facebook's Graph Search

Querying for relationships between objects

What is a Graph?

A graph is a way of representing connections between entities. In the picture here, each user is a node connected by edges to other nodes.

Facebook contains not just people, but things, such as businesses, musicians, politicians, etc. that can all be related to one another in various ways.

Triple Stores

The underlying principle behind graph search is the idea that a user is not searching just for entities and if entities are connected, but also how these entities are related

Graph Searching

facebook provides a few ways to query among nodes to discover things you are already connected to or new things you may be interested in

Under the Hood

Graph Search needed to be designed toward understanding the user intent precisely and serving structured objects

Query Interface

Initial issues with Keyword Queries

Keywords, in general, are good for matching objects in the graph but not for matching connections between the objects.

Solution: Natural Language Interface

Natural Language Interface: Components

Entity Recognition and Resolution

Lexical Analysis

Semantic Parsing

Entity Recognition and Resolution

“People who live in san francisco”

The segment “san francisco” is likely to be an entity of the {city} category

Lexical Analysis

Synonyms: "where to eat sushi” => Sushi restaurants
Fillers: "all my friends photos” => "My friends’ photos"
Related forms: used WordNet to extract related word forms
Inflections
- "photo" vs "photos"
- "people who work at facebook" vs "people who worked at facebook"

Grammar

[start] => [users] $1

[users] => my friend friends(me)

[users] => friends of [users] friends($1)

[users] => {user} $1

[start] => [photos] $1

[photos] => photos of [users] photos($1)

Structure: Weighted Context-Free Grammar (WCFG) to represent the Graph Search query language