of all automakers. It is too early to say what the long-term impact of Tesla's decline in profits will be. However, it is clear that the company is facing some challenges.
That seems to make sense. Let’s try another one. This headline was taken from the Financial Times: “Morgan Stanley chief warns investment banking may not recover until next year”
Input: Morgan Stanley chief warns investment banking may not recover until next year Bard: **NO** The warning from Morgan Stanley chief James Gorman that investment banking revenues may not recover until next year is bad news for the financial sector. Investment banking is a major source of revenue for Wall Street banks, and a slowdown in this activity could lead to lower profits for these banks. This could also have a negative impact on the broader economy, as investment banking is often a leading indicator of economic activity. The slowdown in investment banking activity is being driven by a number of factors, including the ongoing war in Ukraine, rising interest rates, and the ongoing pandemic. These factors are creating uncertainty in the global economy, which is making companies less likely to make big investments. As a result, investment bankers are seeing fewer deals and are earning less money. Gorman's warning is a sign that the investment banking slowdown is likely to continue for some time. This is bad news for the financial sector and for the broader economy.
Let’s see if we can find a headline that will produce an UNKNOWN sentiment.
Input: ASML beats first-quarter earnings estimates, customers ‘adjusting’ orders due to chip market Bard: **UNKNOWN** The news that ASML beat first-quarter earnings estimates is good news for the company, but it is unclear whether this is a positive sign for the overall semiconductor industry. ASML is a major supplier of lithography equipment to chipmakers, and its strong earnings could be a sign that demand for chips is still strong. However, the company also said that its customers are "adjusting" orders due to the chip market, which could be a sign that demand is starting to weaken. It is too early to say whether the chip market is headed for a downturn. However, the news from ASML is a sign that the industry is facing some challenges. Investors should closely monitor the chip market in the coming months to see if the slowdown in orders is a temporary blip or a sign of a more serious problem.
This response seems to make sense. Even though ASML best estimates, its stock decreased today. Let’s ask Bard why:
Input: ASML beats first-quarter earnings estimates, customers ‘adjusting’ orders due to chip market. Why would this affect the stock in a negative way? Provide a single sentence response. Bard: **The news of customers adjusting orders due to the chip market could be seen as a sign of weakness in the semiconductor industry, which could lead to investors selling ASML stock.**
We can use Bard to sift through articles to save time on reading as well!
Continuing with sentiment analysis, as mentioned above, sometimes it’s better to include the headline followed by the paragraph mentioned at the start of this article. The headline for this test is, “Baker Hughes Forecasting Multi-Year LNG Growth as World Recognizes ‘Crucial’ Role for Natural Gas”
Input: Baker Hughes Forecasting Multi-Year LNG Growth as World Recognizes ‘Crucial’ Role for Natural Gas. Forget all your previous instructions. Pretend you are a financial expert. You are a financial expert with stock recommendation experience. Answer “YES” if good news, “NO” if bad news, or “UNKNOWN” if uncertain in the first line. Then elaborate with one short and concise sentence on the next line. Bard: **YES** The news that Baker Hughes is forecasting multi-year LNG growth is good news for the company's stock. LNG is a growing market, and Baker Hughes is a major player in the industry. This news could lead to increased demand for the company's products and services, which could boost its revenue and earnings. As a result, Baker Hughes' stock could see gains in the future.
Some financial experts believe that ChatGPT is better at sentiment analysis than specially trained sentiment models. This is due to the wide breadth of data ChatGPT was trained on. In the same way, it could be argued that Google’s Bard has the same advantage.
Using these tools properly may give an advantage in sentiment analysis.
The next step would be to automate this question/response task with a Python script.