@echogarden/text-segmentation 0.7.0 → 0.8.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/exports/Exports.d.ts +3 -0
- package/dist/exports/Exports.d.ts.map +1 -0
- package/dist/exports/Exports.js +3 -0
- package/dist/exports/Exports.js.map +1 -0
- package/dist/{exports → segmentation}/TextSegmentation.d.ts +0 -1
- package/dist/segmentation/TextSegmentation.d.ts.map +1 -0
- package/dist/segmentation/TextSegmentation.js +575 -0
- package/dist/segmentation/TextSegmentation.js.map +1 -0
- package/dist/segmentation/WordSequence.d.ts.map +1 -0
- package/dist/segmentation/WordSequence.js.map +1 -0
- package/dist/tests/Test.js +2 -2
- package/dist/tests/Test.js.map +1 -1
- package/dist/utilities/Timer.d.ts +7 -3
- package/dist/utilities/Timer.d.ts.map +1 -1
- package/dist/utilities/Timer.js +41 -39
- package/dist/utilities/Timer.js.map +1 -1
- package/package.json +4 -4
- package/src/exports/Exports.ts +2 -0
- package/src/segmentation/TextSegmentation.ts +743 -0
- package/src/tests/Test.ts +2 -2
- package/src/utilities/Timer.ts +50 -48
- package/dist/Test.d.ts +0 -2
- package/dist/Test.d.ts.map +0 -1
- package/dist/Test.js +0 -120
- package/dist/Test.js.map +0 -1
- package/dist/exports/TextSegmentation.d.ts.map +0 -1
- package/dist/exports/TextSegmentation.js +0 -369
- package/dist/exports/TextSegmentation.js.map +0 -1
- package/dist/exports/WordSequence.d.ts.map +0 -1
- package/dist/exports/WordSequence.js.map +0 -1
- package/src/exports/TextSegmentation.ts +0 -527
- /package/dist/{exports → segmentation}/WordSequence.d.ts +0 -0
- /package/dist/{exports → segmentation}/WordSequence.js +0 -0
- /package/src/{exports → segmentation}/WordSequence.ts +0 -0
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"TextSegmentation.js","sourceRoot":"","sources":["../../src/exports/TextSegmentation.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,wBAAwB,IAAI,wBAAwB,EAAE,qBAAqB,EAAE,mDAAmD,EAAE,gCAAgC,EAAE,uBAAuB,EAAE,yBAAyB,EAA0D,MAAM,yBAAyB,CAAA;AACxT,OAAO,EAAE,gBAAgB,EAAE,sBAAsB,EAAE,wCAAwC,EAAE,gBAAgB,EAAE,eAAe,EAAE,MAAM,6BAA6B,CAAA;AACnK,OAAO,EAAE,yBAAyB,EAAE,MAAM,2CAA2C,CAAA;AACrF,OAAO,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAA;AAChD,OAAO,EAAE,oBAAoB,EAAE,MAAM,2BAA2B,CAAA;AAChE,OAAO,EAAE,YAAY,EAAkB,MAAM,mBAAmB,CAAA;AAEhE,OAAO,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAA;AAE7C,gGAAgG;AAChG,mBAAmB;AACnB,gGAAgG;AAChG,MAAM,CAAC,KAAK,UAAU,WAAW,CAAC,IAAY,EAAE,OAA6B;IAC5E,MAAM,YAAY,GAAG,MAAM,YAAY,CAAC,IAAI,EAAE,OAAO,CAAC,CAAA;IAEtD,OAAO,mBAAmB,CAAC,YAAY,CAAC,CAAA;AACzC,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,mBAAmB,CAAC,YAA0B,EAAE,OAAyC;IAC9G,OAAO,GAAG,EAAE,GAAG,sCAAsC,EAAE,GAAG,OAAO,EAAE,CAAA;IAEnE,MAAM,iBAAiB,GAAY,EAAE,CAAA;IAErC,CAAC;QACA,IAAI,sBAAsB,GAAG,CAAC,CAAA;QAC9B,IAAI,qBAAqB,GAAG,KAAK,CAAA;QACjC,IAAI,2BAA2B,GAAG,KAAK,CAAA;QAEvC,KAAK,IAAI,SAAS,GAAG,CAAC,EAAE,SAAS,GAAG,YAAY,CAAC,MAAM,EAAE,SAAS,EAAE,EAAE,CAAC;YACtE,MAAM,IAAI,GAAG,YAAY,CAAC,SAAS,CAAC,SAAS,CAAC,CAAA;YAE9C,IAAI,CAAC,2BAA2B,IAAI,qBAAqB,IAAI,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC;gBAClF,2BAA2B,GAAG,IAAI,CAAA;YACnC,CAAC;YAED,IAAI,2BAA2B,EAAE,CAAC;gBACjC,MAAM,QAAQ,GAAG,SAAS,GAAG,YAAY,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,YAAY,CAAC,SAAS,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAA;gBAEjG,IAAI,QAAQ,KAAK,EAAE,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;oBAChE,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,sBAAsB,EAAE,GAAG,EAAE,SAAS,GAAG,CAAC,EAAE,CAAC,CAAA;oBAC7E,sBAAsB,GAAG,SAAS,GAAG,CAAC,CAAA;oBAEtC,2BAA2B,GAAG,KAAK,CAAA;gBACpC,CAAC;YACF,CAAC;YAED,IAAI,qBAAqB,KAAK,KAAK,EAAE,CAAC;gBACrC,qBAAqB,GAAG,CAAC,uBAAuB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;YAC5D,CAAC;QACF,CAAC;QAED,IAAI,sBAAsB,GAAG,YAAY,CAAC,MAAM,EAAE,CAAC;YAClD,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,sBAAsB,EAAE,GAAG,EAAE,YAAY,CAAC,MAAM,EAAE,CAAC,CAAA;QACpF,CAAC;IACF,CAAC;IAED,MAAM,kBAAkB,GAAY,EAAE,CAAA;IACtC,MAAM,qBAAqB,GAAY,EAAE,CAAA;IAEzC,CAAC;QACA,MAAM,0BAA0B,GAAG,CAAC,CAAA;QAEpC,KAAK,IAAI,YAAY,GAAG,CAAC,EAAE,YAAY,GAAG,iBAAiB,CAAC,MAAM,EAAE,YAAY,EAAE,EAAE,CAAC;YACpF,MAAM,gBAAgB,GAAG,iBAAiB,CAAC,YAAY,CAAC,CAAA;YAExD,MAAM,qBAAqB,GAAG,gBAAgB,CAAC,KAAK,CAAA;YACpD,MAAM,mBAAmB,GAAG,gBAAgB,CAAC,GAAG,CAAA;YAEhD,qBAAqB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,kBAAkB,CAAC,MAAM,EAAE,GAAG,EAAE,kBAAkB,CAAC,MAAM,EAAE,CAAC,CAAA;YAEhG,IAAI,qBAAqB,GAAG,KAAK,CAAA;YACjC,IAAI,uBAAuB,GAAG,qBAAqB,CAAA;YACnD,IAAI,0BAA0B,GAAG,CAAC,CAAA;YAElC,KAAK,IAAI,SAAS,GAAG,qBAAqB,EAAE,SAAS,GAAG,mBAAmB,EAAE,SAAS,EAAE,EAAE,CAAC;gBAC1F,MAAM,IAAI,GAAG,YAAY,CAAC,SAAS,CAAC,SAAS,CAAC,CAAA;gBAE9C,IAAI,qBAAqB,KAAK,KAAK,EAAE,CAAC;oBACrC,qBAAqB,GAAG,CAAC,uBAAuB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;gBAC5D,CAAC;gBAED,IAAI,0BAA0B,GAAG,0BAA0B,EAAE,CAAC;oBAC7D,MAAM,OAAO,GAAG,IAAI,CAAC,QAAQ,CAAC,yBAAyB,CAAC,CAAA;oBAExD,KAAK,MAAM,CAAC,IAAI,OAAO,EAAE,CAAC;wBACzB,0BAA0B,IAAI,CAAC,CAAA;wBAE/B,IAAI,0BAA0B,IAAI,0BAA0B,EAAE,CAAC;4BAC9D,MAAK;wBACN,CAAC;oBACF,CAAC;gBACF,CAAC;gBAED,IAAI,qBAAqB,IAAI,0BAA0B,IAAI,0BAA0B,IAAI,gCAAgC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;oBACtI,IAAI,wBAAwB,GAAG,SAAS,CAAA;oBAExC,OAAO,wBAAwB,GAAG,mBAAmB,EAAE,CAAC;wBACvD,MAAM,YAAY,GAAG,YAAY,CAAC,SAAS,CAAC,wBAAwB,CAAC,CAAA;wBAErE,IAAI,mDAAmD,CAAC,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;4BAC5E,wBAAwB,EAAE,CAAA;wBAC3B,CAAC;6BAAM,CAAC;4BACP,MAAK;wBACN,CAAC;oBACF,CAAC;oBAED,kBAAkB,CAAC,IAAI,CAAC;wBACvB,KAAK,EAAE,uBAAuB;wBAC9B,GAAG,EAAE,wBAAwB;qBAC7B,CAAC,CAAA;oBAEF,qBAAqB,CAAC,qBAAqB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,CAAA;oBAEhE,uBAAuB,GAAG,wBAAwB,CAAA;oBAClD,0BAA0B,GAAG,CAAC,CAAA;oBAE9B,SAAS,GAAG,wBAAwB,GAAG,CAAC,CAAA;gBACzC,CAAC;YACF,CAAC;YAED,IAAI,uBAAuB,GAAG,mBAAmB,EAAE,CAAC;gBACnD,kBAAkB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,uBAAuB,EAAE,GAAG,EAAE,mBAAmB,EAAE,CAAC,CAAA;gBACrF,qBAAqB,CAAC,qBAAqB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,CAAA;YACjE,CAAC;QACF,CAAC;IACF,CAAC;IAED,MAAM,SAAS,GAAe,EAAE,CAAA;IAEhC,KAAK,MAAM,SAAS,IAAI,kBAAkB,EAAE,CAAC;QAC5C,MAAM,oBAAoB,GAAG,YAAY,CAAC,KAAK,CAAC,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,GAAG,CAAC,CAAA;QAE/E,SAAS,CAAC,IAAI,CAAC,IAAI,QAAQ,CAAC,SAAS,EAAE,oBAAoB,CAAC,CAAC,CAAA;IAC9D,CAAC;IAED,KAAK,MAAM,QAAQ,IAAI,SAAS,EAAE,CAAC;QAClC,MAAM,gBAAgB,GAAY,EAAE,CAAA;QAEpC,IAAI,qBAAqB,GAAG,QAAQ,CAAC,SAAS,CAAC,GAAG,CAAA;QAClD,IAAI,qBAAqB,GAAG,QAAQ,CAAC,SAAS,CAAC,KAAK,CAAA;QAEpD,KAAK,IAAI,SAAS,GAAG,qBAAqB,EAAE,SAAS,GAAG,qBAAqB,EAAE,SAAS,EAAE,EAAE,CAAC;YAC5F,MAAM,WAAW,GAAG,YAAY,CAAC,SAAS,CAAC,SAAS,CAAC,CAAA;YAErD,MAAM,4BAA4B,GACjC,qBAAqB,CAAC,IAAI,CAAC,WAAW,CAAC;gBACvC,CAAC,CAAC,OAAO,CAAC,mCAAmC;oBAC5C,CAAC,WAAW,KAAK,GAAG,IAAI,WAAW,KAAK,GAAG,CAAC;oBAC5C,uBAAuB,CAAC,IAAI,CAAC,YAAY,CAAC,SAAS,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,CAAA;YAEtE,IAAI,4BAA4B,EAAE,CAAC;gBAClC,IAAI,kBAAkB,GAAG,KAAK,CAAA;gBAE9B,OAAO,SAAS,GAAG,qBAAqB,GAAG,CAAC,EAAE,CAAC;oBAC9C,MAAM,QAAQ,GAAG,YAAY,CAAC,SAAS,CAAC,SAAS,GAAG,CAAC,CAAC,CAAA;oBAEtD,IAAI,CAAC,mDAAmD,