aleixac
commented
4 years ago I have my thoughts against trying to store so much metadata in the users FE, that will heavily impact the performance of the plugin (both in terms of memory and communications times). Take in mind that the actual performance is already not the ideal, although better that in the beginning. Would it not be enough to post the actual final credibility label that is shown to the user. At least for the first versions I would minimize both the received and posted data from the FE plugin side.
rdenaux
andreasjberg
daffyDukk
To help with refining the definition of the
/twitter/evaluate/labelendpoint, this issue describes a test case where a user submits an accuracy rating for a credibility assessment. It also describes the desired outcome, namely a corresponding call to the content collector api.Context
A co-inform plugin user has internal UUID
cf40ced2-7cc8-98a3-1fb0-02c96d781626. The user has logged into the plugin and views this tweet. The plugin calls the/twitter/tweet, which is assigned a query id as follows:The plugin, then retrieves the response via the api which returns the following credibility assessment:
The FE uses this to show a label
not credible' for the tweet. The user clicks on a button to **agree** with this credibility assessment, which triggers the FE to call the proposed [twitter/evaluate/label`](https://co-inform.github.io/gateway-api/#/to%20be%20developed/post_twitter_evaluate_label) endpoint.Desired outcome
The GW should issue the following call to content collector api, corresponding to the data the user entered.
The GW calls the
/user/accuracy-reviewwith a POST and the following content objectVariant
Note that if the user disagrees the only changes from the case described above is that:
namewill have valueinaccurateandreviewRating.ratingValuewill also have valueinaccurate.Note
The content collector (CC) will specify an enum for credibility values. It'd be nice to align this with the GW and FE, otherwise the GW will have to map between FE and CC enums.
Issue
When the user clicks on the submit button, the FE calls the
/twitter/evaluate/labelvia POST. Please verify that the data submitted by the FE is sufficient to allow the GW implementation to convert it to the format required by the content collector.Important Consideration
In order for the rating to be useful, we need to know what the label is that we showed to the user, as well as the tweet for which we generated the label.
Via the
itemReviewed.url, the content collector may try to reconstruct the data that was sent to the FE. However, this is dangerous, since the response may have changed as new results become available. Therefore, it may be better to, besides sending theitemReviewed.urlanditemReviewed.identifier, also forward the fullModuleResponsethat was sent to the FE. I'd introduce propertyitemReviewed.moduleResponsefor this. This requres the FE to keep it and send it along with the rating. The main advantage of having this data is that we should be able to see whether some partial responses perform better than others.