Widgets are where the user can dive into their data. There are three types of widgets:
- Document list
Notes are places for the user to take notes and keep track of findings in their other widgets.
Getting the most out of notes
Use markdown language for formatting in notes.
A document list is a searchable widget that displays the documents from a data set the dashboard was built from.
Charts are the widgets that will help the user visualize and gain insights into their data.
Charts can display various types of content:
- document count
- sentiment score
- sentiment polarity
- volume and sentiment
- taxonomy maps
- word clouds
Depending on the chart content, various information can be displayed. This information can comprise built in analytical items — listed below — or any custom fields the user has uploaded.
No matter the software, it may feel confusing or overwhelming when first trying to visualize data analysis output. Within Semantria Storage & Visualization there are several graphs and charts that may help the user get started in the right direction.
Wordclouds help the user get a high-level view frequency and sentiment. Advanced users may try filtering wordclouds to see how topics or themes are discussed within a data set. Consistent with all graphs and charts in Semantria Storage & Visualization, the greener the word in a wordcloud the more positive its sentiment polarity. Sentiment expressed in wordclouds represents the polarity of an analytical item (e.g. a topic, theme, phrase, or entity) and not the average sentiment for an entire document.
Sentiment polarity allows the user to see an ordered breakdown of sentiment and volume for analytical items, like themes, entities, and topics.
In this example, the topic "Staff-Attitude" is mentioned the most, with the majority of these mentions landing at the very top of the sentiment score range, "Very Positive."
Most widgets offer more information than is seen at first glance.
When the user hovers their mouse over a point of interest in a chart — be that a word in a wordcloud or a point on a line graph — they will be presented with more information.
The same goes for clicking on charts. If the user left-clicks on a point of interest in a chart they will be given the option to drill into the the documents supporting that item. This allows for a finer grained interaction with the data set. The user may also use this functionality to hide an item from the chart.
The user can also generate a pie chart to show the relative contributions by different analytics items to an overall dataset. Pie charts will also count non-sentiment bearing metadata.
We encourage users to get creative with how they investigate their data!
Click the icon at the top of the screen to access the widget creation wizard.
Notice the four main elements within the widget wizard:
- Data source allows the user to select which data source they'd like to power the widget (e.g. Customer reviews - Quarter 1 or Customer reviews - Quarter 2)
- Items to display allows the user to select whether they'd like Semantria Storage & Visualization to display the top n items or a custom set of items
- Widget is the menu from which the user may select their desired widget
- A greyed out widget occurs when the underlying data or configuration doesn't support it; in this case, the configuration does not contain a taxonomy, therefore a taxonomy map cannot be populated
Most widgets require only a single click to generate. However there are a few widgets which require an extra step, such as the sentiment vs. volume over time chart. This is because the user must select what item or items they wish to track.
In most situations the user will be able to customize a widget after they create it. Customization (e.g. the number of items displayed, specific sentiment filtering, and item exclusion) are all handled after the widget is created.
After the widget is created it can be customized using the icons at the top of the widget. These are explained below.
Clone, download, or delete a widget, or display data as a table
Increase or decrease the number of items displayed
Set the analytic items to display (e.g. themes, entities, phrases, etc.)
Set the chart type
Set the sentiment values (of the analytic item displayed) to display (e.g. very negative and somewhat negative).*
Reveal previously hidden analytic items
Filter the data
* The sentiment displayed on some widgets is the average of all items showing. If you show positive and negative items and exclude neutral items you might still see "neutral" sentiment as this is the average of positive and negative items.
Filtering allows the user to segment insights and find relationships within a large dataset. To open the filter panel, click the icon. It's located on any chart or document list widget tile.
