Signals User Guide¶

Create an account¶

Begin by Creating an Account. Once you’ve logged in, check out some of the pre-built public dashboards in the folders on the homepage.

Create your first dashboard¶

To create your own dashboard, click “Create a new Dashboard”.

Connect to data¶

Connect to your database, upload a CSV or Excel file, or use one of our Data Connectors for easy access to a wealth of publicly available data. Add as many data source to your dashboard as needed.

If you’re bringing your own data, specify which fields contain Textual, Temporal, or Geographical information for analysis and normalization. All other fields will be brought in as pivot points and additional visualizations.

Note

See the Mapping Data guide for more details when uploading CSV or Excel Files

Create visualizations¶

Your dashboard will contain visualizations based on the type of data in your analysis by default. See the guide on Processing Data for information on how to interpret these visualizations. Every visualization is clickable, and clicking will apply a filter on your data based on your selection.

Now you’re ready to Create Your Own Widgets using your other datapoints to home in on interesting segments within your data and discover topics and themes within your textual data.

Share your results¶

Make your dashboard public or share with colleagues.

Enterprise Data Connectors¶

The Enterprise Data Connector application allows users to connect with a variety of internal data sources.

Examples include:

• Hive
• mySQL
• Microsoft SQL Server
• PostgreSQL
• Oracle SQL

3rd Party Data Connectors¶

3rd Party Data Connectors usually have a pop-up window that allows you to enter your credentials with the 3rd party in order to authenticate with Signals.

• salesforce
• Foresee
• Surveymonkey
• Gmail
• Livechat
• Intercom
• JIRA
• UserVoice
• Zendesk
• Google Analytics
• Trello

salesforce¶

Foresee¶

Surveymonkey¶

Gmail¶

Gmail allows you to connect with a Gmail account and enter a query that will pull data matching the query.

Livechat¶

The LiveChat connector allows you to select a date range and the type of user you want to analyze chats for: Agent, Visitor, or both.

JIRA¶

With the JIRA Data Connector, you can select which projects you want to analyze. You can also specify a ticket type if desired. If nothing is specified, Signals will pull all of your JIRA data.

UserVoice¶

The UserVoice data connector requires a login and api key to connect to your uservoice data. The permissions associated with the login correspond with what data will be available through the connector.

Users can specify which objects you want to analyze.

Public Data Connectors¶

• iOS Store Reviews
• Google Play Store Reviews
• Home Depot Product Reviews
• Lowes Product Reviews
• Best Buy Product Reviews
• Wal Mart Product Reviews
• Etsy Shop Reviews
• Etsy Product Reviews
• Amazon Product Reviews

• Twitter (with twitter account credentials)
• Facebook (with facebook account credentials)
• Weibo search
• Youku
• YouTube Comments

• Consumer Financial Protection Bureau
• Indeed
• Consumer Affairs

iOS Store Reviews¶

Home Depot Product Reviews¶

Lowes Product Reviews¶

Best Buy Product Reviews¶

Wal Mart Product Reviews¶

Amazon Product Reviews¶

Twitter¶

Facebook¶

YouTube Comments¶

Consumer Financial Protection Bureau¶

Indeed¶

Consumer Affairs¶

Signals SDK¶

The Signals SDK should be used to upload files more than 500 MB.

It can also be used to develop custom connections and/or schedule uploads.

We offer a node.js wrapper for our SDK. https://www.npmjs.com/package/signals-api

To generate an API key:

1. go to Settings in the left-hand menu from the Signals homepage.
2. In the Settings page, click the button to generate an api key towards the bottom.
3. This API key will be generated and downloaded in JSON format to be included in your application.

After the API key has been generated, you can always download it again from the Settings page.

The “Revoke” button will immediately invalidate the API key from further use.

Textual¶

The textual data is the data that you want to run unsupervised machine learning to generate buzzwords, a topic model, and sentiment analysis upon. The next section, on Processing Data, discusses the outputs of this process.

Temporal¶

Identifying where you temporal data is within your dataset allows the Signals Analytics Engine to normalize dates, allowing you the flexibility to roll-up data on different granularities like yearly, monthly, daily, hourly, etc.

Temporal data can be represented in many different formats:

• If your data has a field that contains Day, Month, Year, and/or time in any order, you can simply select the “Date” option in the Map Fields menu.
• If your Day, Month and Year are all contained in separate columns, use the corresponding Map Fields menu items for each to indicate this.
• If your date is represented in some other format, select “Date” and the Signals Analytics Engine will do a best-guess to normalize the data.

Geographical¶

The Signals Analytics Engine will create a geographical hierarchy based on the input you provide, giving you access to compare your data at any level, whether it’s by country, region, or zip code.

If your data contains multiple geographical or spatial data points, you can indicate where that data resides using the Map Fields menu.

Some of the Geographical options are absolute and therefore don’t need additional information:

• For example if your data has a Lat/Long field, Long/Lat field, or one Latitude and one Longitude field, there is no need to specify country, city, or any other geographical information.
• Same goes for IP address and phone number.
• Street, City, State, and Country however, are not 100% by themselves, so map as much information as possible if your geographical data is not absolute.