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;wBACzE,MAAK;oBACN,CAAC;oBAED,IAAI,uBAAuB,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;wBAC5C,kBAAkB,GAAG,IAAI,CAAA;oBAC1B,CAAC;oBAED,IAAI,QAAQ,KAAK,GAAG,IAAI,kBAAkB,EAAE,CAAC;wBAC5C,MAAK;oBACN,CAAC;oBAED,SAAS,IAAI,CAAC,CAAA;gBACf,CAAC;gBAED,gBAAgB,CAAC,IAAI,CAAC;oBACrB,KAAK,EAAE,qBAAqB;oBAC5B,GAAG,EAAE,SAAS,GAAG,CAAC;iBAClB,CAAC,CAAA;gBAEF,qBAAqB,GAAG,SAAS,GAAG,CAAC,CAAA;YACtC,CAAC;QACF,CAAC;QAED,IAAI,qBAAqB,GAAG,qBAAqB,EAAE,CAAC;YACnD,gBAAgB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,qBAAqB,EAAE,GAAG,EAAE,qBAAqB,EAAE,CAAC,CAAA;QACpF,CAAC;QAED,KAAK,MAAM,SAAS,IAAI,gBAAgB,EAAE,CAAC;YAC1C,MAAM,kBAAkB,GAAG,YAAY,CAAC,KAAK,CAAC,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,GAAG,CAAC,CAAA;YAE7E,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,IAAI,MAAM,CAAC,SAAS,EAAE,kBAAkB,CAAC,CAAC,CAAA;QACjE,CAAC;IACF,CAAC;IAED,MAAM,MAAM,GAAuB;QAClC,KAAK,EAAE,YAAY;QACnB,SAAS;QACT,qBAAqB;KACrB,CAAA;IAED,OAAO,MAAM,CAAA;AACd,CAAC;AAED,MAAM,yBAAyB,GAAG,IAAI,GAAG,EAAkB,CAAA;AAE3D,MAAM,CAAC,KAAK,UAAU,YAAY,CAAC,IAAY,EAAE,OAA6B;IAC7E,IAAI,CAAC,OAAO,EAAE,CAAC;QACd,OAAO,GAAG,EAAE,CAAA;IACb,CAAC;IAED,OAAO,GAAG,EAAE,GAAG,0BAA0B,EAAE,GAAG,OAAO,EAAE,CAAA;IAEvD,IAAI,OAAO,CAAC,QAAQ,EAAE,CAAC;QACtB,OAAO,CAAC,QAAQ,GAAG,oBAAoB,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAA;IAC1D,CAAC;IAED,MAAM,aAAa,GAAG,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAA;IAE7C,IAAI,kBAAkB,GAAG,yBAAyB,CAAC,GAAG,CAAC,aAAa,CAAC,CAAA;IAErE,IAAI,CAAC,kBAAkB,EAAE,CAAC;QACzB,kBAAkB,GAAG,iCAAiC,CAAC,OAAO,CAAC,CAAA;QAE/D,yBAAyB,CAAC,GAAG,CAAC,aAAa,EAAE,kBAAkB,CAAC,CAAA;IACjE,CAAC;IAED,IAAI,YAAY,GAAG,IAAI,YAAY,EAAE,CAAA;IAErC,SAAS,0BAA0B,CAAC,WAAmB,EAAE,SAAiB;QACzE,MAAM,wBAAwB,GAAG,IAAI,CAAC,SAAS,CAAC,WAAW,EAAE,SAAS,CAAC,CAAA;QAEvE,IAAI,UAAU,GAAG,WAAW,CAAA;QAC5B,IAAI,0BAA0B,GAAG,WAAW,CAAA;QAE5C,SAAS,0BAA0B;YAClC,IAAI,UAAU,GAAG,0BAA0B,EAAE,CAAC;gBAC7C,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,0BAA0B,EAAE,UAAU,CAAC,CAAA;gBACvE,YAAY,CAAC,OAAO,CAAC,QAAQ,EAAE,0BAA0B,EAAE,IAAI,CAAC,CAAA;gBAEhE,0BAA0B,GAAG,UAAU,CAAA;YACxC,CAAC;QACF,CAAC;QAED,KAAK,MAAM,IAAI,IAAI,wBAAwB,EAAE,CAAC;YAC7C,IAAI,IAAI,KAAK,GAAG,EAAE,CAAC;gBAClB,UAAU,IAAI,CAAC,CAAA;gBAEf,SAAQ;YACT,CAAC;YAED,0BAA0B,EAAE,CAAA;YAE5B,UAAU,IAAI,IAAI,CAAC,MAAM,CAAA;YAEzB,0BAA0B,EAAE,CAAA;QAC7B,CAAC;QAED,0BAA0B,EAAE,CAAA;IAC7B,CAAC;IAED,MAAM,WAAW,GAAG,IAAI,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CAAA;IAErD,IAAI,kBAAkB,GAAG,CAAC,CAAA;IAE1B,IAAI,WAAW,EAAE,CAAC;QACjB,KAAK,MAAM,KAAK,IAAI,WAAW,EAAE,CAAC;YACjC,MAAM,OAAO,GAAG,KAAK,CAAC,OAAQ,CAAC,CAAC,CAAE,CAAA;YAClC,MAAM,gBAAgB,GAAG,OAAO,CAAC,CAAC,CAAC,CAAA;YACnC,MAAM,cAAc,GAAG,OAAO,CAAC,CAAC,CAAC,CAAA;YAEjC,IAAI,gBAAgB,GAAG,kBAAkB,EAAE,CAAC;gBAC3C,0BAA0B,CAAC,kBAAkB,EAAE,gBAAgB,CAAC,CAAA;YACjE,CAAC;YAED,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,cAAc,CAAC,CAAA;YACjE,YAAY,CAAC,OAAO,CAAC,QAAQ,EAAE,gBAAgB,EAAE,KAAK,CAAC,CAAA;YAEvD,kBAAkB,GAAG,cAAc,CAAA;QACpC,CAAC;QAED,0BAA0B,CAAC,kBAAkB,EAAE,IAAI,CAAC,MAAM,CAAC,CAAA;IAC5D,CAAC;IAED,IAAI,OAAO,CAAC,6BAA6B,EAAE,CAAC;QAC3C,YAAY,GAAG,MAAM,yBAAyB,CAAC,IAAI,EAAE,YAAY,CAAC,CAAA;IACnE,CAAC;IAED,OAAO,YAAY,CAAA;AACpB,CAAC;AAED,KAAK,UAAU,yBAAyB,CAAC,cAAsB,EAAE,YAA0B;IAC1F,MAAM,eAAe,GAAG,MAAM,kBAAkB,EAAE,CAAA;IAElD,IAAI,eAAe,KAAK,SAAS,EAAE,CAAC;QACnC,OAAO,YAAY,CAAA;IACpB,CAAC;IAED,IAAI,cAAc,GAAG,KAAK,CAAA;IAE1B,MAAM,eAAe,GAAG,IAAI,YAAY,EAAE,CAAA;IAE1C,KAAK,IAAI,SAAS,GAAG,CAAC,EAAE,SAAS,GAAG,YAAY,CAAC,MAAM,EAAE,SAAS,EAAE,EAAE,CAAC;QACtE,MAAM,SAAS,GAAG,YAAY,