Fuzzy Full Text Search

[jdgamble555 (Jonathan Gamble)]

Is it possible to do a fuzzy full text search now by using var queries?

For those who don’t know, right now you can do a full text search by using anyoftext or alloftext in DQL (@hash / @exact). You can also do a fuzzy search by using the match function (@trigram), however it is one word only.

To me the match function is not that useful since most searches will contain more than 1 word. We need a match function that matches phrases like the full-text search.

Question

1. Couldn’t we write a match function right now that matches all words by combining several match functions. Surely this has already been done? I do realize the words won’t necessarily be next to each other, but it is still the next best thing.
2. Could we possibly store all our text fields without spaces (maybe -) to simulate larger words? Some trigram indexes count the space in the index. Would this blow up our database too much?
3. Perhaps a mixture…
   a. Get all full-text results
   b. Get fuzzy results by matching results including all words
   c. Score full-text higher?

I realize we can’t score now by how many times the words are there, but we can score by closeness to the search phrase and not just search term…

This does not take speed into account, just thinking about functionality.

Just some thoughts.

J

Related Posts:

---

Feature request: full text search with tf-idf Scoring

MichelDiz commented : This feature does not currently exist. There is not much to do. But I have an idea that can help you. But you will have to use Facets. https://docs.dgraph.io/query-language/#sorting-using-facets The idea is to map the text by your application and then create a word structure. So you can track its usage. The only problem is that if you need the words to be unique in the system, you will need to manually add the number of repetitions. Inside the facets by the Map_text_body…

Partial full-text search

Hi guys, I’m doing a tutorial and cannot figure out how I can do a partial full-text search. Let’s say I would like to find “Michael” by “mic” or “Artyom” by “art”. Is it possible? Thanks!

How to implement keyword based relevance sorting

{ me(func: Type(Document)) @filter(allofterms(keywords, "All target Keywords here")) { count(uid) } } I was about to work on it, but reading further I noticed that this is really not possible. Dgraph can’t count the number of terms found in a predicate. It counts nodes/edges only. I think you should do something similar to Not saying that your problem is the same, just the example I did would be similar.