AsterixDB Support of Similarity Queries

Table of Contents

• Motivation
• Data Types and Similarity Functions
• Similarity Selection Queries
• Similarity Join Queries
• Using Indexes to Support Similarity Queries

Motivation [Back to TOC]

Similarity queries are widely used in applications where users need to find objects that satisfy a similarity predicate, while exact matching is not sufficient. These queries are especially important for social and Web applications, where errors, abbreviations, and inconsistencies are common. As an example, we may want to find all the movies starring Schwarzenegger, while we don’t know the exact spelling of his last name (despite his popularity in both the movie industry and politics :-)). As another example, we want to find all the Facebook users who have similar friends. To meet this type of needs, AsterixDB supports similarity queries using efficient indexes and algorithms.

Data Types and Similarity Functions [Back to TOC]

AsterixDB supports edit distance (on strings) and Jaccard (on sets). For instance, in our TinySocial example, the friend-ids of a Facebook user forms a set of friends, and we can define a similarity between the sets of friends of two users. We can also convert a string to a set of grams of a length “n” (called “n-grams”) and define the Jaccard similarity between the two gram sets of the two strings. Formally, the “n-grams” of a string are its substrings of length “n”. For instance, the 3-grams of the string schwarzenegger are sch, chw, hwa, …, ger.

AsterixDB provides tokenization functions to convert strings to sets, and the similarity functions.

Similarity Selection Queries [Back to TOC]

The following query asks for all the Facebook users whose name is similar to Suzanna Tilson, i.e., their edit distance is at most 2.

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $ed := edit-distance($user.name, "Suzanna Tilson")
    where $ed <= 2
    return $user

The following query asks for all the Facebook users whose set of friend ids is similar to [1,5,9,10], i.e., their Jaccard similarity is at least 0.6.

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $sim := similarity-jaccard($user.friend-ids, [1,5,9,10])
    where $sim >= 0.6f
    return $user

AsterixDB allows a user to use a similarity operator ~= to express a condition by defining the similarity function and threshold using “set” statements earlier. For instance, the above query can be equivalently written as:

    use dataverse TinySocial;

    set simfunction "jaccard";
    set simthreshold "0.6f";

    for $user in dataset('FacebookUsers')
    where $user.friend-ids ~= [1,5,9,10]
    return $user

In this query, we first declare Jaccard as the similarity function using simfunction and then specify the threshold 0.6f using simthreshold.

Similarity Join Queries [Back to TOC]

AsterixDB supports fuzzy joins between two sets. The following query finds, for each Facebook user, all Twitter users with names similar to their name based on the edit distance.

    use dataverse TinySocial;

    set simfunction "edit-distance";
    set simthreshold "3";

    for $fbu in dataset FacebookUsers
    return {
        "id": $fbu.id,
        "name": $fbu.name,
        "similar-users": for $t in dataset TweetMessages
                            let $tu := $t.user
                            where $tu.name ~= $fbu.name
                            return {
                            "twitter-screenname": $tu.screen-name,
                            "twitter-name": $tu.name
                            }
    };

Using Indexes to Support Similarity Queries [Back to TOC]

AsterixDB uses two types of indexes to support similarity queries, namely “ngram index” and “keyword index”.

    use dataverse TinySocial;

    create index fbUserIdx on FacebookUsers(name) type ngram(3);

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $ed := edit-distance($user.name, "Suzanna Tilson")
    where $ed <= 2
    return $user

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $ed := edit-distance-check($user.name, "Suzanna Tilson", 2)
    where $ed[0]
    return $ed[1]

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $sim := similarity-jaccard($user.friend-ids, [1,5,9,10])
    where $sim >= 0.6f
    return $user

    use dataverse TinySocial;

    for $user in dataset('FacebookUsers')
    let $sim := similarity-jaccard-check($user.friend-ids, [1,5,9,10], 0.6f)
    where $sim[0]
    return $user

    use dataverse TinySocial;

    for $i in dataset('FacebookMessages')
    where contains($i.message, "phone")
    return {"mid": $i.message-id, "message": $i.message}

    use dataverse TinySocial;

    drop index FacebookMessages.fbMessageIdx if exists;
    create index fbMessageIdx on FacebookMessages(message) type keyword;

    for $o in dataset('FacebookMessages')
    let $jacc := similarity-jaccard-check(word-tokens($o.message), word-tokens("love like ccast"), 0.2f)
    where $jacc[0]
    return $o

    use dataverse TinySocial;

    create index fbUserIdx_fids on FacebookUsers(friend-ids) type keyword;

    for $c in dataset('FacebookUsers')
    let $jacc := similarity-jaccard-check($c.friend-ids, {{3,10}}, 0.5f)
    where $jacc[0]
    return $c