Theophile Cantelo has created Foudinge, a knowledge graph connecting restaurants and chefs. Leveraging Large Language Models (LLMs), Foudinge extracts information from various online sources like blogs, guides, and social media to establish relationships between culinary professionals and the establishments they've worked at or own. This allows for complex queries, such as finding all restaurants where a specific chef has worked, discovering connections between different chefs through shared work experiences, and exploring the culinary lineage within the restaurant industry. Currently focused on French gastronomy, the project aims to expand its scope geographically and improve data accuracy through community contributions and additional data sources.
The notebook demonstrates how Vision Language Models (VLMs) like Donut and Pix2Struct can extract structured data from document images, surpassing traditional OCR in accuracy and handling complex layouts. Instead of relying on OCR's text extraction and post-processing, VLMs directly interpret the image and output the desired data in a structured format like JSON, simplifying downstream tasks. This approach proves especially effective for invoices, receipts, and forms where specific information needs to be extracted and organized. The examples showcase how to define the desired output structure using prompts and how VLMs effectively handle various document layouts and complexities, eliminating the need for complex OCR pipelines and post-processing logic.
HN users generally expressed excitement about the potential of Vision-Language Models (VLMs) to replace OCR, finding the demo impressive. Some highlighted VLMs' ability to understand context and structure, going beyond mere text extraction to infer meaning and relationships within a document. However, others cautioned against prematurely declaring OCR obsolete, pointing out potential limitations of VLMs like hallucinations, difficulty with complex layouts, and the need for robust evaluation beyond cherry-picked examples. The cost and speed of VLMs compared to mature OCR solutions were also raised as concerns. Several commenters discussed specific use-cases and potential applications, including data entry automation, accessibility for visually impaired users, and historical document analysis. There was also interest in comparing different VLMs and exploring fine-tuning possibilities.
Summary of Comments ( 16 )
https://news.ycombinator.com/item?id=43242818
Hacker News users generally expressed skepticism about the value proposition of the presented knowledge graph of restaurants and chefs. Several commenters questioned the accuracy and completeness of the data, especially given its reliance on LLMs. Some doubted the usefulness of connecting chefs to restaurants without further context, like the time period they worked there. Others pointed out the existing prevalence of this information on platforms like Wikipedia and guide sites, questioning the need for a new platform. The lack of a clear use case beyond basic information retrieval was a recurring theme, with some suggesting potential applications like tracking career progression or identifying emerging culinary trends, but ultimately finding the current implementation insufficient. A few commenters appreciated the technical effort, but overall the reception was lukewarm, focused on the need for demonstrable practical application and improved data quality.
The Hacker News post titled "Show HN: Knowledge graph of restaurants and chefs, built using LLMs" generated a moderate amount of discussion, with a focus on the practical application and potential limitations of the project.
Several commenters expressed interest in the project's potential, particularly regarding its use for restaurant recommendations. One commenter highlighted the difficulty of finding good restaurants in unfamiliar cities and suggested the knowledge graph could be helpful in this scenario, particularly if it allowed users to filter by cuisine type and other specific criteria. They also inquired about the possibility of incorporating user reviews or ratings into the system.
Another user echoed this sentiment, pointing out that existing restaurant recommendation platforms often rely on outdated or inaccurate information. They envisioned the project as a valuable tool for both diners and restaurant owners, providing a centralized and up-to-date resource for restaurant information.
However, some commenters expressed concerns about the project's reliance on LLMs. One commenter pointed out the potential for hallucinations and inaccuracies in LLM-generated data, emphasizing the importance of thorough verification and fact-checking. They also questioned the long-term viability of relying solely on LLMs for data collection and maintenance, suggesting that a more robust approach might involve incorporating human input and curation.
The creator of the project engaged with the commenters, acknowledging the challenges of LLM-based data generation and outlining plans to address these concerns. They mentioned plans to implement a feedback mechanism to flag inaccurate information and explore methods for verifying the accuracy of LLM-generated data. They also discussed potential future features, such as incorporating user reviews, dietary information, and real-time menu updates.
A recurring theme in the comments was the need for a practical application or interface for the knowledge graph. Commenters suggested various use cases, including a dedicated search engine for restaurants, a mobile app for on-the-go recommendations, and integration with existing restaurant platforms.
Finally, one commenter raised a broader point about the ethical implications of using LLMs to scrape data from the web, questioning the potential impact on website owners and the overall ecosystem of online information. This sparked a brief discussion about the responsible use of LLMs and the importance of respecting website terms of service. While not directly related to the project itself, this comment highlighted the broader ethical considerations surrounding LLM-driven data collection.