Tingle- A Personalized Search Engine

Discover our personalized engine app, empowering users to customize their online experience.

Comment

Inspiration

With explosive growth of information and technology accessibility, current search engines are catered to deliver customizable experience to users of different age groups and skill levels. Tingle is a simple yet effective. Hidden sensitive content can be triggers for individuals. Personalized search engine that would customize your experience by providing a Safe, reliable, configurable, and efficient search experience for its users according to their demographics and technology expertise.

What it does

The user can enter a topic of choice in either of the two search engines provided-> 1. Google & 2. Youtube. Tingle provides users with source, theme, and summary of search results without leaving the search experience. It flags potentially disturbing keywords or themes and allows users to adjust confidence levels and block sources. Tingle can search across multiple media types and adapts to user preferences over time. It also offers a safe search experience for kids with parental controls and notifies parents of concerning search behavior. Users can use an interactive chat tool to navigate their search history and create a knowledge journal.

How we built it

Employed a custom Google search engine and the YouTube API to retrieve links. Utilized LlamaIndex to load and index scraped content and transcriptions for Google links and YouTube videos, respectively, storing the data in Astradb using session state. Employed SimpleDirectoryReader for document loading. Implemented LlamaIndex VectorStoreIndex to establish a vector store, enabling the model to swiftly retrieve context from the data. Developed a query engine and employed OpenAI for summarization, trigger identification, and theme selection. Employed the PDF parser of Llama Hub to parse a trigger guideline PDF for reference.

Challenges we ran into

Navigating the learning curve with Astradb and Llama Index as newcomers. Grappling with the complexities of UI development due to limited prior experience. Facing constraints with data scraping capabilities. Managing restrictions imposed by the Google and YouTube APIs.

Usability

Here's how such a platform enhances usability:

  1. Customized Preferences: Users can set up personalized preferences based on their interests, age, language, and other relevant factors. This customization ensures that search results align with their specific needs and preferences.

  2. Like, Dislike, and Block Options: Users have the ability to interact with search results by liking, disliking, or blocking specific pages. This feature allows users to provide feedback on the relevance and quality of search results, enabling the search engine to fine-tune its recommendations over time.

  3. Parental Control Features: For parents, the ability to monitor and control their children's online activities is paramount. With the personalized dSearch engine, parents can block access to inappropriate or undesirable content, ensuring a safe and age-appropriate browsing experience for their children.

  4. Safe Browsing Environment: By leveraging user feedback and preferences, the dSearch engine can filter out potentially harmful or objectionable content, creating a safer browsing environment for users of all ages.

  5. Enhanced Relevance: Personalized search results are more relevant and tailored to the user's interests, leading to a more efficient and satisfying search experience. This increased relevance saves users time and effort by presenting them with the most pertinent information upfront.

  6. Continuous Improvement: The dSearch engine continuously learns from user interactions and feedback, refining its algorithms to deliver increasingly accurate and personalized search results over time. This iterative process ensures that the search engine remains responsive to evolving user needs and preferences.

What we learned

  1. Learned about llamaindex and astradb capabilities.
  2. importance of prompt engineering

What's next for Tingle- A Personalized Search Engine

  1. Architectured to add more configurations over time without major impact to key functionalities.
  2. Currently dependant on certain API’s to derive search results, in future can expand to crawl and build self search capabilities
  3. Could be scaled to add more integrations from social micro blogs link X, Threads, etc.
  4. Seamless experience across multiple software platforms and devices.
  5. Easy plug and play capability to adapt to new AI models.

Built With

Share this project:

Updates