Below is a screenshot of the search interface.

Plain Text and Binary Search Queries

The search interface takes either a regular plain text entry or enables you to use binary search parameters. The screenshot above shows a binary search parameter. In this case, use "&" and "|" to tell the platform whether the terms you are seeking should be shown together ("&" for yes, and "|" for no). For example, the case above will prioritze search results that have the term "ipo" or the term "financial result".

Putting ipo & (financial & result) would prioritize statements where both terms appear.

Note that the search results are ranked by relevance, so you will see results that have a mix of the terms based on relevance to your search.

Specifying Dates

There are two ways to specify dates. The simplest is putting in a start and end date for your query. In this case, you will search through the knowledge graph for the specified periods. If the end date is blank, then tomorrow's date will be used. A blank start date will use a date that is 50 days earlier than the end date.

If you plan on rerunning your queries, another approach is to set the days from end date option. This will always set the start date and end date range to whatever number of days you set prior to the end date. If you leave the end date blank, this means the start and end date range will be the period of "x" days you set from today, up until tomorrow.

Using Probability Scores

Click Advanced Options to toggle the probability scores. These scores allow you to limit queries to topics that we've built models for. These topics are...

• COVID-19: a probability score that the statement is about COVID-19.
• ESG: a probability score that the statement is related to climate change, environmental issues, or social and governance issues.
• Funding: a probability score that the statement is related to a funding or financial announcement.
• Sentiment: the sentiment of a statement, with -100 to -11 being negative, -10 to 10 being neutral, and 10 to 100 being positive. Note that this looks at the positivity of the words themselves, rather than their sentiment towards a specific topic, as such a task requires bespoke models.

The Entity Analysis option uses the same query approach as the Dashboard interface. The difference is the level of analysis completed, which causes the query to be signficantly slower.

The graphs for this part of the site are described below:

• Proportion of sentences that contain the term or search query by day. This includes additional filters, like probability scores.
• Average sentiment of statements meeting search criteria, by day.
• Number of statements by sentiment type.
• A table of up to 150 statements that meet your search criteria.
• A list of entities that have been extracted from the sentences themselves. These are typically proper nouns appearing in the statements. The brackets show the number of sentences those entities appear in.

You can speed up queries a number of ways:

• Limit the time frame for your search. The fewer days being analyzed, the faster the query.
• Add probability filters for the search.
• Limit your search to titles.
• Turn Plain Query Format off and use the binary search approach.