The advanced search provides a clear and structured way to build complex queries including logical operators and filters for metadata elements such as source, sentiment, location, etc. You can access this dialog in two different ways – through the header menu “Advanced Filter” to inspect or modify the currently active global filter, or by clicking on the gear icon to edit a bookmark definition.
Main Search Dialog
As shown in the screenshot below, the plus and minus symbols adjust the number of restrictions, disregarding empty fields. The system supports three logical operators: ‘and’, ‘or’, ‘and not’, including the option to use brackets to combine these operators as part of nested constructs. Please note that ‘and’ combinations will take precedence over ‘or’, which can be changed by applying brackets.
List of Search Attributes
The logical operators of the advanced search can be applied to a range of content features and metadata elements:
- Text / Title / Text or Title. Specifies in which parts of a document to search. The full text search can also be restricted to matches which have to co-occur within the same sentence. The topic management section contains additional information on the various options to define a query.
- Country of Publication. Contains a drop down selector for restricting the search to documents published in certain countries.
- Target Location. Search for locations mentioned in documents, which can include the names of cities, states and countries.
- Document Keywords. Documents with specific keyword annotations (this feature differs from the full text search, as only up to ten keywords are being annotated for any given document).
- Named Entities. Documents that mention specific people (e.g. “Elon Musk”), organizations (e.g. “Tesla, Inc.”) or locations (e.g., “San Carlos, US”).
- Source Identifier. Restricts the search to certain social media accounts or Web domains, using platform-specific prefixes such as twitter:cnn, http:cnn.com, etc. (note the difference in spelling of a source identifier such as http:ibm.com and a URL http://www.ibm.com).
- Uniform Resource Locator (URL). Restricts the results set to documents that include the specified pattern in their URL. We recommend to use the “Source Identifier” (see above) and not the URL pattern when targeting entire domains such as bbc.co.uk.
- Sentiment. Documents classified as either “positive”, “negative” or “neutral”.
- Reach. Restricts the search to sources with a certain level of authority, normalized in a zero to one interval (measured in terms of average traffic statistics in the case of Websites, for example, or a metric ratio of followers and followed accounts in the case of Twitter).
- Length of Text. Document of a certain length in terms of word count, character count or number of sentences contained.
- Content Source / Content Language. Narrows the search to one of the available data sources (news, Twitter, etc.) or specific languages. These settings are typically done via the header menu, but can be useful for nested queries with brackets – e.g., including all news articles but requiring a minimum reach value for Twitter postings.
Features to Support the Editing Process
List Editor. The options “list of phrases” for text elements and “is (list)” for other attributes provide access to a multi-line editor that facilitates the definition of queries that require a list of attributes is required, for example a set of ten source identifiers.
Autocomplete. The search for some of the fields, for example known entities, provides an autocomplete function. Entering text in a search bar triggers a drop-down menu with matching suggestions retrieved from the entire document repository. This feedback is useful to check whether a match exists before running the actual query.