Implement English-to-Cree search, based on WordNet senses

[aarppe]

Here's a mock-up HTML version representing what an implementation of searching with English words would look like, when matched with WordNet senses under which the Cree entries would be organized based on their manual classifications according to WN (in some order, here in decreasing corpus frequency).

---

1. WAIT (V): stay in one place and anticipate or expect something
   [ (v) wait#1 (100) ]
• pêhêw (VTA-1)
1. s/he waits for s.o.
• pêhow (VAI-1)
1. s/he waits
2. s/he expects, s/he is in expectation (of someone's arrival)
• pêhtâw (VTI-2)
1. s/he waits for s.t.
• aswaham (VTI-1)
1. s/he waits eagerly for s.t., s/he lies in wait for s.t.
2. s/he catches s.t. as it drips
• pa-pêhêw (VTA-1)
1. s/he is still waiting for s.o.


2. WAIT (V): look forward to the probable occurrence of
   [ (v) wait#3 (33) - (v) await#1 (25) - (v) expect#3 (12) - (v) look#8 (1) ]
• pêhow (VAI-1)
1. s/he waits
2. s/he expects, s/he is in expectation (of someone's arrival)
• âswamêw (VTA-1)
1. s/he awaits s.o. (as food) with an open mouth
• âswahtam (VTI-1)
1. s/he awaits s.t. (as food) with an open mouth
• âswahcikêw (VAI-1)
1. s/he awaits food with an open mouth (e.g. baby, bird)


3. WAIT (V): wait before acting
   [ (v) wait#2 (49) - (v) hold_off#2 (0) - (v) hold_back#4 (0) ]


4. WAIT (N): the act of waiting (remaining inactive in one place while expecting something)
   [ (n) waiting#1 (2) - (n) wait#2 (2) ]
• pêhowin (NI-1)
1. waiting
2. expecting someone


5. WAIT (N): time during which some action is awaited
   [ (n) wait#1 (2) - (n) delay#1 (2) - (n) time_lag#1 (0) - (n) postponement#1 (0) - (n) hold#4 (0) ]


6. WAIT (V): serve as a waiter or waitress in a restaurant
   [ (v) wait#4 (2) - (v) waitress#1 (0) ]

---

The above is created with the following incantation, run in the ALTLab repo in: altlab/eng/bin/:

./sort-cw-entries-by-wn-senses.sh ../../crk/dicts/Wolvengrey_WN_RW.tsv ../dicts/WordNet-3.0 ../../crk/generated/ahenakew_wolfart_bloomfield.fst+cg.freq-sorted.txt | ./organize-cw-wn-senses-by-wn-synsets.sh | ./pretty-print-wn-sense-search.sh 'wait' '' html | less

What this would require are the following:

• Whether the Wordnet senses pertaining to a WN synset should be shown or not should be considered.
• Most if not all Cree entries in all the dictionaries have been classified (manually or otherwise) according to WN senses
• All Cree entries with WN senses that belong to the same synset are grouped together (so we would need to incorporate the WN ontology).
• WN synsets are shown in their descending frequency (as known from the SEMCOR corpus, and included in the WN source data).
• Cree entries under synsets need to be sorted in some way. The above implementation uses descending corpus-frequencies from the AW+B corpus, though an alternative could be the glossary counts. Nevertheless, one might want to be able to manually adjust and improve this ordering.
• Potentially some Cree entries might be missing from some WN synset / sense. This can be fixed by adding the WN semantic class to the Cree entry in question.

Some further thinking would be needed for the following:

• multiword searches should be tokenized using a list of known multiword phrases, when applicable, e.g. blow up or give up.
• The English phrase translation module should perhaps be expanded to be able to analyze the inflected word forms of any WordNet nouns or verbs.
• for English search words that do not match with the WN classification of any Cree word entry, we could either apply the semantic-vector method, or in the case of such a word nevertheless occurring in WordNet, providing the Cree entries that have a semantic classification from the same synset as the search term, or it that is not the case, then CW entries for which the semantic classification is graph-edge-wise closest to the search term.

[aarppe]

@nienna73 @DBDacanay I couldn't resist creating a full-fledged mock-up example case, illustrating what the English-to-Cree search could actually look like, describing the broad strokes and open questions in this meta-issue. When you need a break, you might consider whether the above would make sense, or might benefit from some further tweaking.