Word activity card and flow

[wcjord]

The view when a user first clicks a token, accompanied by sound of the text-to-speech of the word. The user can decide which of the icons to click first though the translation icon is prominent in the center. The seed/sprout/flower in the upper right represents the user's progress with its lemma.

Not automatically selecting translation makes for one more click, however, it might be advantageous to learning to allow them to discern the translation from some other cues that will be often present. It also has the opportunity to create an interesting game dynamic where they might first want to go through some of the other activities, and get hints about the translation, before going for the translation activity.

When the user clicks the translation icon, the translation activity begins. They are presented with 3-4 possible meanings of the word. The translation activity is worth the most points on a word. A user can gain 5pts. They can also lose as many as 15 from wrong answers.

The distractors above are different conjugations of the lemma. This would be for an advanced learner who should already know the meaning. A learner with less demonstrated proficiency on the word would get distractors that vary more widely in meaning such those below.

Translation activity with distractor varying more widely in meaning.

After the user completes the translation activity, and sees the success animation with points, they now see the translation displayed. When a user opens a word card, we'll decide whether we think they should do the translation activity. If not, then the translation appears without going through the activity.

When the user clicks the light-gray box, the placeholder for part of speech icon, they get a part of speech activity.

After the part of speech activity and success animation. They can now see the light-gray placeholder icon of the next most important, morphological feature for the word. In the case of a verb, this is Tense, represented by the silhouette of an hourglass.

Upon click of the light-gray emoji-add icon, the user gets an emoji activity. There is no right/wrong in the activity and the user gets +1 XP upon whatever selection they make. The user can choose to do this before the translation activity as a hint towards the meaning of the word.

Upon selection of the play audio button beside the word itself, we could give them an activity. We show the phonetic spelling and four choices of what the audio could be. We can make the distractors more or less similar to adjust difficulty.

For an English learner of Spanish, this might not be sufficiently challenging or good for learning. In this case, clicking the audio just plays it and displays the phonetic spelling. For learners of languages with entirely different alphabets (like me learning Russian), this could be a good activity.

[wcjord]

We could also give an activity for the lemma. This might be relevant when the word and lemma were very different in spelling.

[wcjord]

There are several different activity types in this flow with individual needs for content generation:

• word meaning
• emoji
• morphological
• lemma
• audio

Translation:
The current practice activity generation endpoint with word meaning as activity type works for v1. We can optimize content quality by using the contextual translation for the correct option and distractor generation for incorrect. We could potentially optimize API costs by pulling distractors from the contextual translations of other tokens, potentially within the message.

Emoji:
The choices for the emoji can be obtained via LemmaDictionaryRepo.get. There will be no distinction between correct/incorrect choices. Once the user has made their selection for a lemma, we'll need someplace to store them. Let's store them in the user's analytics room with a p.construct.emoji event, with a state key being lemma + cat.

Morphs:
We will deliver activities for features that are both 1) in the token.morph map and 2) in the list of features marked as appropriate for activities for that part of speech. Distractors can be randomly chosen from other labels that exist for that feature.

Lemma:
The correct answer is token.lemma.text. The distractors can be pulled from other words in the user's vocabulary, maybe those that are closest in spelling (minimum Levenshtein distance) and/or have the same part of speech. Before we've implemented this activity or, if it's not deemed level appropriate, then clicking the seed/sprout/flower can take the user directly to the vocab details page.

Audio
The text to speech audio can be generated locally from token.content.text and distractors. Distractors can be other tokens in the message. The phonetic spelling of the word should be generated for each form of the word. This should potentially be added to the lemma definition endpoint.

@ggurdin These are my plans for getting and storing information for the activities. Give it a read please and let me know if you have any questions/comments. I'll shoot to start this tomorrow morning.