CAAC,OAAO,CAAC,SAAS,CAAC,CAAA;QACjD,MAAM,eAAe,GAAG,SAAS,CAAC,WAAW,CAAA;QAE7C,MAAM,IAAI,GAAG,YAAY,CAAC,SAAS,CAAC,SAAS,CAAC,CAAA;QAE9C,IAAI,yBAAyB,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;YAC1C,IAAI,CAAC,cAAc,EAAE,CAAC;gBACrB,MAAM,eAAe,CAAC,UAAU,EAAE,CAAA;gBAElC,cAAc,GAAG,IAAI,CAAA;YACtB,CAAC;YAED,MAAM,UAAU,GAAG,CAAC,GAAG,eAAe,CAAC,uBAAuB,CAAC,IAAI,CAAC,CAAC,CAAA;YAErE,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;gBAChD,MAAM,kBAAkB,GAAG,eAAe,GAAG,UAAU,CAAC,CAAC,CAAC,CAAA;gBAC1D,MAAM,gBAAgB,GAAG,eAAe,GAAG,UAAU,CAAC,CAAC,GAAG,CAAC,CAAC,CAAA;gBAE5D,MAAM,WAAW,GAAG,cAAc,CAAC,SAAS,CAAC,kBAAkB,EAAE,gBAAgB,CAAC,CAAA;gBAElF,eAAe,CAAC,OAAO,CACtB,WAAW,EACX,kBAAkB,EAClB,KAAK,CACL,CAAA;YACF,CAAC;QACF,CAAC;aAAM,CAAC;YACP,MAAM,QAAQ,GAAG,cAAc,CAAC,SAAS,CAAC,SAAS,CAAC,WAAW,EAAE,SAAS,CAAC,SAAS,CAAC,CAAA;YAErF,eAAe,CAAC,OAAO,CAAC,QAAQ,EAAE,SAAS,CAAC,WAAW,EAAE,SAAS,CAAC,aAAa,CAAC,CAAA;QAClF,CAAC;IACF,CAAC;IAED,OAAO,eAAe,CAAA;AACvB,CAAC;AAED,uDAAuD;AACvD,MAAM,UAAU,wCAAwC,CAAC,YAA0B,EAAE,UAAkB;IACtG,MAAM,2BAA2B,GAAG,IAAI,GAAG,EAAkB,CAAA;IAE7D,MAAM,2BAA2B,GAAG,IAAI,YAAY,EAAE,CAAA;IAEtD,SAAS,0BAA0B,CAAC,SAAiB,EAAE,iBAAyB;QAC/E,IAAI,UAAU,GAAG,iBAAiB,CAAA;QAElC,kFAAkF;QAClF,KAAK,MAAM,IAAI,IAAI,SAAS,EAAE,CAAC;YAC9B,MAAM,aAAa,GAAG,UAAU,GAAG,IAAI,CAAC,MAAM,CAAA;YAE9C,MAAM,SAAS,GAAG,2BAA2B,CAAC,SAAS,CAAA;YAEvD,IAAI,IAAI,KAAK,GAAG,IAAI,SAAS,IAAI,SAAS,CAAC,aAAa,IAAI,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK,GAAG,EAAE,CAAC;gBACvF,2BAA2B,CAAC,SAAS,CAAC,IAAI,IAAI,GAAG,CAAA;gBACjD,2BAA2B,CAAC,SAAS,CAAC,SAAS,GAAG,aAAa,CAAA;YAChE,CAAC;iBAAM,CAAC;gBACP,MAAM,QAAQ,GAAG,UAAU,CAAC,SAAS,CAAC,UAAU,EAAE,aAAa,CAAC,CAAA;gBAEhE,2BAA2B,CAAC,OAAO,CAAC,QAAQ,EAAE,UAAU,EAAE,IAAI,CAAC,CAAA;YAChE,CAAC;YAED,UAAU,GAAG,aAAa,CAAA;QAC3B,CAAC;IACF,CAAC;IAED,KAAK,IAAI,SAAS,GAAG,CAAC,EAAE,SAAS,GAAG,YAAY,CAAC,MAAM,EAAE,SAAS,EAAE,EAAE,CAAC;QACtE,MAAM,SAAS,GAAG,YAAY,CAAC,UAAU,CAAC,SAAS,CAAC,CAAA;QACpD,MAAM,eAAe,GAAG,SAAS,CAAC,WAAW,CAAA;QAE7C,MAAM,qBAAqB,GAAG,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,YAAY,CAAC,OAAO,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,CAAA;QAE/F,8FAA8F;QAC9F,IAAI,qBAAqB,KAAK,eAAe,EAAE,CAAC;YAC/C,MAAM,SAAS,GAAG,UAAU,CAAC,SAAS,CAAC,qBAAqB,EAAE,eAAe,CAAC,CAAA;YAE9E,0BAA0B,CAAC,SAAS,EAAE,qBAAqB,CAAC,CAAA;QAC7D,CAAC;QAED,2BAA2B,CAAC,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAA;QACnD,2BAA2B,CAAC,GAAG,CAAC,2BAA2B,CAAC,MAAM,GAAG,CAAC,EAAE,SAAS,CAAC,CAAA;QAElF,oEAAoE;QACpE,IAAI,SAAS,KAAK,YAAY,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YAC3C,IAAI,UAAU,CAAC,MAAM,KAAK,SAAS,CAAC,SAAS,EAAE,CAAC;gBAC/C,MAAM,SAAS,GAAG,UAAU,CAAC,SAAS,CAAC,SAAS,CAAC,SAAS,EAAE,UAAU,CAAC,MAAM,CAAC,CAAA;gBAE9E,0BAA0B,CAAC,SAAS,EAAE,SAAS,CAAC,SAAS,CAAC,CAAA;YAC3D,CAAC;QACF,CAAC;IACF,CAAC;IAED,OAAO,EAAE,2BAA2B,EAAE,2BAA2B,EAAE,CAAA;AACpE,CAAC;AAED,SAAS,oBAAoB,CAAC,YAA0B,EAAE,IAAY;IACrE,MAAM,iBAAiB,GAAY,EAAE,CAAA;IACrC,MAAM,WAAW,GAAG,YAAY,CAAC,OAAO,CAAA;IAExC,IAAI,WAAW,CAAC,CAAC,CAAC,CAAC,WAAW,GAAG,CAAC,EAAE,CAAC;QACpC,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,CAAC,EAAE,GAAG,EAAE,WAAW,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,CAAC,CAAA;IACtE,CAAC;IAED,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,WAAW,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAC7C,MAAM,KAAK,GAAG,WAAW,CAAC,CAAC,CAAC,CAAA;QAE5B,MAAM,iBAAiB,GAAG,WAAW,CAAC,CAAC,GAAG,CAAC,CAAC,EAAE,SAAS,IAAI,CAAC,CAAA;QAE5D,IAAI,KAAK,CAAC,WAAW,GAAG,iBAAiB,EAAE,CAAC;YAC3C,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,GAAG,EAAE,KAAK,CAAC,WAAW,EAAE,CAAC,CAAA;QAC7E,CAAC;QAED,IAAI,KAAK,CAAC,aAAa,EAAE,CAAC;YACzB,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,WAAW,EAAE,GAAG,EAAE,KAAK,CAAC,SAAS,EAAE,CAAC,CAAA;QAC3E,CAAC;IACF,CAAC;IAED,CAAC;QACA,MAAM,aAAa,GAAG,WAAW,CAAC,WAAW,CAAC,MAAM,GAAG,CAAC,CAAC,EAAE,SAAS,CAAA;QAEpE,IAAI,aAAa,IAAI,aAAa,GAAG,IAAI,CAAC,MAAM,EAAE,CAAC;YAClD,iBAAiB,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,aAAa,EAAE,GAAG,EAAE,IAAI,CAAC,MAAM,EAAE,CAAC,CAAA;QACnE,CAAC;IACF,CAAC;IAED,OAAO,iBAAiB,CAAA;AACzB,CAAC;AAED,gGAAgG;AAChG,iBAAiB;AACjB,gGAAgG;AAChG,SAAS,iCAAiC,CAAC,OAA4B;IACtE,MAAM,uBAAuB,GAAG,gBAAgB,CAAC,OAAO,CAAC,QAAQ,IAAI,EAAE,CAAC,IAAI,EAAE,CAAA;IAC9E,MAAM,2BAA2B,GAAG,sBAAsB,CAAC,OAAO,CAAC,QAAQ,IAAI,EAAE,CAAC,IAAI,EAAE,CAAA;IACxF,MAAM,8BAA8B,GAAG,wCAAwC,CAAC,OAAO,CAAC,QAAQ,IAAI,EAAE,CAAC,IAAI,EAAE,CAAA;IAC7G,MAAM,6CAA6C,GAAG,8BAA8B,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,UAAU,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAA;IACzH,MAAM,kBAAkB,GAAG,OAAO,CAAC,kBAAkB,IAAI,EAAE,CAAA;IAE3D,IAAI,YAAY,GAAG;QAClB,GAAG,kBAAkB;QACrB,GAAG,uBAAuB;QAC1B,GAAG,2BAA2B;QAC9B,GAAG,8BAA8B;QACjC,GAAG,6CAA6C;QAChD,GAAG,gBAAgB;QACnB,GAAG,eAAe;KAClB,CAAA;IAED,MAAM,WAAW,GAAG,wBAAwB,CAAC;QAC5C,GAAG,YAAY;QACf,GAAG,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,iBAAiB,EAAE,CAAC;QACrD,GAAG,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,iBAAiB,EAAE,CAAC;KACrD,CAAC,CAAA;IAEF,MAAM,kBAAkB,GAAG,WAAW,CAAC,WAAW,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAA;IAErE,OAAO,kBAAkB,CAAA;AAC1B,CAAC;AAED,KAAK,UAAU,kBAAkB;IAChC,IAAI,CAAC;QACJ,MAAM,eAAe,GAAG,MAAM,MAAM,CAAC,mCAAmC,CAAC,CAAA;QAEzE,OAAO,eAAe,CAAA;IACvB,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,SAAS,CAAA;IACjB,CAAC;AACF,CAAC;AAWD,MAAM,OAAO,YAAY;IACxB,SAAS,CAAO;IAChB,KAAK,CAAc;IAEnB,YAAY,SAAgB,EAAE,KAAmB;QAChD,IAAI,CAAC,SAAS,GAAG,SAAS,CAAA;QAC1B,IAAI,CAAC,KAAK,GAAG,KAAK,CAAA;IACnB,CAAC;IAED,IAAI,IAAI;QACP,OAAO,IAAI,CAAC,KAAK,CAAC,IAAI,CAAA;IACvB,CAAC;IAED,IAAI,SAAS;QACZ,OAAO;YACN,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,UAAU,CAAC,WAAW;YACxC,GAAG,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,SAAS;SACnC,CAAA;IACF,CAAC;CACD;AAED,MAAM,OAAO,QAAS,SAAQ,YAAY;IACzC,OAAO,GAAa,EAAE,CAAA;CACtB;AAED,MAAM,OAAO,MAAO,SAAQ,YAAY;CACvC;AAaD,MAAM,CAAC,MAAM,0BAA0B,GAAwB;IAC9D,QAAQ,EAAE,EAAE;IACZ,kBAAkB,EAAE,EAAE;IACtB,6BAA6B,EAAE,IAAI;CACnC,CAAA;AAMD,MAAM,CAAC,MAAM,sCAAsC,GAAoC;IACtF,mCAAmC,EAAE,IAAI;CACzC,CAAA"}
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"WordSequence.d.ts","sourceRoot":"","sources":["../../src/exports/WordSequence.ts"],"names":[],"mappings":"AAAA,qBAAa,YAAY;IACxB,OAAO,EAAE,SAAS,EAAE,CAAK;IAEzB,OAAO,CAAC,IAAI,EAAE,MAAM,EAAE,eAAe,EAAE,MAAM,EAAE,aAAa,EAAE,OAAO,QAUpE;IAED,YAAY,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,YAEhD;IAEA,gBAAgB,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,oCAIrD;IAED,SAAS,CAAC,KAAK,EAAE,MAAM,UAEtB;IAED,aAAa,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,eAEjD;IAEA,iBAAiB,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,uCAItD;IAED,UAAU,CAAC,KAAK,EAAE,MAAM,GAAG,SAAS,CAEnC;IAED,KAAK,iBAEJ;IAED,KAAK,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,gBAMzC;IAED,IAAI,SAAS,aAEZ;IAED,IAAI,SAAS,WAEZ;IAED,IAAI,QAAQ,WAEX;IAED,IAAI,UAAU,cAEb;IAED,IAAI,SAAS,cAEZ;IAED,IAAI,kBAAkB,gBAErB;IAED,IAAI,gBAAgB,aAEnB;IAED,IAAI,qBAAqB,gBAExB;IAED,IAAI,mBAAmB,aAEtB;IAED,IAAI,IAAI,WAEP;IAED,IAAI,MAAM,WAET;CACD;AAED,MAAM,WAAW,SAAS;IACzB,IAAI,EAAE,MAAM,CAAA;IAEZ,WAAW,EAAE,MAAM,CAAA;IACnB,SAAS,EAAE,MAAM,CAAA;IAEjB,aAAa,EAAE,OAAO,CAAA;CACtB"}
