I am embedding a contact list .csv file with multiple columns (first_name, last_name, title, industry, location) using the text-embedding-ada-002 engine from OpenAI.
To return contacts based on semantic search sentences such as “find me all the managers in the hospitality industry”, ChatGPT recommended embedding each column individually and then combine each column’s embedding array into one big array to be uploaded to Pinecone and queried.
Thus I currently have an index with 7680 dimensions (5x each column’s 1536 dimension embedding) and when I search the db I add a single search phrase embedding (ie. find me all the managers in the hospitality industry) 5 times into one big embedding array so the dimensions match the database that is 7680.
However this approach seems to produce dimensions that are unnecessarily large and the results aren’t quite accurate (ie. returning teachers closer to the top of list when I search for professors, which while directionally correct doesn’t return the most accurate results at the top of list first)
Is this the wrong way to perform embeddings in this case and are there ways to make the results more accurate? Thank you very much.
Very clever @Jasper! Definitely, a better way to think about the source data than just raw columnar data.
I think that’s one of the biggest tripping points for people who are new to vector databases. The traditional methods of data storage (relational, key-value, columnar) all depend on the data being in a specific order with assigned values. But with vectors, it’s not the individual components of the data that matter, but the set’s overall value. Adding the additional context of “ is a at <location” transforms the raw source into something that a human or AI would more naturally process. And that allows for a better vector representation of the data as a whole.
If I want to filter out irrelevant results that would naturally show up near the end of the returned data (ie. if topK is large enough), would it come down to trail and error to find a similarity score cutoff that works well? (ie. if found that results < 0.75 will start returning contacts from Canada even though asked for US then set the filter cutoff at 0.75 when making the search)
Search now more accurate overall with only 1536 dimensions compared to the previous 7680.
However some items that shouldn’t be in the top results appear there.
For example, the search “find me all the teachers in france” returns a contact from the United States as the second result with a score of 0.8012. The “location_country” of this contact is United States, however the job title is “Enseignant” which is French for teacher which may have caused the result to be placed higher up the list.
Not sure if the results are as accurate as they can currently be with existing embedding libraries and I would need to do a hybrid of vector + keyword search to get more accurate results.
(ie. use vector search to identify closest categories based on the csv/database columns that exist and run a traditional search filter using the returned categories)
I am no expert, but I read somewhere that you can add information to the metadata, that can then be used for non-semantical querying.
Vector databases are specifically designed for unstructured data and yet provide some of the functionality you’d expect from a traditional relational database. They can execute CRUD operations (create, read, update, and delete) on the vectors they store, provide data persistence, and filter queries by metadata. When you combine vector search with database operations, you get a powerful tool with many applications.
Source:
I have a similar problem where it’s not accurate at traditional database querying:
Example: If I wanted the find the user with the most posts.
It doesn’t filter the data in order as the overall value (Vectorised Position) is not based on the numerical value.
Solution 1:
I read a solution to this is to base vectorised values in relation to the numerical values, but this seems unintuitive and non-extendable.
Solution 2:
Embed the Numerical values as metadata and Fitler the data from a traditional sense.
Im creating this using Langchain js, and Solution 2 requires hardcoding the filter of metadata. I don’t know if this is the best solution, but I feel nervous about allowing a prompts response to set this filter+ it adds an additional Api call.
Yes! Meatadata is the place to store additional data, you want to use for filtering data when querying your searches.
While I am still not completely sure what you want to query (you have to query something - vector database needs a vector query one way or the other), I think COUNT() is not supported.
You can store this data in Pinecone, but I think it would be a long way around I guess. How I would do it.
For each user, I would create something in the lines of: ‘User with username Jasper has 10 posts.’. I would embedd this data and upsert it directly into Pinecone with an ID I would hand pick and know. I would then add some metadata { username, num_posts, datetime }. Okey, all good and well so far.
Every time a user would create a new post I would query the index with ‘How many posts does Jasper have?’ and set the top-k to 3. I would check each results metadata it the username is indeed the one I seek. Then just update the vector with new value, embedd again and update the index.
Next comes the querying. Here you could have a problem and would have to test how good the solution is. The biggest problem is, how to get the one with the MOST posts. Here you will have to see if querying actually returns results with results that make sense.
All in all, I think for your usecase vector database doesn’t provide what you want. Or I am not understanding your usecase completely (which is also possible ). Either way, let me know if I missed the point of your problem or if it helped you think it through.
I don’t know what I want, so I want to catch ‘every’ use case of a human trying to find data from a database (I’ve started small with a CSV, but want to scale on successful solutions), Im using my companies Slack Analytics as a proof of concept and to explore its limitations.
So looking into Langchain possibilities,
And you are right I can query with a filter of metadata
for example:
I can iterate with whatever the top messagesPosted value from the results from top-k
R: Ian = 5000
R: jef = 200
R: eli = 500
Q: MessagesPosted : {‘$gt’:5000}
R: Paul = 12200
…
Q: MessagesPosted : {‘$gt’:12200}
R: Paul = 144000
…
Q: MessagesPosted : {‘$gt’:144000}
…
…
But this seems very un-optimised and when if this was a SQL database I could just have 1 SQL query to return the row which contains the highest MessagesPosted
What I’ve concluded is that:
I’m trying to use a vectorised database just like it was a traditional database when I should see it more as a semantical database with semantical queries (Is the query/value subjective?). It will cover some of my ‘every’ use cases.
So what did you decide to go with? The Vectorised DB approach or Text to SQL?
I’m working on a similar problem and am kind of in the same situation as you. Especially as my data has a heck of a lot more numerical values. What would be the best method to move forward?
Could you fine-tune GPT to re-interpret the results of your vector query? Use the questions you would normally ask with examples of the expected answers. Then, when you tell the fine tuned model to interpret the results of the query for “find me all the teachers in france” it would know to ignore the “Enseignant” because you included an example in your fine-tuning data?
I have a similar issue, where I have a records with multiple numbers and text fields, and considering fine-tuning a free / open source model. Could you please let me know how this may be done?
). Either way, let me know if I missed the point of your problem or if it helped you think it through.