A dialog act is a specific component of speech that labels a piece of dialog according to its category of meaning. The dialogue act attaches to many of the most common kinds of dialog, often on a line-by-line or statement-by-statement basis. Many researchers tag parts of dialog with dialog act labels in order to more accurately observe or model the dialogue.
Dialog act tags show what a speaker or writer is trying to convey. Some of these are fairly basic labels that express a more general function, where others may be more complex. Some of the most basic dialog acts include binaries. For example, a dialog tag of “yes or no answer” corresponds to the two possible future outcomes, a yes or a no.
Certain kinds of dialog act designations refer to general language tasks. For example, the dialog tag “salutation” could be attached to any simple greeting such as the word “hello.” Some other simple dialog tags more specifically show a function. For example, the dialog act “instruction” is a command to carry out something, where the dialog tag “acknowledge” simply shows that the speaker has heard a preceding piece of dialogue from someone else.
Some simple dialog acts and tags take specific suffixes for parsing. The dialog act tag, “query,” can take either a “yes/no” or “open” suffix to designate either a yes or no question or an open question. The “reply” dialog tag can have a suffix of “yes,” “no,” or “open.”
In many cases, researchers display dialog acts in a column next to the column of dialog. This helps the reader to understand that each specific dialog act tag corresponds to something that a speaker is saying. By reading down through both columns, the reader can understand how the dialog acts represent the functions of the included dialog.
The designation of dialog acts is critically important to a range of software technologies. One of these is the artificial intelligence technology, where humans try to use machines to mimic human thought or expression. Dialog act markers can also help with analyzing natural language for the purposes of speech recognition, or other kinds of software. They can be critically important any time a computer or machine needs to analyze and parse meaning from recorded or written human speech.
In specific research applications, dialog acts are often expressly conveyed through visual charts or graphs. Researchers may use a kind of bubble map to show where the most common dialog acts exits in a conversation. Alternately, researchers can simply use the two rows above, with some color coding or letters and numbers that visually track different kind of dialog acts.