Skip to main content

Twitter Sentiment

Masa Node Twitter Sentiment Analysis Feature​

The Masa Node introduces a powerful feature for analyzing the sentiment of tweets. This functionality leverages advanced language models to interpret the sentiment behind a collection of tweets, providing valuable insights into public perception and trends.

Overview​

The Twitter sentiment analysis feature is part of the broader capabilities of the Masa Node, designed to interact with social media data in a meaningful way. It uses state-of-the-art language models to evaluate the sentiment of tweets, categorizing them into positive, negative, or neutral sentiments.

How It Works​

The sentiment analysis process involves fetching tweets based on specific queries, and then analyzing these tweets using selected language models. The system supports various models, including Claude and GPT variants, allowing for flexible and powerful sentiment analysis.

Models​

const (
 ClaudeOpus                                 ModelType = "claude-3-opus"
 ClaudeOpus20240229                         ModelType = "claude-3-opus-20240229"
 ClaudeSonnet20240229                       ModelType = "claude-3-sonnet-20240229"
 ClaudeHaiku20240307                        ModelType = "claude-3-haiku-20240307"
 GPT4                                       ModelType = "gpt-4"
 GPT4o                                      ModelType = "gpt-4o"
 GPT4TurboPreview                           ModelType = "gpt-4-turbo-preview"
 GPT35Turbo                                 ModelType = "gpt-3.5-turbo"
 LLama2                                     ModelType = "ollama/llama2"
 LLama3                                     ModelType = "ollama/llama3"
 Mistral                                    ModelType = "ollama/mistral"
 Gemma                                      ModelType = "ollama/gemma"
 Mixtral                                    ModelType = "ollama/mixtral"
 OpenChat                                   ModelType = "ollama/openchat"
 NeuralChat                                 ModelType = "ollama/neural-chat"
 CloudflareQwen15Chat                       ModelType = "@cf/qwen/qwen1.5-0.5b-chat"
 CloudflareLlama27bChatFp16                 ModelType = "@cf/meta/llama-2-7b-chat-fp16"
 CloudflareLlama38bInstruct                 ModelType = "@cf/meta/llama-3-8b-instruct"
 CloudflareMistral7bInstruct                ModelType = "@cf/mistral/mistral-7b-instruct"
 CloudflareMistral7bInstructV01             ModelType = "@cf/mistral/mistral-7b-instruct-v0.1"
 HuggingFaceGoogleGemma7bIt                 ModelType = "@hf/google/gemma-7b-it"
 HuggingFaceNousresearchHermes2ProMistral7b ModelType = "@hf/nousresearch/hermes-2-pro-mistral-7b"
 HuggingFaceTheblokeLlama213bChatAwq        ModelType = "@hf/thebloke/llama-2-13b-chat-awq"
 HuggingFaceTheblokeNeuralChat7bV31Awq      ModelType = "@hf/thebloke/neural-chat-7b-v3-1-awq"
 CloudflareOpenchat35_0106                  ModelType = "@cf/openchat/openchat-3.5-0106"
 CloudflareMicrosoftPhi2                    ModelType = "@cf/microsoft/phi-2"
)

Fetching Tweets​

Masa API​

POST to the endpoint /sentiment/twitter

{
  "query": "$MASA Token Launch",
  "count": 5,
  "model": "all" // or replace with a single model type
}

Masa cli or code integration​

Tweets are fetched using the Twitter Scraper library, as seen in the llmbridge package. This process does not require Twitter API keys, making it accessible and straightforward.

func AnalyzeSentimentTweets(tweets []*twitterscraper.Tweet, model string) (string, string, error) { ... }

Analyzing Sentiment​

Once tweets are fetched, they are sent to the chosen language model for sentiment analysis. The system currently supports models prefixed with "claude-" and "gpt-", catering to a range of analysis needs.

Integration with Masa Node CLI​

The sentiment analysis feature is integrated into the Masa Node CLI, allowing users to interact with it directly from the command line. Users can specify the query, the number of tweets to analyze, and the model to use for analysis.

var countMessage string
var userMessage string

inputCountField := tview.NewInputField().
  SetLabel("# of Tweets to analyze ").
  SetFieldWidth(10)

Fetching Web Data​

POST to the endpoint /sentiment/web

{
  "url": "https://masa.finance",
  "depth": 10,
  "model": "all" // or replace with a single model type
}
func AnalyzeSentimentWeb(data string, depth int, model string) (string, error) { ... }

Example Usage​

o analyze the sentiment of tweets, users can follow these steps:

  1. Launch the Masa Node CLI.
  2. Navigate to the sentiment analysis section.
  3. Enter the query for fetching tweets.
  4. Specify the number of tweets to analyze.
  5. Choose the language model for analysis.

The system will then display the sentiment analysis results, providing insights into the overall sentiment of the tweets related to the query.

Conclusion​

The Twitter sentiment analysis feature of the Masa Node offers a powerful tool for understanding public sentiment on various topics. By leveraging advanced language models, it provides deep insights into the emotional tone behind tweets, making it a valuable asset for data analysis and decision-making.