3.8 KiB
Earlybird Light Ranker
Note: the light ranker is an old part of the stack which we are currently in the process of replacing. The current model was last trained several years ago, and uses some very strange features. We are working on training a new model, and eventually rebuilding this part of the stack entirely.
The Earlybird light ranker is a logistic regression model which predicts the likelihood that the user will engage with a tweet. It is intended to be a simplified version of the heavy ranker which can run on a greater amount of tweets.
There are currently 2 main light ranker models in use: one for ranking in network tweets (recap_earlybird), and
another for
out of network (UTEG) tweets (rectweet_earlybird). Both models are trained using the train.py script which is
included in this directory. They differ mainly in the set of features
used by the model.
The in network model uses
the src/python/twitter/deepbird/projects/timelines/configs/recap/feature_config.py file to define the
feature configuration, while the
out of network model uses src/python/twitter/deepbird/projects/timelines/configs/rectweet_earlybird/feature_config.py.
The train.py script is essentially a series of hooks provided to for Twitter's twml framework to execute,
which is included under twml/.
Features
The light ranker features pipeline is as follows:
Some of these components are explained below:
-
Index Ingester: This is an indexing pipeline that handles the tweets as they are generated. It is the main input of Earlybird, producing Tweet Data (the basic information about the tweet, such as the text, URLs, media entities, facets, etc.) and Static Features (features that can be computed directly from a tweet, such as whether it has a URL, cards, quotes, etc.). All information computed here is stored in the index and flushed as each real-time index segment becomes full. They are loaded back later from disk when Earlybird restarts. Note that the features may be computed in a non-trivial way (like deciding the value of hasUrl), as they could be computed and combined from some more "raw" information in the tweet and from other services.
-
Signal Ingester: This is the ingester for Realtime Features, which are per-tweet features that can change after the tweet has been indexed. They mostly include social engagements like retweetCount, favCount, replyCount, etc., along with some (future) spam signals that are computed with later activities. These features are collected and computed in a Heron topology by processing multiple event streams and can be extended to support more features.
-
User Table Features: This is another set of features per user, which are from User Table Updater, a different input that processes a stream written by our user service. It is used to store sparse real-time user information. These per-user features are propagated to the tweet being scored by looking up the author of the tweet.
-
Search Context Features: These are basically the information of the current searcher, such as their UI language, their own produced/consumed language, and the current time (implied). They are combined with Tweet Data to compute some of the features used in scoring.
The scoring function in Earlybird uses both static and real-time features. Examples of static features used are:-
- Whether the tweet is a retweet
- Whether the tweet contains a link
- Whether this tweet has any trend words at ingestion time
- Whether the tweet is a reply
- A score for the static quality of the text, computed in TweetTextScorer.java in the Ingester. Based on factors such as offensiveness, content entropy, "shout" score, length, and readability.
- Tweepcred (see top-level README.md)
Examples of real-time features used are:
- Number of tweet likes/replies/retweets
- pToxicity and pBlock scores provided by health models