Using StorySurvey: admin interface

This documentation is work in progress.

You have to log in with one of the accounts which is hardcoded as admin.

The SQL table ss2embeddings has columns for survey (“project”), original phrase (“old”) and then hundreds of columns containing the embeddings for each original phrase.

This table can be (re-)constructed with the function generate_embeddings() (there is a button in the Cluster panel). This should not be necessary as the embedding for each new phrase is added to the table ss2embeddings when the respondent enters it.

The SQL table ss2recodes has xxxxxxx columns for survey (“project”), original phrase (“old”) and then hundreds of columns containing the embeddings for each original phrase.

Explorer

Here you can click to explore the factor structure

Map

A view of the complete map with all links with frequency of 2 or above

Project recodes

All the current recodes on the current project, including factors which have been hidden.

Project settings

TBC

Project answers

TBC

Project questions

If you change the text of the questions and there are already answers recorded, these answers will become detached from the original questions.

Each question can also have a short version which is used instead of the long question in the maps.

Project recodes ace: The recodes advanced editor

The tab you need is Project recodes ace which is the first tab.

Basic idea

The idea is that you recode daughters like Y into mothers like X by adding X; in front of Y.

This is very similar to the idea from Causal Map that increased income; cash from sale of pig is an example of increased income, but with StorySurvey the more detailed part of the label is already created by the respondent, and the analyst’s job (or the NLP’s job) is only to classify cash from sale of pig as an example of increased income.

Also, in Causal Map, the different lower-level labels should be meaningfully different variations of a central idea, whereas in StorySurvey that is only half the job, the other half is simply about merging different formulations of exactly the same idea.

Unlike in Causal Map, in StorySurvey the purpose of the recodes is not mainly to be able to sometimes display outputs with all the different variations. In StorySurvey, the daughters in a recode set are more like the different quotes in Causal Map – they serve to give colour to, and illustrate, each individual factor.