Speaker Interference
Symbol @n… …n@
Definition

Speaker interference occurs when speakers speak at the same time (also referred to as 'cross-talk'), or when various noises interfere with a particular speaker as he or she is speaking.

An interference may be either passive or active. There will always be one passive and one active interference occurring in combination for any given instance of speaker interference. One speaker speaks actively, while the other speaks passively.

Passive Interference: refers to the utterance being interrupted.
Active Interference: refers to the utterance doing the interrupting.

Pairs always occur in numerical order, beginning with one [1].

Rules for Transcribing the @n… …n@
 

1. Both interferences are marked with a number and the 'at' sign [@].

2. The number refers to the number of the interference within the dialog. Numbers begin with 1 and continue as needed.

3. Interferences are grouped in pairs, so for every active interference, there is a matching passive interference. These are marked with the same number.

Passive interferences are marked at the end of the word being interrupted (word-final). In passive cases, the number precedes the @-sign.

Active interferences are marked before the word that is interrupting (word-initial) of the element in question. In active cases, the number follows the @-sign.

When to tag speaker interferences:
In general, elements of speech are tagged as interferences when the utterances of different speakers occur simultaneously. Only certain kinds of utterances, however, are transcribed as speaker interferences. Note that there is a difference between what elements can be passively interfered, and what can be actively interfering.
  Passively interfered turn elements to be marked are:

1. Lexical items

2. All human noises (e.g. breathing, coughing)

3. Pauses during speech

4. Filled pauses

5. Technical noises

Add picture
  Actively interfering turn elements to be marked are:

 

1. Lexical items

2. Human noises

3. Filled pauses

Add picture
When not to tag speaker interferences

1. Pauses can only be tagged as passive interferences because they cannot actively cause any interference themselves. Utterances which are overlapped by pauses actually interfere with nothing.

Example:

Speaker 1: hello , what's1@ your name ?
Speaker 2: .........@1again , <P><Noise> ......... I am Vicky .

Within the turn of the second speaker, the words of the first speaker being uttered (and therefore heard) during <P> are referred to by the <Noise> -tag (in this example it is the phrase "your name").

2. Breathings, pauses and noises occurring as passive interferences are annotated with an additional carat following the 'at' sign [@]. They are only annotated within a turn, never at a turn's end.
Overlapped breathings and human noises are annotated as: <B> <Noise>and <Noise> <Noise> , where the second

<Noise>-tag represents the overlapping utterance of the other speaker.

Examples:

Speaker 1: I don't know .<B> <Noise>
Speaker 2: never mind.

Speaker 1: no .
Speaker 2: <Noise> <Noise> what was that ?

The reason why <B> and <Noise> may occur at the end of turns is because it is necessary not to cut turns too close to a lexical item.

So in general, whenever this technical aspect of transcription requires that you not position turncuts too close to a lexical item, additional acoustic contributions by another speaker are referred to as <Noise> .

Example:

Speaker 1: Do you want an explanation for that ? <Noise>
Speaker 2: <Noise>Yes , please .

Segregation and numeration of interferences:
Pairs of interfering and interfered groups are numbered chronologically, beginning with 1. Several elements may belong to the same interference (and so are marked with the same number) under the following conditions:

1. As long as there is no element occurring within the utterance of the passively interfered speaker which is not part of the interference (but could generally be one like lexical items, human noise, filled pauses, pauses).

Example:

Speaker 1: I .....think1@.....so.........................but2@ …
Speaker 2: .......@1do.........<P> <Noise> .....
. @2you ?

In this case 'so' is not part of any interference (because <P> cannot be treated as an active interference). So the interferences involving 'think' and 'but' must be considered as different instances of interference.

Elements that are not tagged as active speaker interferences (like technical noises in the active-interfering state) are not to be seen as interference-separators

 

2. As long as there is no turn change:

Example:

Speaker 1: awesome , .....isn't1@ it1@ ?
Speaker 2: ......................@1yeah .
Speaker 1: cool2@ , good2@ bye .
Speaker 2: @2see @2you .

In this case 'isn't it,' 'cool,' and 'good' (response: "yeah" and "see you") belong to different turns and therefore to different interferences.

Given the example:

Speaker 1: yes , that's something I would like1@ to1@ do1@ .
Speaker 2: ....................................@1are @1you sure2@ ?
Speaker 1: .......................................................@2yeah , definitely .

a problem might occur if 'sure' is at the same time part of the active interference 1 and the passive interference 2. A combination of tags like@1sure2@ is not allowed. So you have to decide which interference-tag you are going to assign.

In this example the alternative 'sure2@' is more reasonable because otherwise the second interference might be completely ignored.

To visualize these concepts in a more general way, consider the following diagram, which illustrates the basic principle of how speaker interference works:

 
A
1
2
3
4
Speacking
--> -->
Speacking
Speacking
Not Speaking
 
B  
Not Speaking
 
Interrupting
Interrupting
Speacking
Explanation
 

At interval 1 only Speaker A is speaking. As indicated by the small horizontal arrows [à], Speaker A continues to speak through interval 2. Speaker B has not yet begun to speak or to interfere.

At interval 2 Speaker B interrupts Speaker A. Thus, Speaker B is actively interfering, because it is this speaker's utterance which is causing or initiating the interference. In other words, the interference occurs because of Speaker B's action. Speaker A, on the other hand, is being passively interfered because this speaker is being interrupted. As this is the first instance of interference, it would be marked with the number one.

At interval 3 Speaker B interrupts Speaker A a second time. This instance is distinct from that of the previous interval because Speaker A had stopped speaking, and then started again. Therefore, this instance of speaker interference would be marked with the number two.

At interval 4 only Speaker B is speaking this time, so no interference is occurring here.

Another way to conceptualize the differences between active and passive interference is to imagine that Speaker A - the passively-interfered speaker - has no control over being interrupted, whereas the active-interference of Speaker B is due solely to the fact that Speaker B has total control over when he chooses to speak, and who he might interrupt at the same time.

Transcription

The above diagram would be transcribed as follows:

1. Interval 1 would not be marked as speaker interference.

2. Interval 2 would look like:
Speaker A: (SPEAKING)1@
Speaker B: @1(Interrupting)

3. Because interval 3 is a separate instance of speaker interference, it would look like:
Speaker A: (SPEAKING)2@
Speaker B: @2(Interrupting)

4. Interval 4 would not be marked as speaker interference.

Multiparty interferences

In multiparty recordings, several partners may be speaking at the same time. Only one turn may be passively interfered and all the other dialog partners will actively interfere, although more than one speaker can be passively interfered. Here again the number of each active interference is the same as the number of the passive one. This identifies the connection between the affected elements by the interference.

Example:

Speaker 1: hello , ....~Jim1@ is1@ ....speaking. ....how2@ are2@ you ?
Speaker 2: .... .... @1hello .
....<P><Noise> .... .... @2oh . .... .... ..<P><Noise>fine3@ .
Speaker 3: .... .....@1hello . ... <P><Noise> ... .... ..... .... ..... .... ....... ......@3great .