Commit Graph

905 Commits (53fb86c447bc322c13212ae8eaa102f6a3e57604)

Author SHA1 Message Date
Keisuke Kuroyanagi 941734695b Implement Ver4PatriciaTriePolicy::getNextWordAndNextToken.
Bug: 12810574
Change-Id: Idea44f03c477964f58e65fbf2b55e3fcd77a2934
2014-02-04 12:42:14 +09:00
Keisuke Kuroyanagi 38f341a2a5 Add a jni method to iterate words in a dictionary.
Bug: 12810574
Change-Id: I87b13559765f0262dfd06c2ba40d43af6437ea7f
2014-02-03 14:51:58 +09:00
Keisuke Kuroyanagi 9221772ab7 Get bigram information via getWordProperty().
Bug: 12810574
Change-Id: I2750a5659ccbc3e31307c28e35dd9a1dbdffbeee
2014-02-03 11:47:26 +09:00
Keisuke Kuroyanagi 94d9a2309f Extend jni interface to get bigrams in WordProperty.
Bug: 12810574
Change-Id: Ia4b88d02ea8790a5c47d32376cc0b84c3e071ddd
2014-01-31 20:32:44 +09:00
Keisuke Kuroyanagi 17e7768a3e Fetch bigram information in getWordProperty().
Bug: 12810574
Change-Id: Ic67e964b01c2753f2eec82b079dc0e5650f340bb
2014-01-31 20:24:23 +09:00
Keisuke Kuroyanagi 20b7786ece Add BigramProperty and ShortcutProperty in WordProperty.
Bug: 12810574
Change-Id: Ia9e497c49bb2cf10897bae3a79317113325db819
2014-01-31 15:26:43 +09:00
Jean Chalard b0df28f4cc Merge "[HD02] Make the date header attribute mandatory." 2014-01-31 03:58:55 +00:00
Keisuke Kuroyanagi 748f32feba Merge "Rename UnigramProperty to WordProperty." 2014-01-31 03:06:44 +00:00
Keisuke Kuroyanagi e9085da8cb Merge "Simplify UnigramProperty." 2014-01-31 03:06:33 +00:00
Keisuke Kuroyanagi 080a35e959 Rename UnigramProperty to WordProperty.
Bug: 12810574
Change-Id: If5ddd803948aaf6e491ddcbaa5436fb3af3f7257
2014-01-31 11:06:42 +09:00
Jean Chalard a137c21c9c [HD02] Make the date header attribute mandatory.
Bug: 11281748
Change-Id: I48a6f3d95ca89ced0c51335527201ecabfb7998e
2014-01-30 18:19:21 +09:00
Keisuke Kuroyanagi 5f7f6a1615 Simplify UnigramProperty.
Bug: 12810574
Change-Id: I606f559b5dd340b7525280ecead2c2c7ec920a78
2014-01-30 17:10:23 +09:00
Jean Chalard 324adefca3 [HD01] Small initial refactoring
Bug: 11281748
Change-Id: I3c64c75ecd17729bdacf03048311d76ca9f37ae2
2014-01-30 17:08:13 +09:00
Keisuke Kuroyanagi 21e2380234 Improve the double letter word promotion logic.
The old logic only checks the top 3 suggestions before the
final adjustment. It leads to instability in scores.

Bug: 10700674
Change-Id: I986aed2aefd66c1fba6196a8f100fcb5bc838a38
2014-01-28 16:33:20 +09:00
Keisuke Kuroyanagi 51c5ec10f9 Add length check for dict update operations.
Bug: 12602903
Change-Id: I6835dee8bf8b356f0f1cf6c0531bee5b3415a13f
2014-01-24 16:32:42 -08:00
Keisuke Kuroyanagi 3e0777e726 Use memmove for all place src and dest can be same.
Change-Id: I03e5c8f95cb22c7096ff545f1181f914b80a6c58
2014-01-20 11:27:01 -08:00
Keisuke Kuroyanagi d78a447d10 Add a dedicated method to control time in native code.
Bug: 12443085
Change-Id: I41a5cf6c895cb59e54af98b40efded36afcd3247
2014-01-15 14:55:57 -08:00
Tadashi G. Takaoka 20a8a4a3d9 Fix native method signature
Change-Id: Ia2d566f162f746c6b2b2b874f7278d86abd9e839
2014-01-15 18:56:40 +09:00
Keisuke Kuroyanagi 9638693479 Fix native crash in BigramDictionary::addWordBigram()
Bug: 12479653
Change-Id: I7d263f744f711202000524a47769dc2d3d0e0de6
2014-01-14 18:51:49 -08:00
Keisuke Kuroyanagi 84c52a3c81 Boundary check for ProximityInfoState::getProximityTypeG().
Bug: 12481052
Change-Id: Ic0b64d2fa17eb8f65f3ad1a7b468cdf6dddf7e70
2014-01-10 20:15:36 +09:00
Keisuke Kuroyanagi 05113c1847 Check whether the word is OOV or not for amanatto update.
Bug: 11736529
Change-Id: I27973212da7e922b2185dbb4febfb5bb4dfbdcbb
2014-01-08 21:02:35 +09:00
Keisuke Kuroyanagi b68dd6cd0b Merge "Differentiate exact matches' minor errors." 2013-12-19 04:14:11 +00:00
Keisuke Kuroyanagi afa6c07d30 Differentiate exact matches' minor errors.
Bug: 8844931
Change-Id: I06754643188ccd28ba8c873878f7761787d91ff6
2013-12-19 12:37:56 +09:00
Ken Wakasa abe3a35abf Add C++ compiler warning options
Change-Id: I0d79d450c5cca887d02a1087128c1d3eb0bf0b7f
2013-12-19 00:12:01 +09:00
Keisuke Kuroyanagi d8f35f7b4c Move methods for outputting from Suggest.
Bug: 8844931
Change-Id: I83dde6c37b75ed1e1ae4e0972e169d30ada8faf5
2013-12-18 16:47:23 +09:00
Keisuke Kuroyanagi 9ee9095528 Move methods for scoring from Traversal to Scoring.
Bug: 8844931

