dialign-online

Tutorial to try dialign-online for Interactive Purposes

In this tutorial, you will:

• learn how to run dialign-online in interactive mode and export the transcript of the created dialogue ; and
• learn how to run dialign-online on a single dialogue transcript in order to directly compute online metrics for each turn.

Running dialign-online in Interactive Mode

dialign-online can be ran in interactive mode to simulate an ongoing chat dialogue.

Input Transcript (optional)

dialign-online can take an optional dialogue transcript that represents the starting point of the dialogue. If no dialogue transcript is provided, dialign-online starts from an empty dialogue.

In this tutorial, the transcript "excerpt-alice-partial.tsv" is used as a starting point. Its content is reproduced below:

Human:	okay well but the books like  there are many eh facts in the book which are against the law of physics like the piece of cake which is which makes eh Alice bigger or smaller i don't remember but which
Agent:	indeed
Human:	yeah eh does the piece of cake makes Alice bigger or smaller ?
Agent:	the piece of cake grows Alice
Human:	grows Alice okay and what makes her smaller ?
Agent:	sorry could you repeat that ?
Human:	what makes eh Alice smaller ?

The dialogue format used by dialign-online is the same tsv format (tab-separated values) as the one used for corpus studies.

Running an Interactive Session

To start an interactive session, just run dialign-online with or without a dialogue transcript:

java -jar dialign-online.jar -f excerpt-alice-partial.tsv
# if no dialogue transcript is provided, simply do:
# java -jar dialign-online.jar

Results :

dialign-online prints the current dialogue history and then prompts for the next step. To obtain a description of the possible options, just enter "h" (as "help"):

Scoring and Adding Utterances

Next, let's try to score a new utterance with the following features:

• Locutor: "Agent"
• Utterance: "throughout the book there are many making her smaller"

To do that, enter "s" (as "score") and specify the "locutor" and the "utterance". This returns the following result:

Scoring the utterance returns multiple information:

• information related to verbal alignement:
  • DER (Dynamic Expression Repetition) which scores between 0.0 (not all) and 1.0 (completely) how much the utterance reuses shared lexical patterns
  • Reused shared lexical patterns: the list of shared lexical patterns that are reused in this utterance
  • Established shared lexical patterns: the list of shared lexical patterns that are reused for the first time (thus "established")
• information related to self-repetitions:
  • DSER (Dynamic Self-repetition Expression Repetition): which scores between 0.0 (not all) and 1.0 (completely) how much the utterance reuses self-repetitions
  • Reused self-repetition patterns: the list of self-repetitions that are reused in this utterance.

Then, the scored utterance can be added to the dialogue history ("y") or discarded ("n", default choice). Here, let's add the utterance ("y"):

It is also possible to directly add an utterance by pressing "a". For instance, let's directly add the following utterance:

• Locutor: "Human"
• Utterance: "i'm sorry i didn't understand"

Directly adding this utterance gives:

Then, let's score the following agent response:

• Locutor: "Agent"
• Utterance: "throughout the book there are many passages where Alice shrinks"

This gives the following scoring results:

Note that this last utterance both verbally aligns and verbally self-repeats. Let's add the utterance to the dialogue history by pressing "y".

Exporting the Created Dialogue Transcript and Closing the Interactive Session

To conclude this tutorial, let's export the created dialogue. To do this, simply press "e" and enter a filepath:

Exporting produces a tsv file with the following columns (the full file is available at dialign-online-excerpt-alice.tsv):

locutor	utterance	der	sharedExpressions	establishedSharedExpressions	dser	selfRepetitions

• locutor: the locutor of the utterance
• utterance: the text content of the utterance
• der: the "Dynamic Expression Repetition" score of the utterance
• sharedExpressions: the list of shared lexical patterns reused in this utterance
• establishedSharedExpressions: the list of shared lexical patterns established in this utterance
• dser: the "Dynamic Self-repetition Expression Repetition" score of the utterance
• selfRepetitions: the list of lexical self-repetitions used in the utterance

Finally, the interactive session can be closed by pressing "q" (as "quit").

Running dialign-online on a Dialogue Transcript

dialign-online can be ran directly on a file to compute online metrics for each turn.

To do this, an input dialogue transcript and an output tsv file should be provided. For instance with:

• input dialogue transcript: excerpt-alice-partial.tsv
• output tsv file: dialign-online-excerpt-alice-partial.tsv

dialign-online can be executed as follows:

java -jar dialign-online.jar -f excerpt-alice-partial.tsv -o dialign-online-excerpt-alice-partial.tsv

The output file format is the same as exporting from the interactive session (see above).