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"WordSequence.js","sourceRoot":"","sources":["../../src/exports/WordSequence.ts"],"names":[],"mappings":"AAAA,MAAM,OAAO,YAAY;IACxB,OAAO,GAAgB,EAAE,CAAA;IAEzB,OAAO,CAAC,IAAY,EAAE,eAAuB,EAAE,aAAsB;QACpE,MAAM,WAAW,GAAG,eAAe,CAAA;QACnC,MAAM,SAAS,GAAG,WAAW,GAAG,IAAI,CAAC,MAAM,CAAA;QAE3C,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC;YACjB,IAAI;YACJ,WAAW;YACX,SAAS;YACT,aAAa;SACb,CAAC,CAAA;IACH,CAAC;IAED,YAAY,CAAC,UAAkB,EAAE,QAAgB;QAChD,OAAO,IAAI,CAAC,aAAa,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IACzE,CAAC;IAED,CAAC,gBAAgB,CAAC,UAAkB,EAAE,QAAgB;QACrD,KAAK,IAAI,CAAC,GAAG,UAAU,EAAE,CAAC,GAAG,QAAQ,EAAE,CAAC,EAAE,EAAE,CAAC;YAC5C,MAAM,IAAI,CAAC,SAAS,CAAC,CAAC,CAAC,CAAA;QACxB,CAAC;IACF,CAAC;IAED,SAAS,CAAC,KAAa;QACtB,OAAO,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,EAAE,IAAI,CAAA;IACjC,CAAC;IAED,aAAa,CAAC,UAAkB,EAAE,QAAgB;QACjD,OAAO,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;IAChD,CAAC;IAED,CAAC,iBAAiB,CAAC,UAAkB,EAAE,QAAgB;QACtD,KAAK,IAAI,CAAC,GAAG,UAAU,EAAE,CAAC,GAAG,QAAQ,EAAE,CAAC,EAAE,EAAE,CAAC;YAC5C,MAAM,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,CAAA;QACzB,CAAC;IACF,CAAC;IAED,UAAU,CAAC,KAAa;QACvB,OAAO,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAA;IAC3B,CAAC;IAED,KAAK;QACJ,OAAO,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,MAAM,CAAC,CAAA;IAClC,CAAC;IAED,KAAK,CAAC,UAAkB,EAAE,QAAgB;QACzC,MAAM,cAAc,GAAG,IAAI,YAAY,EAAE,CAAA;QAEzC,cAAc,CAAC,OAAO,GAAG,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;QAEjE,OAAO,cAAc,CAAA;IACtB,CAAC;IAED,IAAI,SAAS;QACZ,OAAO,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IAC7C,CAAC;IAED,IAAI,SAAS;QACZ,OAAO,IAAI,CAAC,UAAU,EAAE,IAAI,CAAA;IAC7B,CAAC;IAED,IAAI,QAAQ;QACX,OAAO,IAAI,CAAC,SAAS,EAAE,IAAI,CAAA;IAC5B,CAAC;IAED,IAAI,UAAU;QACb,OAAO,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAA;IACvB,CAAC;IAED,IAAI,SAAS;QACZ,OAAO,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAA;IACrC,CAAC;IAED,IAAI,kBAAkB;QACrB,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,aAAa,KAAK,IAAI,CAAC,CAAA;IAClE,CAAC;IAED,IAAI,gBAAgB;QACnB,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,aAAa,KAAK,IAAI,CAAC,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IAC3F,CAAC;IAED,IAAI,qBAAqB;QACxB,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,aAAa,KAAK,KAAK,CAAC,CAAA;IACnE,CAAC;IAED,IAAI,mBAAmB;QACtB,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,aAAa,KAAK,KAAK,CAAC,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IAC5F,CAAC;IAED,IAAI,IAAI;QACP,OAAO,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,EAAE,CAAC,CAAA;IAC/B,CAAC;IAED,IAAI,MAAM;QACT,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAA;IAC3B,CAAC;CACD"}
|
|
@@ -1,527 +0,0 @@
|
|
|
1
|
-
import { buildWordOrNumberPattern as buildWordSplitterPattern, phraseSeparatorRegExp, sentenceSeparatorTrailingPunctuationCharacterRegExp, sentenceSeparatorCharacterRegExp, whitespacePatternRegExp, letterPatternGlobalRegExp, startsWithWhitespacePattern, endsWithWhitespacePattern } from '../patterns/Patterns.js'
|
|
2
|
-
import { cldrSuppressions, additionalSuppressions, leadingApostropheContractionSuppressions, nounSuppressions, tldSuppressions } from '../patterns/Suppressions.js'
|
|
3
|
-
import { eastAsianCharRangesRegExp } from '../patterns/EastAsianCharacterPatterns.js'
|
|
4
|
-
import { WordSequence } from './WordSequence.js'
|
|
5
|
-
import { getShortLanguageCode } from '../utilities/Utilities.js'