Contributor¶

Contributor data is used to tie all the insights back to an individual to help organizations close the loop with their customers.

This field should represent the person or entity that generated the record or document in your data.

Signals Analytics Engine¶

The Signals Analytics Engine leverages Machine Learning and Deep Learning algorithms to help navigate and pivot a large set of textual data.

Built on top of our proprietary Bayesian Neural Network and Generative Model, Signals dynamically identifies semantic topic groups based on the context in your input data.

This is all done in a three-step process:

1. The engine starts by performing NLP in over 24 languages.

In this step, your input documents are tokenized into corresponding N-Grams (N>=2), lemmatized (words with the same root are grouped together e.g. run & ran), stemmed, spam/junk and stop words are filtered out, and part-of-speech tagging and named entity extraction are performed. A large N-Gram-based content network is then created based on your input data files.

2. The engine runs a Multi-Model approach on top of the N-Gram-based content network.

This includes using our proprietary text analytics algorithms extended from Bayesian Neural Network, Generative Model, LSTM (Long Short Term Memory), and Seq2Seq NLU. In this step, data input is clustered into semantically meaningful groups.

The groups are generated and visualized by statistical significance (i.e. the percentage attributed to each topic category in the Semantic Topic Visualization). Each topic is tagged with top representative terms in Buzzwords.

3. Signals automatically processes all geographical (Where), temporal (When), contributor (Who), as well as any other structured data.

It joins the data with the N-Gram-based content network for you to pivot and construct analytics questions against your dataset.

Reprocessing¶

There are a number of Advanced Options that can be applied to customize the output of the Analytics Engine. When applying advanced options your data will need to be reprocessed. This can either be done directly through Edit Mode, or through clicking the reprocess button on your dataset under Manage Data.

Dashboard Options¶

Once you’ve created a dashboard, you can Permission Levels, delete, or modify properties by clicking on dashboard menu:

Dashboard Properties¶

This is where you can add tags, an image, or even a video, as well as see who else has access.

Note

If a dashboard has been shared with you, you might see a dashboard menu that looks like this:

This means you do not have permission to share with others. For more on permissions, see Administration

Analytics Engine Output¶

The Signals Analytics Engine generates new datapoints based on your unstructured data. This information is represented in the following way.

Buzzwords¶

The Signals Analytics Engine reads through every piece of text in a dataset and compares every possible combination of words.

The Buzzwords visualization is made of every statistically significant N-Gram found in the textual data. In order for an N-Gram to be deemed statistically significant, we look at how often they occur together in the dataset vs. how often they occur individually.

A word-cloud view of N-Grams

N-Grams in a detail list visualization

Topics¶

Topics are generated by performing unsupervised machine learning on top of the N-Grams to determine hidden themes and group the documents accordingly.

Any document can occur in more than one topic, therefore the % of documents contained in each topic will add up to > 100%

The Semantic Topics visualization can be represented in a donut chart or a network graph.

The words around each slice are just the top one or two N-Grams included in that topic. Each topic is represented by a slice of the donut chart. The order (starting at 12:00 and moving clockwise) and size of the slice is determined by the statistical relevance or “tightness” of that topic. The color of the slice represents the sentiment.

Semantic Topics represented in a network graph of related N-Grams. Each bubble represents an N-Gram, the size indicating the count, and the color indicating the sentiment. The lines show the level of connection, or co-ocurrence, between two N-Grams.

Temporal Trends¶

Temporal information accompanying structured and unstructured data is paramount in understanding quantitative events and their potential underlying relationships across disparate data sets. Signals utilizes time-series predictive analysis, deep learning, and event analysis to uncover trends and patterns across structured and unstructured data.

Topic Model trended over time. Each smaller bar represents a topic while the bar groups represent a time period.

Contributors¶

Contributors help to identify an individual’s influence in a data set.

Name: a list of unique identifiers mapped as “name” Count: the # of feedback from an individual or unique ID. Sentiment: The aggregated sentiment score on all the documents contributed by a contributor.

Interactions and Filters¶

Every interaction in the dashboard essentially applies a filter on your data. This method is designed to help you access insights learned from your unstructured or structured data, and tie it to other dimensions.

When interacting, the filters you’ve applied will appear at the bottom of the page next to the name of the dataset the filter is applied to.

Clicking on the name of the filter will remove it.

Filters can also be saved to a dashboard and can be set on one of three levels:

1. Dashboard
2. Tab
3. Widget

Tab-Level Filters¶

Tab level filters can be applied on a per-tab basis. To do this, click the icon and select one of the two methods:

1. Merge with the current query will allow you to set a permanent filter, but still interact with the data. Interactions with other tabs can still affect the data.
2. Override the current query will apply the filter selections you make and freeze the visualizations so that the tab is no longer interactive. Selections on other tabs will not affect the data.

The main datapoints are available for selection from the Tab Detail page. Other datapoints can be accessed by clicking on the “+” icon next to Extra Structured Vis Parameters

Widget-Level Filters¶

Widget-level filters will override all other filters for that widget.

