The purpose of the gender detection is to determine whether a profile is a male, a female or a brand. If no gender is identified, no gender will be assigned to that mention.
The algorithm looks at the name and the language of a user. Based on that we determine the gender.
Example: Emeline in French refer to a female name.
If the gender based on the name is ambiguous (example: Camille in French) or if the name is unknown (if it is a nickname for instance), the algorithm will look at the description in the account of the user, its timeline (the 200 last tweets/posts of the user) and the colours used in the profile (for Twitter online) in order to find clues for what gender it may be. All theses elements will help determine the most likely gender.
Example: the description is "Proud mother of 3". The algorithm will find that "mother" refers to a female gender.
For web mentions, we look at the name of the page author and the language associated with it. If for a specific language our algorithm has no doubts what the gender is, we assign one based on that name. Otherwise the gender remains unknown and won't be displayed in our graphs as the one below.
Note: Since some mentions remain without a gender the total number of males and female mentions may not add up to the total volume of mentions.