How Comprende works, and how we checked it

Taste: the Qloo taste graph

Your favourites, from any domain, are looked up in Qloo and passed as signals to Qloo's Insights API. Qloo returns films and series originally released in Spanish-speaking countries that audiences with similar tastes enjoy, scores the titles we measured against the same signals, and reports which favourite drove each match. Titles are ranked by how much more they fit your taste than a typical one; between close matches, the better-known and more recent title (Qloo popularity, release year) goes first. Qloo describes what groups of people tend to like; it does not predict what one person will do.

Difficulty: words you know, in real dialogue

For each Spanish-original film or series with subtitles in the open OPUS OpenSubtitles corpus, we count the words in its dialogue and rank Spanish words by how many titles use them. If you know the 3,000 most common words, the share of a title's running words you know is its coverage for you. Proper names and interjections count as known.

Studies of reading and listening put adequate comprehension at about 95% coverage and comfortable, unassisted comprehension near 98% (Hu & Nation 2000; Nation 2006). Comprende uses 95% for “watch now” and 90–95% for “a good stretch”.

Your vocabulary comes from a yes/no word check: 48 real words from six frequency bands and 16 made-up words. Saying yes to made-up words lowers the estimate, the standard correction for guessing (Meara & Buxton 1987). Words spelled almost the same in English or French (cognates such as “perfecto” or “fotografía”) are left out of the check, because they can be recognised without knowing Spanish.