You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for this interesting project using the open-source model. I have a question about the representation choice. Inspired by your blog post example, suppose I have one single sentence, "Mary will try not to eat lamb," and for this single sentence, we generate a graph as "Mary--avoids-->lamb". Can we still use this as a Knowledge graph representation, as it seems more like a triplets graph representation, not KG? Also, KG and graph of concepts were used interchangeably, so this clarification will help me in this regard.
The text was updated successfully, but these errors were encountered:
It is GRAPH representation if you deal with nodes/edges appropriately.
I merged/concatenate/dedup same nodes and edges. Then vectorized them.
We can extend question into multi questions (use simple whatever LLM), then search though NODES only. from these NODES then we search ALL EDGES connected (1 hop, more if you will, I can do 3 and it took sometime). Run rerank (we build our own with Gemms-2-2B) wirh series of above extended questions, then voila you have just queried a graph knowledge DB.
Thanks for this interesting project using the open-source model. I have a question about the representation choice. Inspired by your blog post example, suppose I have one single sentence, "Mary will try not to eat lamb," and for this single sentence, we generate a graph as "Mary--avoids-->lamb". Can we still use this as a Knowledge graph representation, as it seems more like a triplets graph representation, not KG? Also, KG and graph of concepts were used interchangeably, so this clarification will help me in this regard.
The text was updated successfully, but these errors were encountered: