--lex_sqrt

Switch

--lex_sqrt

Description

Applies a square-root transformation to the final lexicon/category value produced by --add_lex_table (or --add_corp_lex_table), as opposed to --sqrt which transforms the underlying word/ngram frequencies that feed into that value.

Argument and Default Value

No argument. Boolean switch, default: False (no transform applied to the lexicon value).

Details

Verified: unlike --sqrt (which adds a trailing $16to16-style segment), --lex_sqrt folds its suffix directly into the lexicon name portion of the table, joined with an underscore, e.g. feat$cat_met_a30_2000_cp_w_16to4$msgs$user_id — no $ before 16to4. The number is int(featValueFunc(16)) (truncated, not rounded) — for sqrt this coincides with rounding since sqrt(16) == 4.0 exactly, but --lex_log shows the truncation clearly (int(log(17)) == 2, not 3).

--sqrt and --lex_sqrt can be combined in the same command — they're two independent value functions (valueFunc for the underlying word/ngram counts, featValueFunc for the resulting lexicon category value) passed into the same feature-extraction call, and each can add its own suffix.

Other Switches

Required Switches:

• --add_lex_table or --add_corp_lex_table

• -l LEX_TABLE

Optional Switches:

• --lex_log, --lex_boolean, --lex_anscombe (mutually exclusive alternatives — only one lexicon value transform can be active at a time)

Example Commands

dlatkInterface.py -d dla_tutorial -t msgs -c user_id --add_lex_table -l met_a30_2000_cp --weighted_lexicon --lex_sqrt