Change-Id: I6137d06e597f4e7b81aaf29555199e18984f2c39
2013-12-18 07:13:44 +00:00
Keisuke Kuroyanagi d5f5dd74c2 Quit suggest words that do not include proximity chars.
Bug: 10416722

Change-Id: I74d27c6a22b4ca47d02eb238b2fbcc49f5dee99c
2013-12-18 14:31:10 +09:00
Keisuke Kuroyanagi b5215d456a Increment version number of ver4 format to 401.
Bug: 11073222
Change-Id: I3878aef02040b1ef73165645fc70ede586289c51
2013-12-16 23:02:54 +09:00
Keisuke Kuroyanagi 9b08a9e611 Fix: incompatibility of sparse table.
Bug: 11073222
Change-Id: Id17178efd8f9f3828d821cc718d35606e8bd4b1a
2013-12-16 22:46:29 +09:00
Keisuke Kuroyanagi c1163c8518 Align block sizes for ver4 format to java side.
Bug: 11073222
Change-Id: I09646cbc81529576d5e34dcfa7bc29083a64076f
2013-12-16 20:42:32 +09:00
Jean Chalard a245d15da5 Have dicttool use the native library to generate v4 dicts.
Yay !

Change-Id: Iea8ced9e81031b9ab7eff05ad9ef7215be248de9
2013-12-13 18:18:20 +09:00
Ken Wakasa 2fa3693c26 Reset to 9bd6dac470
The bulk merge from -bayo to klp-dev should not have been merged to master.

Change-Id: I527a03a76f5247e4939a672f27c314dc11cbb854
2013-12-13 17:13:32 +09:00
Keisuke Kuroyanagi 6bc5acaa79 Fix: Suggested words from user history are invalid.
- Suggestions form user history can contain invalid words.
- isValidWord always returns false.

Bug: 11139426
Change-Id: I6075b275603332ddb00f4a9284afcaa82d824270
2013-10-22 05:23:48 -07:00
Keisuke Kuroyanagi 6d91e4ce74 Fix offdevice test build.
Change-Id: I42143bdc4f6f5aa4f55530de84819ab722a22f72
2013-10-08 16:05:42 +09:00
Keisuke Kuroyanagi 6d3d53955e Merge "Fix a bug: wrong parent position is written when moving." 2013-10-08 04:48:53 +00:00
Keisuke Kuroyanagi 802e6383ea Fix a bug: wrong parent position is written when moving.
Change-Id: I93d33ba9105eb85327108cca5706d8db3399de6a
2013-10-08 12:44:00 +09:00
Jean Chalard 86f7711407 Merge "Don't auto-correct to non-whitelist shortcuts" 2013-10-07 10:04:18 +00:00
Jean Chalard f309f2288b Don't auto-correct to non-whitelist shortcuts
Bug: 11076722
Change-Id: I0de10ab24d33ec31f45664fe92d80d5a7bec9c20
2013-10-07 19:01:57 +09:00
Keisuke Kuroyanagi cfb018ba6d Fix bug and Add large test for decaying dictionary.
- GC gets failure when the dictionary become empty.
- Useless unigrams are sometimes not removed.

