search

pareco-spike-team / research

Research
2 stars 3 forks source link

[bug] missing tags/connections on articles #4

Open Hurix opened 4 years ago

Hurix commented 4 years ago

On a search for "Aisling" or "Aisling Duval", it correctly shows everything that is tagged and not tagged. The bug is that a few articles also show up without any connection. The articles should always have the respective tags which they were found with. Right?

Hurix commented 4 years ago

I see existing tags that are not connected to a relevant article show with yellow background in the articles text. Is that a conscious decision to not add the connection and make it green?

MorganPersson commented 4 years ago

Example of article that should have a tag: 3301-08-15 Aisling Duval Reviews her Security.

MorganPersson commented 4 years ago

Well, the search actually searches tags and a regex match on articles so stuff like week in review will show up. If you look in the dataImport/updateArticleTags.js you see some article titles will not be tagged.

Yellow background = The tag Green background = An article

I rebuilt my DB locally and my example in my other comment is connected so the example is not valid.

Hurix commented 4 years ago

If you talk about the nodes in the map, that is correct. I referred to the right text panel. There is also yellow marked words and green marked words. The green words also appear as tags below the article and I suspect they are responsible for the map lines I described as connections. There are articles with existing tags in yellow background that should be green and create a connection line in the map, like other articles do for the same words.

Hurix commented 4 years ago

Missclicked sorry

MorganPersson commented 4 years ago

I talked about the node map. So, when you click on a round article, the text shows up in the right panel. All tags associated with that article is also loaded at that time and shown in the right panel. If you double-click the ball in the left panel all those tags are now "expanded" and the tags ad linked articles are shown. The yellow tags you see in the right panel is some code that tries to suggest new tags.

Hurix commented 4 years ago

You mentioned that rebuilding the DB fixed it for one example. So the tagging of the import is inconsistent across multiple imports? As far as I can see, the import script tags only those articles that explicitly have tags assigned (via markdown). So the inconsistency should be in the source file but the import process should be consistent. Your example of reimporting to fix the tags of one article hints against this logic. I am confused about what is actually happening.

The bugreport here is still the same: Right panel shows suggested tags that match existing tags - should be real tags that get added in the db as part of the import process (even tho markdown may not have that tag explicitly).

To formulate it as a feature if you like: Automatically add article-tag relation to articles while importing articles (new or from file), where the tag is found in the text but no real tag is existing yet.

MorganPersson commented 4 years ago

I emptied my DB, I suspect I had mucked around in it so that may be the cause of the inconsistency in my case. Then I ran the import.sh script. As far as I can tell the import is consistent.