|
|
6
|
-
export { WordSequence, type WordEntry } from './WordSequence.js'
|
|
7
|
-
|
|
8
|
-
import { buildRegExp } from 'regexp-composer'
|
|
9
|
-
|
|
10
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
11
|
-
// Exported methods
|
|
12
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
13
|
-
export async function segmentText(text: string, options?: SegmentationOptions) {
|
|
14
|
-
const wordSequence = await splitToWords(text, options)
|
|
15
|
-
|
|
16
|
-
return segmentWordSequence(wordSequence)
|
|
17
|
-
}
|
|
18
|
-
|
|
19
|
-
export async function segmentWordSequence(wordSequence: WordSequence, options?: WordSequenceSegmentationOptions) {
|
|
20
|
-
options = { ...defaultWordSequenceSegmentationOptions, ...options }
|
|
21
|
-
|
|
22
|
-
const segmentWordRanges: Range[] = []
|
|
23
|
-
|
|
24
|
-
{
|
|
25
|
-
let segmentStartWordOffset = 0
|
|
26
|
-
let nonWhitespaceWordSeen = false
|
|
27
|
-
let newlineSeenInCurrentSegment = false
|
|
28
|
-
|
|
29
|
-
for (let wordIndex = 0; wordIndex < wordSequence.length; wordIndex++) {
|
|
30
|
-
const word = wordSequence.getWordAt(wordIndex)
|
|
31
|
-
|
|
32
|
-
if (!newlineSeenInCurrentSegment && nonWhitespaceWordSeen && word.endsWith('\n')) {
|
|
33
|
-
newlineSeenInCurrentSegment = true
|
|
34
|
-
}
|
|
35
|
-
|
|
36
|
-
if (newlineSeenInCurrentSegment) {
|
|
37
|
-
const nextWord = wordIndex < wordSequence.length - 1 ? wordSequence.getWordAt(wordIndex + 1) : ''
|
|
38
|
-
|
|
39
|
-
if (nextWord === '' || !whitespacePatternRegExp.test(nextWord)) {
|
|
40
|
-
segmentWordRanges.push({ start: segmentStartWordOffset, end: wordIndex + 1 })
|
|
41
|
-
segmentStartWordOffset = wordIndex + 1
|
|
42
|
-
|
|
43
|
-
newlineSeenInCurrentSegment = false
|
|
44
|
-
}
|
|
45
|
-
}
|
|
46
|
-
|
|
47
|
-
if (nonWhitespaceWordSeen === false) {
|
|
48
|
-
nonWhitespaceWordSeen = !whitespacePatternRegExp.test(word)
|
|
49
|
-
}
|
|
50
|
-
}
|
|
51
|
-
|
|
52
|
-
if (segmentStartWordOffset < wordSequence.length) {
|
|
53
|
-
segmentWordRanges.push({ start: segmentStartWordOffset, end: wordSequence.length })
|
|
54
|
-
}
|
|
55
|
-
}
|
|
56
|
-
|
|
57
|
-
const sentenceWordRanges: Range[] = []
|
|
58
|
-
const segmentSentenceRanges: Range[] = []
|
|
59
|
-
|
|
60
|
-
{
|
|
61
|
-
const minimumSentenceLetterCount = 2
|
|
62
|
-
|
|
63
|
-
for (let segmentIndex = 0; segmentIndex < segmentWordRanges.length; segmentIndex++) {
|
|
64
|
-
const segmentWordRange = segmentWordRanges[segmentIndex]
|
|
65
|
-
|
|
66
|
-
const segmentStartWordIndex = segmentWordRange.start
|
|
67
|
-
const segmentEndWordIndex = segmentWordRange.end
|
|
68
|
-
|
|
69
|
-
segmentSentenceRanges.push({ start: sentenceWordRanges.length, end: sentenceWordRanges.length })
|
|
70
|
-
|
|
71
|
-
let nonWhitespaceWordSeen = false
|
|
72
|
-
let sentenceStartWordOffset = segmentStartWordIndex
|
|
73
|
-
let currentSentenceLetterCount = 0
|
|
74
|
-
|
|
75
|
-
for (let wordIndex = segmentStartWordIndex; wordIndex < segmentEndWordIndex; wordIndex++) {
|
|
76
|
-
const word = wordSequence.getWordAt(wordIndex)
|
|
77
|
-
|
|
78
|
-
if (nonWhitespaceWordSeen === false) {
|
|
79
|
-
nonWhitespaceWordSeen = !whitespacePatternRegExp.test(word)
|
|
80
|
-
}
|
|
81
|
-
|
|
82
|
-
if (currentSentenceLetterCount < minimumSentenceLetterCount) {
|
|
83
|
-
const matches = word.matchAll(letterPatternGlobalRegExp)
|
|
84
|
-
|
|
85
|
-
for (const _ of matches) {
|
|
86
|
-
currentSentenceLetterCount += 1
|
|
87
|
-
|
|
88
