Veritone advanced search provides a set of techniques that can help you create more effective search queries. The use of grouping, excluding keywords with NOT or -, contextual searching, word aliasing, proximity searching, and searching for abbreviations are ways you can customize searches using a set of special instructions.
Grouping Search Query Elements
Grouped or “Nested” searches use parentheses to clarify relationships between terms and specify the order in which search terms and operators should be read. Keywords inside parentheses are required to be read first, followed by information outside parentheses. If there are multiple nested parentheses, the innermost parenthetical expression is processed first, then the next, and so on until the entire query has been interpreted.
Example: ((Kobe OR Bryant) AND Lakers) OR basketball
Create a Group
Veritone makes it easy to select and create custom groupings of items in search queries containing two or more elements.
1. Press and hold SHIFT and click the first query element to be added to the group. The selected query item turns dark gray.
2. Press and hold SHIFT and click a query item to the immediate left or right of the first selected item. The selected query item turns dark gray. Continue selecting all query items to be added to the group.
3. Click on any of the selected items and select Group Selection from the drop-down list.
The selected items change to light blue and are enclosed in a blue dotted line.
Ungroup a Set of Query Elements
1. Press and hold SHIFT and click each element in the group. The selected query items turn dark gray.
2. Click on any element in the group and select Ungroup Selection from the drop-down list. The blue dotted line no longer encloses the group.
Exclude Keywords With NOT or -
Using the NOT and - operators in a Keyword search lets you easily exclude information so results are more relevant. To exclude a particular term, type _NOT_ or add the - character immediately before it. Exclusion searches must include at least one term that is not preceded by the NOT or - operator to be successful.
There are multiple ways to effectively perform a Keyword search using exclusion operators. In the sample below, we demonstrate how to build the same search query using both operators.
Search for transcripts that contain the word “peace” but that do not include all three of the words “Lakers,” “Shaq,” and “heat.” Transcripts that contain the word “peace” and one or two of the excluded words will display in the results.
peace AND _NOT_(lakers shaq heat)
With the NOT operator, words to be excluded can be grouped together using parenthesis.
peace AND -lakers -shaq -heat
With the - operator, the excluded words must each be preceded by the - character instead of being grouped inside parentheses.
A Contextual search uses closely related keywords or phrases to help you find proper nouns and non-English words. For example, Mahershala Ali is an actor who won the Academy Award for Best Supporting Actor in the film “Moonlight.” Because the name “Mahershala Ali” is not recognized as a common word or phrase, a search for it will most likely return no results. Alternatively, a Contextual search for "Moonlight actor" or “Best Supporting Actor for Moonlight” (or other variation) finds multiple results with various spellings of the actor’s name.
- more shell Ali
- Marshall Ali
- martial Ali
- my Hershel Ali
Once you see how the name is transcribed, enter it as a Keyword and search to determine how often it is transcribed using the modified spelling. If a search returns multiple results, the term may serve as an effective Alias for the name. (See the next section on Aliasing for details.)
If you frequently use specific proper nouns and have identified the way they are transcribed in the system, Veritone can alias them for you so that they are recognized in search queries. To request aliasing for a proper noun, send an email to [email protected] with the proper noun and all aliases you’d like associated with it.
The search function WITHIN allows you to search for instances where one subject occurs in the proximity of another subject. This function is very similar to AND in that both words or phrases connected by this operator are found together. But WITHIN allows you to control how close these words or phrases occur in proximity to each other, while AND simply ensures that both are found within a 5-minute mention.
For example, search for “Peyton Manning” AND “MVP” returns the results shown below:
To find mentions where “Peyton Manning” and “MVP” are spoken within 10 words of each other, enter your search as “Peyton Manning” w/10 MVP.
Searching for Abbreviations
When searching for abbreviations, it's recommended to create your Keyword query with the following four variations, separated by commas:
- All capital letters, no periods, no spaces, no quotes (e.g., USA)
- All capital letters with periods after each letter, no spaces, no quotes (e.g., U.S.A.)
- All capital letters with spaces and quotes (e.g., “U S A”)
- All capital letters with periods, spaces, and quotes (e.g., “U. S. A.”)
USA, U.S.A., “U S A”, “U. S. A.”
When searching for abbreviations that contain “&”, enclose the entire abbreviation in quotes (e.g., “AT&T”).