Outlink topic model, a language-agnostic model, expects the following input arguments:
{ "lang": "en", "page_title": "Toni Morrison", "threshold": 0.3 }
where lang and page_tile are required, threshold is optional.
It makes prediction based on outlinks associated with a Wikipedia article:
{"prediction": {"article": "https://en.wikipedia.org/wiki/Toni Morrison", "results": [{"topic": "Culture.Biography.Biography*", "score": 0.9615437984466553}, {"topic": "Culture.Biography.Women", "score": 0.8872146010398865}, {"topic": "Geography.Regions.Americas.North_America", "score": 0.7186043858528137}, {"topic": "Culture.Literature", "score": 0.5775054097175598}]}}
We want to return meaningful HTTP error messages in Lift Wing if:
- no matching article for the given title
- missing a language
- missing an article title
- missing both language and title
- incorrect argument name e.g. lang vs language
- threshold value provided not a float
Similar to how we've done on revscoring models in T300270, we will use tornado.web.HTTPError and add the status code and message.