International Corpus of English

Using VOICE-Online

6 Filters

This section offers information on how the filter function can be accessed and on how filters can be selected, combined and disabled again.

The filter function allows 'filtering' the corpus, i.e. selecting specific corpus texts for searches. If the filter function is activated, i.e. one or more filters are selected, subsequent searches are limited to the corpus texts that have been selected, i.e. the searches are carried out only in these texts. If no filter is active (which is the default setting), searches are carried out on the entire corpus, i.e. all 151 corpus texts.

The two classes according to which texts can be selected via the filter function are domain and spet (speech event type).

6.5 Combining filters

The filter function also offers the option of combining several filters which select (or exclude) different filter categories. Filters can be combined via the operators in the Edit Corpus Filter pop-up which are explained in the following.

6.5.1 Adding filters

A new filter can be added by clicking on the plus-symbol in the bottom left-hand corner of the pop-up box. Another grey box with the caption drop filter here appears for which a new filter category can be selected (see description for selecting filters above).
Screenshot of pop-up with 2nd filter
Adding a filter

6.5.2 Filter relationships: and/or operators

Two filter categories can relate to each other in two ways, which are represented by the and/or operators in the pop-up window.
Screenshot of operators
The and/or operators
Operator Description Example selects
and Both filter categories (i.e. conditions) need to be fulfilled by a text in order for the text to be selected for the search. domain PO AND spet wgd all texts in the Professional Organizational domain (PO) which are also working group discussions (wgd).
or One of the filter categories (i.e. conditions) needs to be fulfilled by a text in order for the text to be selected for the search. domain ED OR domain LE all texts from the Educational domain (ED) and from the Leisure domain (LE).

In its default setting, the and relationship is activated whenever a new filter is added. The and is indicated in black letters, while the or is represented in grey letters. The relationship can be changed by clicking on or, which is then indicated in black letters. Filter combinations that yield no matching texts

Obviously, some filter combinations will result in no matching texts. For example, a text cannot fulfill both the condition domain ED AND the condition domain LE as each text in VOICE is only part of one domain. Similarly, not all speech event types are represented in all domains in VOICE. For example, while selecting the filter categories domain PO AND spet mtg produces matching texts, the selection domain PO AND spet sed produces no matching texts, as there are no seminar discussions in the Professional Organizational domain in VOICE.

Whenever a filter combination produces no matching texts, this is indicated in the application area with the caption no matching texts for this selection. The filter selection then needs to be changed in order to find matching texts and render search results.
Screenshot of non-existent combination of filters
Filter combination that yields no matching texts A note of caution: nonsense combinations

The filter function is designed to be very flexible and powerful and allows all possible combinations of filter categories and operators. However, this means that using inappropriate filter combinations may result in no matches, and that some 'nonsense' combinations of filters are also possible.

An example: The filter combination not domain PO OR domain ED selects all texts except those in the PO domain. The second filter category (domain ED) is superfluous as the texts from the ED domain are already selected through the not domain PO filter category.

The responsibility of checking whether the combined filter categories produce a meaningful selection of texts rests with the corpus user.

6.5.3 Introducing filter hierarchies

Combinations of different filter categories can also be put into a hierarchical order. A filter category can be shifted to a higher hierarchical level by clicking on the grey double arrow located at the right side of the grey filter box.
Screenshot of introducing a filter hierarchy
Introducing a new filter hierarchy
Introducing such a hierarchy of filters allows the two different filter relationships (i.e. and/or) to be combined at different levels.
An example:
Screenshot of example of filter hierarchies
Example of filter hierarchy
The filter combination in the example above selects all texts from the ED domain and then further specifies that the texts also need to be either conversations (con) or seminar discussions (sed). Or, in a more formulaic representation: domain ED AND (spet con OR spet sed).

Clicking on the plus-symbol adds a filter at the current hierarchical level. Clicking on the minus-symbol removes the filter or shifts it back to the main level of hierarchy.