Bug: 10197478
Change-Id: I8d1479c01efba61a81f03bc077da6bcb4797a940
2013-10-07 17:31:01 +09:00
Keisuke Kuroyanagi 15b1a0fd14 Parameter optimization.
[Category diff]
+1     194
-1      21
+2      15
-2       1
+3       1
-3      15
+4     152
-4      90
+5     108
-5     197
+6      94
-6     136
+7      28
-7     132

[Weighted category diff]
+1     230
-1      23
+2      15
-2       1
+3       1
-3      15
+4     180
-4     106
+5     122
-5     224
+6     107
-6     158
+7      29
-7     157

show diff for ./en_user_log_phones_2011_08.csv
+1     171
-1       9
+2       7
-2       1
+3       1
-3       7
+4     123
-4      76
+5      84
-5     157
+6      66
-6     119
+7      20
-7     103

Before
---- count ----
Count(1):75.64%, 177913
Count(2):1.06%, 2510
Count(3):0.56%, 1319
Count(4):6.63%, 15612
Count(5):9.87%, 23223
Count(6):2.79%, 6580
Count(7):3.41%, 8038
---- weight ----
Weight(1):96.88%, 2009907
Weight(2):0.24%, 5157
Weight(3):0.09%, 2025
Weight(4):0.77%, 16095
Weight(5):1.11%, 23077
Weight(6):0.39%, 8172
Weight(7):0.48%, 10146
---- score ----
Conservative score = 66.406
Normal score = 81.512
Agressive score = 89.064
Insane score = 92.841

After
---- count ----
Count(1):75.71%, 178087
Count(2):1.07%, 2524
Count(3):0.55%, 1305
Count(4):6.66%, 15677
Count(5):9.83%, 23133
Count(6):2.78%, 6539
Count(7):3.37%, 7930
---- weight ----
Weight(1):96.89%, 2010118
Weight(2):0.24%, 5171
Weight(3):0.09%, 2011
Weight(4):0.78%, 16184
Weight(5):1.1%, 22972
Weight(6):0.39%, 8109
Weight(7):0.48%, 10014
---- score ----
Conservative score = 66.431
Normal score = 81.531
Agressive score = 89.08
Insane score = 92.855

No diffs for 10+ freq user log.

Change-Id: Ica4a7604507a2c91aaa7daf5bacf4fce4039f547
2013-10-07 12:39:51 +09:00
Keisuke Kuroyanagi 3e5f4b53b0 Merge "Check the elapsed time for decaying." 2013-10-04 07:16:17 +00:00
Keisuke Kuroyanagi 67c855ea6f Check the elapsed time for decaying.
Bug: 6669677
Change-Id: I0401a4c0908c702ce65abfa5e017dd16cb1296f9
2013-10-04 16:11:11 +09:00
Xiaojun Bi 0b62621446 Merge "Optimize the parameters for the typing algorithm" 2013-10-04 03:25:43 +00:00
Xiaojun Bi 2ed09c214a Optimize the parameters for the typing algorithm
The new set of parameters are optimized for both
keystroke savings and regression test scores.

This change is corresponding to the patch 6 of
Change-Id: Ib8439a85de26cbbba58837344717f0acbfcf0ab0
in the regression test repository.

Here is the link to the doc analyzing the
“nnow → know” case:
https://docs.google.com/a/google.com/document/d/1XIHfAHLlP1f3F8R-u7yxPGWcP58tokygzN_lQKXFFgM/edit

Here is the link to the spreadsheet showing
the diff of the words with frequencies >= 10
in en_user_log_phones_2011_08.csv.
https://docs.google.com/a/google.com/spreadsheet/ccc?key=0Am_RQlV4zC6SdFFVTlhJT0RLUENCWWZOZ3JQTUpMQ2c&usp=drive_web#gid=0

Updated TRT Results for this change:
[Category diff]
+1     783
-1     717
+2      20
-2      34
+3      34
-3      20
+4     575
-4     777
+5     765
-5     755
+6     785
-6     666
+7     479
-7     472

[Weighted category diff]
+1     999
-1     965
+2      21
-2      35
+3      35
-3      21
+4     804
-4     980
+5     970
-5     971
+6     954
-6     831
+7     645
-7     625

Bug: 10613502

Change-Id: I5cd447673a196ad5af641a2118602bec879bdefb
2013-10-03 19:30:26 -07:00
Keisuke Kuroyanagi 864ee881bd Merge "Fix UserHistoryDictionaryTests." 2013-10-03 12:02:34 +00:00
Keisuke Kuroyanagi 11f7cae094 Fix UserHistoryDictionaryTests.
Bug: 6669677
Bug: 10667710

Change-Id: I6cdc6a6c9cacc7f276fda3a26ec31e3eb928471c
2013-10-03 20:55:34 +09:00
Keisuke Kuroyanagi 9cb1e78a44 Fix: Native crash when an empty attribute is written.
Bug: 10964805
Change-Id: Ia3cd891d20aee2226092f0f646d8f280901786fc
2013-10-03 17:30:54 +09:00
Keisuke Kuroyanagi 3d92954120 Fix build.
Change-Id: Ib33b27214e67378646e44f5b45bda298f960524f
2013-10-02 23:02:42 +09:00
Keisuke Kuroyanagi 7c4dcf1e91 Use probability table for decaying dictionaries.
Bug: 6669677
Change-Id: Ib5994a2e343dfcdaf23755e523f52846709b1c6c
2013-10-02 18:01:47 +09:00
Keisuke Kuroyanagi c76bbceedc Stochastic decay.
Bug: 6669677
Change-Id: Ib2d9228b951c77dab7a8675ce9db60677e87e771
2013-10-02 17:51:12 +09:00