sentence-transformers How to use dataisgod/bge-large-fiqa-financial with sentence-transformers:
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("dataisgod/bge-large-fiqa-financial")
sentences = [
"Withdraw funds with penalty or bear high management fees for 10 years?",
"I personally invest in 4 different ETFs. I have $1000 to invest every month. To save on transaction costs, I invest that sum in only one ETF each month, the one that is most underweight at the time. For example, I invest in XIC (30%), VTI (30%), VEA (30%), and VWO (10%). One month, I'll buy XIC, next month VTA, next month, VEA, then XIC again. Eventually I'll buy VWO when it's $1000 underweight. If one ETF tanks, I may buy it twice in a row to reach my target allocation, or if it shoots up, I may skip buying it for a while. My actual asset allocation never ends up looking exactly like the target, but it trends towards it. And I only pay one commission a month. If this is in a tax-sheltered account (main TFSA or RRSP), another option is to invest in no-load index mutual funds that match the ETFs each month (assuming there's no commission to buy them). Once they reach a certain amount, sell and buy the equivalent ETFs. This is not a good approach in a non-registered account because you will have to pay tax on any capital gains when selling the mutual funds.",
"Emergency funds are good to keep yourself out of debt, for whatever reason. Job loss is a big place where an emergency fund can help you out. It buys you time to find another job before hauling out the credit cards for your groceries, falling behind on your mortgage and car payments, etc. But it can just as easily be used for major car repairs, serious medical issues, home repairs, etc. ... anything that needs to be done quickly, and isn't a discretionary item. The bigger your cash reserves, the better, especially now that the economy is bad.",
"Here's the purely mathematical answer for which fees hurt more. You say taking the money out has an immediate cost of $60,000. We need to calculate the present value of the future fees and compare it against that number. Let's assume that the investment will grow at the same rate either with or without the broker. That's actually a bit generous to the broker, since they're probably investing it in funds that in turn charge unjustifiable fees. We can calculate the present cost of the fees by calculating the difference between: As it turns out, this number doesn't depend on how much we should expect to get as investment returns. Doing the math, the fees cost: 220000 - 220000 * (1-0.015)^40 = $99809 That is, the cost of the fees is comparable to paying nearly $100,000 right now. Nearly half the investment! If there are no other options, I strongly recommend taking the one-time hit and investing elsewhere, preferably in low-cost index funds. Details of the derivation. For simplicity, assume that both fees and growth compound continuously. (The growth does compound continuously. We don't know about the fees, but in any case the distinction isn't very significant.) Fees occur at a (continuous) rate of rf = ln((1-0.015)^4) (which is negative), and growth occurs at rate rg. The OPs current principal is P, and the present value of the fees over time is F. We therefore have the equation P e^((rg+rf)t) = (P-F) e^(rg t) Solving for F, we notice that the e^rg*t components cancel, and we obtain F = P - P e^(rf t) = P - P e^(ln((1-0.015)^4) t) = P - P (1-0.015)^(4t)"
]
embeddings = model.encode(sentences)
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4] 
