Weaviate Shares Late-Interaction Multi-Vector Retrieval Update
Weaviate posted about late-interaction multi-vector retrieval for embedding PDF pages as images.
Connor Shorten retweeted a post from Weaviate stating the company stopped asking PDFs to become text before searching them. The post describes late-interaction multi-vector retrieval that embeds PDF pages as images. A generated headline in the packet reads Weaviate Launches Multi-Vector Retrieval For Direct PDF Image Search. A generated source summary adds that the approach enables search over charts and tables.
Combined views
753
4 posts, first seen 22h ago
Reactions
4 postsWA 
Weaviate AI Database@weaviate_ioWe stopped asking PDFs to become text before we searched them. With late-interaction multi-vector retrieval, you can embed each PDF page as an image and search the charts, tables, and layouts that text extraction misses. We tested it on 92 pages of NVIDIA investor decks. A…
- likes: 19
- replies: 1
- bookmarks: 14
- reposts: 7
PV Prem Viswanathan@prempvThis is very cool. Late interaction techniques over multimodal docs can truly help tackle the long tail of retrieval. cc @lateinteraction
- likes: 9
- replies: 0
- bookmarks: 2
- reposts: 1
WA Weaviate AI Database@weaviate_ioWe stopped asking PDFs to become text before we searched them. With late-interaction multi-vector retrieval, you can embed each PDF page as an image and search the charts, tables, and layouts that text extraction misses. We tested it on 92 pages of NVIDIA investor decks. A…
- likes: 19
- replies: 1
- bookmarks: 14
- reposts: 7
PV Prem Viswanathan@prempvThis is very cool. Late interaction techniques over multimodal docs can truly help tackle the long tail of retrieval. cc @lateinteraction
- likes: 9
- replies: 0
- bookmarks: 2
- reposts: 1
