`nx_g2p_gate.nx` source

↩ module page · 116 lines · 4986 B
1// nx_g2p_gate.nx -- REFEREE for the text->phoneme front-end (nx_g2p).
2//
3//   MAPPING: words map to the expected phoneme-id sequence (incl. digraphs):
4//     "cat"->[K,AE,T]  "see"->[S,IY]  "moon"->[M,UW,N]  "ship"->[SH,IH,P]
5//     "the"->[TH,EH]   "rock"->[R,AA,K]
6//   INVENTORY: voiced flags correct (vowels/nasals voiced; fricatives/stops not).
7//   DETERMINISM: same text -> same sequence.
8//   END-TO-END: g2p("see") -> records -> nx_voice_read -> the S segment reads
9//     UNVOICED (hiss) and the IY segment reads VOICED (tone)  ==> text became
10//     speech with correct voicing.
11//
12// Evidence -> stdout + knowledge/status/g2p_gate.log.  Exit 0 GREEN / 1 RED.
13// Sovereign x86_64 throughout.  license_tier: ORIGINAL
14import "nx_syscalls_x86_64.nx"
15import "nx_g2p.nx"
16import "nx_voice_read.nx"
17import "nx_voiceprint.nx"
18
19func gp(logfd: i64, s: *u8) -> i64 {
20    var n: i64 = 0
21    while s[n] != (0 as u8) { n = n + 1 }
22    sys_write(1, s, n)
23    if logfd > 0 { sys_write(logfd, s, n) }
24    return 0
25}
26func gn(logfd: i64, v: i64) -> i64 {
27    let bb: *u8 = sys_mmap(28)
28    var m: i64 = v
29    if m < 0 { sys_write(1, "-\x00" as *u8, 1); if logfd > 0 { sys_write(logfd, "-\x00" as *u8, 1) } m = 0 - m }
30    let t: *u8 = sys_mmap(28)
31    var k: i64 = 0
32    if m == 0 { t[0] = 48 as u8; k = 1 }
33    while m > 0 { t[k] = (48 + (m % 10)) as u8; m = m / 10; k = k + 1 }
34    var i: i64 = 0
35    while i < k { bb[i] = t[k - 1 - i]; i = i + 1 }
36    sys_write(1, bb, k)
37    if logfd > 0 { sys_write(logfd, bb, k) }
38    return 0
39}
40func slen(s: *u8) -> i64 {
41    var n: i64 = 0
42    while s[n] != (0 as u8) { n = n + 1 }
43    return n
44}
45// map one word; print; return 1 if it matches expected [e0,e1,e2] (ne ids)
46func chk(logfd: i64, text: *u8, e0: i64, e1: i64, e2: i64, ne: i64, ids: *i64) -> i64 {
47    let cnt: i64 = nx_g2p(text, slen(text), ids, 16)
48    gp(logfd, "  \x00" as *u8); gp(logfd, text)
49    gp(logfd, " -> [\x00" as *u8)
50    var i: i64 = 0
51    while i < cnt { gn(logfd, ids[i]); gp(logfd, " \x00" as *u8); i = i + 1 }
52    gp(logfd, "]\n\x00" as *u8)
53    if cnt != ne { return 0 }
54    if ne >= 1 { if ids[0] != e0 { return 0 } }
55    if ne >= 2 { if ids[1] != e1 { return 0 } }
56    if ne >= 3 { if ids[2] != e2 { return 0 } }
57    return 1
58}
59
60func main() -> i64 {
61    let logfd: i64 = sys_openat_append("knowledge/status/g2p_gate.log\x00" as *u8, 0x1a4)
62    gp(logfd, "G2P-GATE (text -> phoneme)\n\x00" as *u8)
63
64    let ids: *i64 = sys_mmap(16 * 8) as *i64
65    var ok: i64 = 1
66
