Very helpful video on the use of Gemini and Open AI APIs for sentiment analysis. Thanks for uploading. Can you also share the steps to generate the Google API key?
I have a dataset with around 4M reviews, I've tried doing sentiment analysis on it using the Transformers pipeline with distilbert but it was way too slow. can you tell me how many batches should I make If I have to use chatgpt API and what would be a cost and time-efficient solution?
Dear learner, before jumping onto ChatGPT API, we would recommend you to try out programmatic labelling for your specific use case of Sentiment Label Generation - using the free Snorkel framework. Otherwise, to answer your question, batch-size will depend on the context window of the Model you are using. Here's a cost sheet you may refer to for your evaluation. It has details on context window as well: docs.google.com/spreadsheets/d/1A57gqIpTDC6qEgZIal3tsTvFzo2oxg-5M0ky_drAdM8/edit?usp=sharing
is it possible to adjust the prompt so that it can analyze the text and not just identify positive and negative reviews, but also identify recurring issues and summarize what it reads?
Absolutely, yes. You may modify your prompt- asking model to capture issues mentioned in reviews and share that as a separate parameter in the output json. For example, apart from the two parameters we have: Review and Sentiment Label, you may add another one called- Issue.
I am working on a project that requires to perform the sentiment analysis on comments of social media posts which approach is better LLM or NLP? Or got any other suggestions to make it better?
i am writing a project which inolves sentiment analysis of a particular topic on twitter. i have already extracted the data, how will you suggest i label it
i have a project which takes tweets and analyses trends in a given area : traffic trends , political trends , restraunts , etc , instead of NLP can i use gemini or open ai?
I want to use the same approach for aspect-based sentiment analysis, the dataset has sentence, sentiment, target and aspect. the model should predict sentiment, target and aspect i.e. multi-head prediction is it possible with this approach.
Great use case. You may try out the discussed prompt engineering approaches for this use case. Depending on the size of your labelled dataset, you may also try out fine-tuning a smaller Language Model, like quantized Llama 3.
You may use some programmatic labelling technique, like snorkel. Else generate labels, and then manually validate a sample (which is statistically significant) for your error tolerance.
i am working on project about share market trend for last yr using Gemini API can i use this project by replacing dataset with share market dataset? provide some guidelines to this project.
Sentiment analysis project (video) focuses on text, not ideal for stock trends. Here's how to adapt for your project: 1. Use share market data APIs (e.g., Alpha Vantage) for historical data (past year). 2. Transform data (e.g., moving averages) for analysis. 3. Optional: Build a model (machine learning) to predict future trends (be cautious!). 4. Use Gemini to analyze data, identify patterns, or visualize trends. Good luck with your project!
Book FREE 1:1 Mentorship for Gen AI / Data Science
Link 🔗 bit.ly/3wlIIGz
Very helpful video on the use of Gemini and Open AI APIs for sentiment analysis. Thanks for uploading. Can you also share the steps to generate the Google API key?
Check out this tutorial for Gemini API Key: ruclips.net/video/pJAzg8CORYI/видео.html
I have a dataset with around 4M reviews, I've tried doing sentiment analysis on it using the Transformers pipeline with distilbert but it was way too slow. can you tell me how many batches should I make If I have to use chatgpt API and what would be a cost and time-efficient solution?
Dear learner, before jumping onto ChatGPT API, we would recommend you to try out programmatic labelling for your specific use case of Sentiment Label Generation - using the free Snorkel framework.
Otherwise, to answer your question, batch-size will depend on the context window of the Model you are using. Here's a cost sheet you may refer to for your evaluation. It has details on context window as well: docs.google.com/spreadsheets/d/1A57gqIpTDC6qEgZIal3tsTvFzo2oxg-5M0ky_drAdM8/edit?usp=sharing
Become Always nice video
1 no video
is it possible to adjust the prompt so that it can analyze the text and not just identify positive and negative reviews, but also identify recurring issues and summarize what it reads?
Absolutely, yes. You may modify your prompt- asking model to capture issues mentioned in reviews and share that as a separate parameter in the output json.
For example, apart from the two parameters we have: Review and Sentiment Label, you may add another one called- Issue.
love you my brother
I am working on a project that requires to perform the sentiment analysis on comments of social media posts which approach is better LLM or NLP? Or got any other suggestions to make it better?
Ideally try both.
Specific to LLMs, you may use Gemini API that comes free for development purpose.
🔥🔥🔥🔥🔥🔥
i am writing a project which inolves sentiment analysis of a particular topic on twitter. i have already extracted the data, how will you suggest i label it
How did you extract data from Twitter? Can you kindly share? A program or a paid service?
ChatGPT API not generating response and displaying RateLimitError. What to do?
Try reducing batch size and adding increased delay using Python time sleep module.
i have a project which takes tweets and analyses trends in a given area : traffic trends , political trends , restraunts , etc , instead of NLP can i use gemini or open ai?
Gemini API comes for free for development purpose. OpenAI also give $5 worth of free credits.
I want to use the same approach for aspect-based sentiment analysis, the dataset has sentence, sentiment, target and aspect. the model should predict sentiment, target and aspect i.e. multi-head prediction is it possible with this approach.
Great use case. You may try out the discussed prompt engineering approaches for this use case. Depending on the size of your labelled dataset, you may also try out fine-tuning a smaller Language Model, like quantized Llama 3.
where are you fine tunning the gemini model???
In the Few Shot prompting part - where we are giving examplers.
how can I save this model ??
What if you are working with a dataset that has no true values?
Even you don't any able Google model will able to predict the position and negative sentence. Which is not clear . Better to use labelled data
You may use some programmatic labelling technique, like snorkel.
Else generate labels, and then manually validate a sample (which is statistically significant) for your error tolerance.
where is the Google Colab file ?
Check the link in description part of the video.
i am working on project about share market trend for last yr using Gemini API
can i use this project by replacing dataset with share market dataset? provide some guidelines to this project.
Sentiment analysis project (video) focuses on text, not ideal for stock trends.
Here's how to adapt for your project:
1. Use share market data APIs (e.g., Alpha Vantage) for historical data (past year).
2. Transform data (e.g., moving averages) for analysis.
3. Optional: Build a model (machine learning) to predict future trends (be cautious!).
4. Use Gemini to analyze data, identify patterns, or visualize trends.
Good luck with your project!