Filters can be used to:
- Segment out sentiment (show only very negative and somewhat negative sentiment)
- Reveal relationships and sentiment (show entities related to the topic "attitude" with somewhat negative and very negative sentiment)
- Reveal relationships and sentiment (show entities related to the topic "attitude" with somewhat negative and very negative sentiment)
- Drill into metadata (show very negative sentiment and very positive sentiment related to the topic "cleanliness" for location: Best_Western_Plus_Inn_of_Sedona)
The user may add filters to any of three buckets: All of These, Any of These, and None of These. Inherent to each of these buckets is a Boolean operator. They are as follows:
All of These
Tim Cook AND Apple
Any of These
very positive sentiment OR somewhat negative sentiment
None of These
York NOT "New York"
Semantria Storage & Visualization scores sentiment for all the analytical items in a document. Then, it averages these scores and produces a sentiment score for the whole document. Therefore a document with negative sentiment may contain a positive topic. Similarly, a topic with negative sentiment may be part of a document with positive sentiment.
The Explore option now available in most widgets. Exploring is useful for seeing things that collocate with the item being explored.
Clicking Explore under an item does two things:
- It adds the item being explored to the widget filter under "All of these"
- It hides the explored item from that widget
The item that is being explored is hidden because by adding that item to the filter list, the item will occur in every document that is now shown in the widget. If the item is not hidden it will dominate whatever widget being explored, as every document in the filtered document set will contain the item. By automatically hiding the item we eliminate the excess data for the user.
Remove the item from the filter panel and unhide the item.
When the x axis of any graph is time, the user may click and drag on a section to zoom in on a specific period. The user may zoom in as many times as needed.
Clicking “Reset zoom” will take the user back to the starting point of the chart. If the user has saved the dashboard while zooming, the “Reset zoom” button disappears, and the new zoom is the default filter on the widget.
Widgets can be resized to fit up to 12 columns. Some widgets will have a minimum number of columns they can fill and still display data. To resize a widget, first, click Edit Mode on the top right of the dashboard, then click and drag on the edge of a widget to resize it.
Changing the number of analytical items displayed on a chart or graph is simple. The user clicks on the encircled number near the other widget options. This number represents the current quantity of analytical items being presented (e.g. top 5, top 10, top 15, etc.) Set the marker to the desired point on the slider and the chart or graph will automatically update.
The user can quickly cycle through analytical items within a widget by clicking on the analytical item icon and selecting a new analytical item from the dropdown menu. The graph will reload to display information relevant to the selected analytical item.
When an analytical item doesn't support a graph
Upon opening the analytical item menu, the user may notice that some analytical items are greyed out. This is because some analytical items cannot be represented in the selected chart.
For example, if a widget contains a sentiment polarity chart, then the user will not be able to select meta data (e.g. star rating) for the underlying analytical item. This is because meta data is not scored for sentiment. Therefore, all analytical items that are meta data will be greyed on on the dropdown menu.
Cloning filters or entire widgets helps the user structure and compare datasets. The user may want to take the filter used on one widget and apply it to another widget for the purposes of comparison. To do this, simply select "Apply + Clone" from the filter panel. Then, select all relevant widgets and apply the filter clone.
The user can also clone entire widgets by clicking the in the widget tile and then selecting the clone widget option.
Quickly compare data sources
Cloning is especially useful when a dashboard contains multiple data sources. The user can quickly clone a widget and select a data source from the data source dropdown menu next to the widget name. Cloning widgets allows the user to draw comparisons across data sources without needing to manually rebuild graphs and charts.
There are many ways to visualize data in Semantria Storage & Visualization. Text data is "deep," this means complex conversation can be visualized in myriad ways. Let's take a look at a few more useful charts and graphs available to the user.
Measuring sentiment score over time is a great way to add a new dimension to customer satisfaction metrics, like NPS scores. In this widget, the user is able to see how employees feel about key topics over time.
Taxonomy maps are a great way to tell a story. The user can drill into any node of a taxonomy map and develop a more granular view. For example, the node labeled "Delivery" is negative. However, drilling down reveals that consumers have a positive experience with the checkout process and packaging, despite a serious issue with shipping delays.
The volume & sentiment graph represents the document count as a line and tracks the sentiment polarity as an area chart. In the above example the sentiment and volume for the topic "attitude" changes drastically over five years.
It's useful to compare average sentiment score over time for analytical items. In this example data set of Disney Park reviews, the user is able to identify theme trends over time.
