How to Search Spoken Audio by Phrase or Mention

Let's Get Started

FluidDATA is an audio search engine that allows you to quickly find words and phrases in millions of publicly available audio files on the internet. In this post we will cover how FluidDATA can be used to quickly discover content, research topics, look for that half remembered podcast episode, or use it has a language learning resource.

Introducing...

FluidDATA dataset

As of April 2018, FluidDATA has indexed over 6 million audio files from over 150,000 RSS feeds. This is over 500 years worth of audio and our collection is rapidly expanding every day.

If you notice that we are missing a podcast feed feel free to shoot us an email with the feed url at support@bitplatter.com and we'll add it to the FluidDATA collection.

Starting a Search

You can begin your search from the FluidDATA landing page.

landing

Enter a search query in the search box to start a new search. FluidDATA will search our database for the specific word or phrase mentioned in an audio file.

By default, FluidDATA will search for the exact word or phrase as it was spoken. For example, the query San Francisco New York City will match someone saying the exact phrase San Francisco New York City. In this example case you are probably interested in audio where both San Francisco and New York City were mentioned. You will need to use advanced search syntax to do this. In this case the query would be "San Francisco" "New York City". You can read more about advanced searches in the Advanced Searches section.

The search is not case sensitive so New York City and new york city will return the same results.

The search automatically performs word stemming to provide additional relevant results in your query.

A search that results in 0 matches will not count against your monthly usage quota.

Once you enter a search phrase you will be taken to the results page.

Audio Results

The audio results are displayed in results cards in the middle of the results page. Each result card gives pertinent information about each result including the title, author, and the time since it was published.

The most important part of the result card is the player. The audio timeline in the player shows where each search phrase was mentioned in the audio. You can either hit the orange skip button or click the timeline to jump to the location in the audio where the phrase was spoken.

Each result card also shows a snippet of the transcribed audio with the search match highlighted. As the audio plays this snippet will be updated to show each match.

player

Clicking on the artwork in the audio result will take you to the Collections view where you can search just the audio created by that author.

Search Tabs

When a new search is performed the results will be placed in a new search tab near the top of the page. You can use these tabs to quickly navigate between previous searches.

searchtabs

Collections

Everytime you perform a search FluidDATA will also search for audio collections. This provides an additional method for finding specific content, especially if you are searching for content from a specific source or channel. The collection search results will be displayed on the right hand side of the page. Click on one of these results to be taken to the channel search page.

Collection Search

When you click on a Collection in either the right-pane or on the artwork in the audio search result you'll be taken to the collection search page which will allow you to limit your searches to only audio files within that collection. Simply enter your search in the search box and hit enter.

Home

You can access this page by clicking 'Home' in the user panel. This page provides a quick way to access Recommending and Popular collections as well as the most mentioned brands and people for the week. home

Mentions

People

FluidDATA can track mentions of celebrities, politicians, athletes, and other well known personalities. This provides insight into who is being actively discussed across thousands of sources.

Brands

In addition to tracking popular and trending people, FluidDATA can also track popular brands and companies. This allows you to quickly discover what brands and companies are being discussed right now.

Clicking on one of the People or Brands shown in the Home page will take you to the audio results where they were mentioned in the past week. This functionality is expanded even further with FluidDATA Alerts which allows you to get email notifications when certain words or phrases are mentioned. See FluidDATA Alerts below for more information.

Saved Searches

We've found that users are searching for the same things over and over so we've added the ability to save searches so you can quickly get started. Simply click the + button and you'll be able to add up to 5 saved searches.

FluidDATA Alerts

FluidDATA allows you to create 5 FluidDATA Alerts that can send email alerts when a word or phrase is mentioned and lets you view charts that display how these mentions are trending over time. You can add new phrases by clicking the plus icon and remove existing entries by hitting the X icon (displays on hover). This feature requires that you create a Free BitPlatter account.

Alert View

Once you have added an Alert you can select it in the user panel and get insight into how the Alert has been trending. This Alert View shows a Trends table and a Mentions Over Time chart.

The Trends table shows the increase or decrease of the Alert for several date ranges. Clicking on one of these will show all the audio that matched the alert for that date rage.

The Mentions Over Time chart shows the percentage of audio that matches the Alert in our audio collection every month. We can see in the chart below that Facebook was mentioned in about 40% of all our audio in September of 2016.

alerts

All Alerts View

Want to compare how well different phrases, products, or brands are trending? We've got you covered. You can click 'View All' under Alerts to view the trending data for all your alerts.

For instance, in this example we can see that Facebook is mentioned much more than Google and Apple. alerts

Alerts Email Notification

When creating Alerts, you can optionally receive Daily or Weekly notifications of how the term is trending on FluidDATA. You'll receive an email report for all of your alerts that will look like this:

Advanced Search

By default, FluidDATA searches for the exact phrase entered in the query box, but more advanced queries are possible.

Double quotes " are used to identify a single phrase. You can chain these single phrases together with AND +, OR |, or NOT - operators. See the table below for some example searches using the advanced syntax.

Example Query Explanation
"machine learning" +"neural network" This will produce audio results where both machine learning and neural network is mentioned.
"machine learning" -"neural network" This will produce audio results where machine learning was mentioned but neural network was not mentioned.
"machine learning" | "neural network" This will produce audio results where machine learning or neural network was mentioned.

Note that the default operator between quoted phrases is AND and can sometimes be omitted. For example "machine learning" +"neural network" and "machine learning" " neural network" are identical.

Let's get moving

FluidDATA is a great way to search and navigate through massive audio datasets. We are constantly working on new features and improvements to allow users to garner new insight from audio and video across the web, or with their own datasets. It is the perfect tool for eDiscovery, educational research, and podcast listeners and creators. You can start searching and create an account at FluidDATA or checkout the recently released FluidDATA API.

You can also checkout an example of the FluidDATA API on GitHub by clicking here.

Visit our homepage at bitplatter.com or contact us as at support@bitplatter.com if you are interested in integrating FluidDATA into your own workflow.