67    if chk(logfd, "cat\x00" as *u8,  PH_K, PH_AE, PH_T, 3, ids) == 0 { ok = 0 }
68    if chk(logfd, "see\x00" as *u8,  PH_S, PH_IY, 0, 2, ids) == 0 { ok = 0 }
69    if chk(logfd, "moon\x00" as *u8, PH_M, PH_UW, PH_N, 3, ids) == 0 { ok = 0 }
70    if chk(logfd, "ship\x00" as *u8, PH_SH, PH_IH, PH_P, 3, ids) == 0 { ok = 0 }
71    if chk(logfd, "the\x00" as *u8,  PH_TH, PH_EH, 0, 2, ids) == 0 { ok = 0 }
72    if chk(logfd, "rock\x00" as *u8, PH_R, PH_AA, PH_K, 3, ids) == 0 { ok = 0 }
73
74    // inventory voiced flags
75    let rec: *i64 = sys_mmap(8 * 8) as *i64
76    nx_g2p_phoneme(PH_IY, rec); if rec[3] != 1 { ok = 0 }      // vowel voiced
77    nx_g2p_phoneme(PH_M, rec);  if rec[3] != 1 { ok = 0 }      // nasal voiced
78    nx_g2p_phoneme(PH_S, rec);  if rec[3] != 0 { ok = 0 }      // fricative unvoiced
79    nx_g2p_phoneme(PH_T, rec);  if rec[3] != 0 { ok = 0 }      // stop unvoiced
80
81    // determinism
82    let ids2: *i64 = sys_mmap(16 * 8) as *i64
83    let c1: i64 = nx_g2p("cat\x00" as *u8, 3, ids, 16)
84    let c2: i64 = nx_g2p("cat\x00" as *u8, 3, ids2, 16)
85    if c1 != c2 { ok = 0 }
86    var di: i64 = 0
87    while di < c1 { if ids[di] != ids2[di] { ok = 0 } di = di + 1 }
88
89    // END-TO-END: "see" -> records -> nx_voice_read -> S unvoiced, IY voiced
90    let n: i64 = nx_g2p("see\x00" as *u8, 3, ids, 16)
91    let recs: *i64 = sys_mmap(2 * PH_REC * 8) as *i64
92    nx_g2p_phoneme(ids[0], (recs as i64) as *i64)
93    nx_g2p_phoneme(ids[1], (recs as i64 + PH_REC * 8) as *i64)
94    let buf: *u8 = sys_mmap(4096 * 2)
95    nx_voice_read(recs, 2, 120, 100, 100, buf, 4096)
96    let s_dur: i64 = recs[4]            // S duration (samples)
97    let vpS: *i64 = sys_mmap(VP_WORDS * 8) as *i64
98    let vpI: *i64 = sys_mmap(VP_WORDS * 8) as *i64
99    nx_voiceprint_extract((buf as i64) as *u8, 460, vpS)                     // S region
100    nx_voiceprint_extract((buf as i64 + (s_dur + 150) * 2) as *u8, 900, vpI) // IY region
101    gp(logfd, "  e2e see: S voiced-frames=\x00" as *u8); gn(logfd, vpS[0])
102    gp(logfd, " IY voiced-frames=\x00" as *u8); gn(logfd, vpI[0]); gp(logfd, "\n\x00" as *u8)
103    if vpS[0] != 0 { ok = 0 }           // S must be unvoiced
104    if vpI[0] < 1 { ok = 0 }            // IY must be voiced
105
106    if ok == 1 {
107        gp(logfd, "G2P-GATE result=ALL-PASS verdict=GREEN\n\x00" as *u8)
108        if logfd > 0 { sys_close(logfd) }
109        sys_exit(0)
110        return 0
111    }
112    gp(logfd, "G2P-GATE result=FAIL verdict=RED\n\x00" as *u8)
113    if logfd > 0 { sys_close(logfd) }
114    sys_exit(1)
115    return 1
116}
nx_g2p_gate.nx source

`nx_g2p_gate.nx` source
