Metadata Query Extensions

What are metadata query extensions?

With Salience 6, topic queries can now search on metadata extracted by Salience. Metadata available for querying is: presence of a particular type of entity, the sentiment associated with an entity or document.

For example this query:

{entity company} WITH “acquire OR merge OR purchase OR buy” WITH {entity company}

would match any text containing two company entities (whether discovered by Salience or defined by you) in the same sentence as any of acquire, merge, purchase or buy. For example:

“Today Lexalytics announced its intention to acquire Semantria.”

Similarly, this query:

{entity company:sentiment>0.2} NEAR/0 Lexalytics

matches documents where the sentiment for the company Lexalytics is greater than 0.2.

As of Salience 7, you can also pass arbitrary sections along with your document and include it in your query logic, e.g.:

{section (NPS < 5 OR net_promoter_score < 5) AND ("service" in What_Can_We_Improve_Question)}

Why would I want to use metadata queries?

Faster event detection and response
Querying on types of things allows you to look for events fitting a pattern, such as a company announcing new products, or what places a natural disaster is affecting. This allows you to quickly identify and respond to these events.

Faster routing of customer complaints
Querying on sentiment allows you to automatically route negative comments for a particular area directly to the team handling those issues. While this can be performed after Salience processing as well, this allows you to split large volumes of comments at processing time, speeding up the process.

How do I use metadata queries and what results do I get?

Salience 6 adds a new structure to the topic query syntax (see Query Grammar page for details). Your existing queries will continue to work unchanged.

If a document hits a query with a metadata structure, the result will include the matching text. The SalienceTopic structure has been expanded to return which entities matched an entity query.

Are there performance implications?

Metadata queries impose no more overhead than any other query term – querying for any entity of type company is no different than querying for an individual company name.

Can I use Metadata queries with subqueries?

Yes, a subquery is just another operator in the query, so a query such as:
{entity company} AND @BadNews
is perfectly legal.