To apply a widget-level filter, you can either click on the funnel icon in the widget settings menu from your dashboard view, or you can apply a filter from the Create Your Own Widgets in the Options tab.

Widget Settings Menu

Stopwords¶

Signals provides an out-of-the-box list of stopwords that include the main commonly used non-informative words such as: “the”, “an”, “I”, etc.

You can create additional lists that can be applied as needed to cut through noise in your textual data on a per-data-source basis.

Typically a brand new data-source is run with the default stopword list, noisy signals can be easily identified in the dashboard and our in-dashboard editor (below) allows you to select N-Grams, topics, contributors, or other features to suppress.

For example, RSS news feeds typically contain the same few sentences at the end of every article:

Reporting By Laila Bassam in Aleppo and Tom Perry, John Davison and Lisa Barrington in Beirut;

Writing by Angus McDowall in Beirut, editing by Peter Millership

Appears at the end of every news article in some publications. “Reporting By” and “Writing By” will likely be identified as Buzzwords. These terms could potentially link unlrelated documents since they aren’t related to the article topics.

To suppress the noise caused by these terms, in the edit menu, click “Tune up data” and select the terms that are non-informative.

When finished, click submit and your data will begin reprocessing with the feedback you’ve provided.

Junk/Spam¶

Creating a Junk/Spam list can be useful in many different datasets

For example in Twitter data if there is any spam, it will quickly be identified through the visualization. Users can choose to filter out the spam and reprocess the data, indicating to the Analytics Engine that the matching documents should be ignored when processing (generating buzzwords and categories).

This is just one example, but as you can see it is generalizable to many other data-types.

Sentiment¶

Signals Provides an out-of-the-box sentiment package that is trained on a breadth of data sources however, you have the ability to customize sentiment to tailor results to your specific use case.

Certain terms are generally neutral, but when used in the context of a specific data-set or industry, always have a sentiment associatd with them. Or vise-versa, some terms are very negative generally, but are actually commonly used industry lingo among certain data-sets.

Taxonomy¶

Signals supports importing existing taxonomies or you can easily build your own from the ground up. The in-dashboard taxonomy editor allows you to tweak your taxonomy as you analyze your data. Drag and drop buzzwords into your label logic to increase your coverage of existing categories, or as you find new categories.

Deploying Signals¶

A Cloud Deployment is our most common method of deployment. Among other services, Amazon Web Services S3 and EC2 are used to provide an optimal experience in computing and visualization.

Contact us for information on alternative deployment methods.

Account Info¶

Your account info and management console is all available on the Settings page.

Permission Levels¶

Permissions on any asset (dashboard, stopword list, taxonomy, etc.) are set by the creator of that object.

Can View users can view and use the asset but cannot permanently modify it.

Can Edit can view and modify the asset, but cannot share it with others.

Can Share can edit and share the asset with others, but cannot remove others’ access.

Owner can edit, share, and add or remove users with less privilege.

User Groups¶

Group management can be done from the Groups page on the main nav bar.

There are 4 levels of access at a group level:

1. Group Admin
2. Admin
3. Can Edit
4. Can View

The creator of a group is automatically a Group Admin for the group.

Group Admin users:

• have a maximum* access level of Owner to any asset shared with the group.
• can add users to the group and give them Group Admin rights or lower to the group.
• can remove users with less permission (Admin, Edit, and View Only users)

Admin users:

• have a maximum* access level of Owner to any asset shared with the group.
• can add users to the group and give them Admin rights or lower to the group.
• can remove users with less permission (Edit and View Only users)

Can Edit users:

• have a maximum* access level of Can Edit to any asset shared with the group.
• cannot add or remove other users from the group.

Can View users:

• have a maximum* access level of Can View on any asset shared with the group.

*when sharing an asset with a group, the sharer dictates the level of permission granted to the group. The lesser privilege (between permission granted to the asset and permission within the group) will then take precedence for each user. For example if a user grants Admin permissions on a dashboard to a group, group members with Can View in the group will only receive Can View on that object.

United States¶

• 1431 W. Morehead Street Charlotte, NC 28208
• P: 704-215-4955
• webcontact@stratifyd.com

Silicon Valley¶

• 4500 Great America Pkwy Santa Clara, CA 95054
• webcontact@stratifyd.com

China¶

• 9th, Guanghua road Chaoyang District, Beijing SOHO 2
• P: 1-3466-390016
• paul@tasteanalytics.cn

London¶

• 24 Cornwall Rd Dorchester, Dorset DR1 1RX
• P: +44-(0)-7944-724920
• o.bayliss@stratifyd.com

Signals for Students¶

We are a group of data scientists and PhDs from universities, thus we support students expanding their analytics horizon by providing free access to advanced technologies.

Feel free to drop us a line with your EDU account at clientsupport@stratifyd.com. Our specialists will be more than happy to create an account for you.

If you are a lecturer or professor who wants to use Signals in your classroom, please contact us and we will be happy to arrange licenses for your students. Feel free to drop us a line at clientsupport@stratifyd.com.