The user may need to count all the occurrences of an analytical item. This is easy to do with bar or column graphs. Now the user can get a clear idea of how prevalent a theme, category, topic, or entity is within a dataset.
The Crosstab widget allows pivot-line presentation of results. This widget allows for the display of collocations of multiple features. The widget can also be displayed as a wheel with scaled relationship ties.
The above image shows themes compared to topics, which allows users to see the frequency of themes occurring with a particular topic hit. One possible use could be comparing the entity type company to themes to see what themes are commonly occurring with mentions of certain companies.
Each cell contains the number of documents that hit on the column item and row item. In the above example, topics are being compared with themes, so each number represents the documents that hit on the topic in a given row and the phrase in the column.
The coloring is determined by the deviation from 0, and can be disabled under chart options.
Currently users can compare topics, entities (all or by type), phrases, themes or categories. Custom fields and metadata fields are coming soon.
The Conversation Explorer widget is an exploratory tool used to investigate what is being talked about in your content and how that item relates to other items in your content.
For a more in-depth explanation of this widget and its settings please see the dedicated Conversation Explorer page.
Users now have the ability to choose what columns are displayed in the document list widget. This allows for dashboards to be tailored to show exactly what information is pertinent to the end user.
The document widget has a new icon in the widget header as seen in the image below.
Clicking on the Column Chooser icon brings up the display view as seen in the below image.
In the Display view there are two sections, Available in data source and Not available in data source.
If all the data sources in a project have the same metadata mappings then there will not be anything in the latter section, regardless of data source. If data sources within a project have different mappings then depending on the data being displayed, different items will be in each section.
For example, in the above image the data source that is currently being displayed only has ID, Sentiment and Text. Other data sources (collections) have additional metadata, which is why they are listed in the Not available in data source section.
If a data source does not have a field that is being displayed then the column header will have a strike through the name. This can be seen in the above images in the Source column.
Changes are made to document sentiment are document specific and will not percolate across other documents, projects or configurations.
Editing document sentiment is useful for when a document contains nuanced sentiment that a reader can identify but NLP may have missed.
To edit the sentiment of a document hover over the document sentiment in the document view. This will bring up the edit icon.
Clicking on the edit icon will bring up the Edit Document Sentiment view on the right-hand side of the document view. Here users can set the sentiment of the document. Just as with the entity editor changing the Sentiment Polarity will place the Sentiment Score in the middle of that polarity range.
In the document view users can now add and edit entities.
This is useful in cases like where an entity is misspelled or a new entity is located by a reader in a document.
To add an annotated entity to a document click on the Add new Entity plus icon on the right side of the Entities header.
Adding or editing entities in SSV will not apply any changes to configurations.
Adding and editing entities is per document only.
Manually adding an entity in a document will not result in that entity being automatically located in any other document.
Adding an entity will bring up the new entity view as seen below.
In this view users type the entity under Entity, then select the Type of the entity from the dropdown list.
Selecting one of the Sentiment Polarity levels will change the Sentiment Score to the middle score for that polarity.
For example, under the default values of feature sentiment, a negative polarity would be set to -0.5, neutral to 0, and positive would be assigned a 0.6 sentiment score.
The Comments section can be used to explain why a user is adding an entity. Please note that comments are not currently exposed once an entity has been added. Future functionality will expose these comments to users.
Below an example of an added entity can be seen.
In addition to being able to add entities to a document, we have added the ability to edit existing entities. Editing found entities is useful for correcting sentiment, misspellings or even deleting misidentified entities.
Editing an existing entity is as easy as clicking the edit icon that appears when hovering over an entity.
Edited entities are denoted by a *
Clicking on the edit button will pull up an edit view much like the add view.
In the edit view users can change Entity, Type, Sentiment Polarity and Sentiment Score for the entity.
Making any changes to the entity will bring up a comment section for a user to document why changes were made.
An option not found in the add entity view is the option to delete the entity. Clicking on delete will ask for confirmation of the delete and allow the user to enter a comment about why the entity is being deleted.
Updated 4 months ago