1NGRAMMAKE(1)                     User Commands                    NGRAMMAKE(1)
2
3
4

NAME

6       ngrammake - manual page for ngrammake 1.3.4
7

DESCRIPTION

9       Make ngram model from input count file.
10
11              Usage: ngrammake [--options] [in.fst [out.fst]]
12
13       PROGRAM FLAGS:
14
15       --backoff: type = bool, default = false
16
17              Use  backoff smoothing (default: method dependent) --backoff_la‐
18              bel: type = int64, default = 0  Backoff  label  --bins:  type  =
19              int64,  default  =  -1  Number of bins for katz or absolute dis‐
20              counting --check_consistency: type = bool, default = false Check
21              model consistency --count_of_counts: type = string, default = ""
22              Read count-of-counts from file --discount_D: type = double,  de‐
23              fault  = -1 Absolute discount value D to use --interpolate: type
24              = bool, default = false  Use  interpolated  smoothing  (default:
25              method  dependent) --method: type = string, default = "katz" One
26              of:  "absolute",  "katz",  "kneser_ney",   "presmoothed",   "un‐
27              smoothed", "katz_frac", "witten_bell" --norm_eps: type = double,
28              default = 0.001  Normalization  check  epsilon  --witten_bell_k:
29              type = double, default = 1 Witten-Bell hyperparameter K
30
31       LIBRARY FLAGS:
32
33       Flags from: flags.cc
34
35       --help: type = bool, default = false
36
37              show usage information --helpshort: type = bool, default = false
38              show brief usage information --tmpdir: type = string, default  =
39              "/tmp"  temporary  directory --v: type = int32, default = 0 ver‐
40              bosity level
41
42       Flags from: fst.cc
43
44       --fst_align: type = bool, default = false
45
46              Write FST data aligned where appropriate --fst_default_cache_gc:
47              type  =  bool, default = true Enable garbage collection of cache
48              --fst_default_cache_gc_limit: type = int64,  default  =  1048576
49              Cache    byte    size    that    triggers   garbage   collection
50              --fst_read_mode: type = string, default =  "read"  Default  file
51              reading  mode for mappable files --fst_verify_properties: type =
52              bool, default = false Verify FST properties queried by TestProp‐
53              erties  --save_relabel_ipairs:  type = string, default = "" Save
54              input  relabel  pairs  to  file  --save_relabel_opairs:  type  =
55              string, default = "" Save output relabel pairs to file
56
57       Flags from: ngram-output.cc
58
59       --end_symbol: type = string, default = "</s>"
60
61              Class  label for sentence end --start_symbol: type = string, de‐
62              fault = "<s>" Class label for sentence start
63
64       Flags from: symbol-table.cc
65
66       --fst_compat_symbols: type = bool, default = true
67
68              Require symbol tables to match when appropriate --fst_field_sep‐
69              arator: type = string, default = "      " Set of characters used
70              as a separator between printed fields
71
72       Flags from: util.cc
73
74       --fst_error_fatal: type = bool, default = true
75
76              FST errors are fatal; o.w. return objects flagged as bad:  e.g.,
77              FSTs:   kError   property  set,  FST  weights:  not  a  Member()
78              --ngram_error_fatal: type = bool, default =  true  NGram  errors
79              are  fatal  if  true;  otherwise returns objects flagged as bad:
80              e.g., NGramModel::Error() is true
81
82       Flags from: weight.cc
83
84       --fst_weight_parentheses: type = string, default = ""
85
86              Characters enclosing the first weight  of  a  printed  composite
87              weight  (e.g., pair weight, tuple weight and derived classes) to
88              ensure proper I/O of nested composite weights; must have size  0
89              (none) or 2 (open and close parenthesis) --fst_weight_separator:
90              type = string, default = "," Character separator between printed
91              composite weights; must be a single character
92
93
94
95ngrammake 1.3.4                  January 2018                     NGRAMMAKE(1)