LogosLink User's Manual · LogosLink version 2.0.0.2

Canonicity Analytics (Argumentation Model)

Canonicity analytics shows how canonical each locution in an argumentation model is, and how canonical the overall model is.

This is useful to assess to what degree locutions in the model constitute separate sentences that follow standard grammatical rules.

Parameters

Speaker. If you select a speaker, only the locutions of this speaker will be taken into account.
Agent. If the argumentation model has an embedded context and you select an agent, only the locutions of speakers linked to this agent will be taken into account.
Top class threshold. This is the threshold above which canonicities are considered to belong to the top class. Typical values are around 99%. See the Details section for additional information.
Bottom class threshold. This is the threshold below which canonicities are considered to belong to the bottom class. Typical values are around 70%. See the Details section for additional information.

Results

Results are given as some overall data, individual canonicities for each locution, and average canonicities for linked context elements, if there is an embedded context.

Overall data

Overall canonicity. This is the overall canonicity of the argumentation model. See the Details section for additional information on how this is calculated.

An overall canonicity close to 100% indicates that the locutions in the model constitute separate sentences that adhere completely to standard grammatical rules. Values below 100% indicate that the locutions do not constitute separate sentences and/or deviate, in a corresponding degree, from standard grammatical rules.

Locutions

The list shows the locutions in the argumentation model, plus some details for each one.

For each locution, the following data are shown:

Key, Content and Speaker. These show the main details of the locution.
Canonicity. This is the canonicity of the locution, calculated as described in the Details section below.

Locutions are also classified into top, middle and bottom classes, according to their canonicities and the chosen thresholds. Top class locutions are highlighted in green, middle class locutions are not highlighted, and bottom class locutions are highlighted in pink.

Linked Themes

The tree table shows the themes in the embedded context, plus some details for each one.

For each theme, the following data are shown:

Canonicity. This is the average canonicity of the locutions connected to propositions that are linked to the theme.

Linked Positions

The tree table shows the positions in the embedded context, organised by theme, plus some details for each one.

For each position, the following data are shown:

Canonicity. This is the average canonicity of the locutions connected to propositions that are linked to the position with a support degree.

Details

Locution canonicity

Canonicity is a measure of how canonical a locution is. This has two components: how well the locution works as a separate sentence, and how much the locution follows standard grammatical rules.

Since propositions are expected to constitute individual sentenes and adhere to standard grammatical rules, the propositions related to each locution are used as a benchmark against which the locution is compared. In this sense, if the text context of a locution is identical to the text content of the associated proposition, this locution is considered to have a 100% canonicity. The more the locution text deviates from the associated proposition text, the lower its canonicity will be.

In order to measure the similarity of a locution text and the corresponding proposition text, the Jaro-Winkler algorithm is used.

Locutions with no associated propositions are not considered by the analytics. For locutions with mutiple associated propositions, similarity is measured against each one, and then an average is taken.

Overall canonicity

The overall canonicity for the whole argumentation model is calculated as (1 + (T - B)) / 2, where T and B are the proportion of locutions in the top and bottom classes, respectively.