|
-
if (currentSentenceLetterCount >= minimumSentenceLetterCount) {
|
|
89
|
-
break
|
|
90
|
-
}
|
|
91
|
-
}
|
|
92
|
-
}
|
|
93
|
-
|
|
94
|
-
if (nonWhitespaceWordSeen && currentSentenceLetterCount >= minimumSentenceLetterCount && sentenceSeparatorCharacterRegExp.test(word)) {
|
|
95
|
-
let trailingSequenceEndIndex = wordIndex
|
|
96
|
-
|
|
97
|
-
while (trailingSequenceEndIndex < segmentEndWordIndex) {
|
|
98
|
-
const trailingWord = wordSequence.getWordAt(trailingSequenceEndIndex)
|
|
99
|
-
|
|
100
|
-
if (sentenceSeparatorTrailingPunctuationCharacterRegExp.test(trailingWord)) {
|
|
101
|
-
trailingSequenceEndIndex++
|
|
102
|
-
} else {
|
|
103
|
-
break
|
|
104
|
-
}
|
|
105
|
-
}
|
|
106
|
-
|
|
107
|
-
sentenceWordRanges.push({
|
|
108
|
-
start: sentenceStartWordOffset,
|
|
109
|
-
end: trailingSequenceEndIndex
|
|
110
|
-
})
|
|
111
|
-
|
|
112
|
-
segmentSentenceRanges[segmentSentenceRanges.length - 1].end += 1
|
|
113
|
-
|
|
114
|
-
sentenceStartWordOffset = trailingSequenceEndIndex
|
|
115
|
-
currentSentenceLetterCount = 0
|
|
116
|
-
|
|
117
|
-
wordIndex = trailingSequenceEndIndex - 1
|
|
118
|
-
}
|
|
119
|
-
}
|
|
120
|
-
|
|
121
|
-
if (sentenceStartWordOffset < segmentEndWordIndex) {
|
|
122
|
-
sentenceWordRanges.push({ start: sentenceStartWordOffset, end: segmentEndWordIndex })
|
|
123
|
-
segmentSentenceRanges[segmentSentenceRanges.length - 1].end += 1
|
|
124
|
-
}
|
|
125
|
-
}
|
|
126
|
-
}
|
|
127
|
-
|
|
128
|
-
const sentences: Sentence[] = []
|
|
129
|
-
|
|
130
|
-
for (const wordRange of sentenceWordRanges) {
|
|
131
|
-
const sentenceWordSequence = wordSequence.slice(wordRange.start, wordRange.end)
|
|
132
|
-
|
|
133
|
-
sentences.push(new Sentence(wordRange, sentenceWordSequence))
|
|
134
|
-
}
|
|
135
|
-
|
|
136
|
-
for (const sentence of sentences) {
|
|
137
|
-
const phraseWordRanges: Range[] = []
|
|
138
|
-
|
|
139
|
-
let sentenceEndWordOffset = sentence.wordRange.end
|
|
140
|
-
let phraseStartWordOffset = sentence.wordRange.start
|
|
141
|
-
|
|
142
|
-
for (let wordIndex = phraseStartWordOffset; wordIndex < sentenceEndWordOffset; wordIndex++) {
|
|
143
|
-
const currentWord = wordSequence.getWordAt(wordIndex)
|
|
144
|
-
|
|
145
|
-
const isCurrentWordPhraseSeparator =
|
|
146
|
-
phraseSeparatorRegExp.test(currentWord) &&
|
|
147
|
-
(!options.requireSpaceAfterColonsOrSemicolons ||
|
|
148
|
-
(currentWord !== ':' && currentWord !== ';') ||
|
|
149
|
-
whitespacePatternRegExp.test(wordSequence.getWordAt(wordIndex + 1)))
|
|
150
|
-
|
|
151
|
-
if (isCurrentWordPhraseSeparator) {
|
|
152
|
-
let whitespaceSeenOnce = false
|
|
153
|
-
|
|
154
|
-
while (wordIndex < sentenceEndWordOffset - 1) {
|
|
155
|
-
const nextWord = wordSequence.getWordAt(wordIndex + 1)
|
|
156
|
-
|
|
157
|
-
if (!sentenceSeparatorTrailingPunctuationCharacterRegExp.test(nextWord)) {
|
|
158
|
-
break
|
|
159
|
-
}
|
|
160
|
-
|
|
161
|
-
if (whitespacePatternRegExp.test(nextWord)) {
|
|
162
|
-
whitespaceSeenOnce = true
|
|
163
|
-
}
|
|
164
|
-
|
|
165
|
-
if (nextWord === '"' && whitespaceSeenOnce) {
|
|
166
|
-
break
|
|
167
|
-
}
|
|
168
|
-
|
|
169
|
-
wordIndex += 1
|
|
170
|
-
}
|
|
171
|
-
|
|
172
|
-
phraseWordRanges.push({
|
|
173
|
-
start: phraseStartWordOffset,
|
|
174
|
-
end: wordIndex + 1
|
|
175
|
-
})
|
|
176
|
-
|
|
177
|
-
phraseStartWordOffset = wordIndex + 1
|
|
178
|
-
}
|
|
179
|
-
}
|
|
180
|
-
|
|
181
|
-
if (phraseStartWordOffset < sentenceEndWordOffset) {
|
|
182
|
-
phraseWordRanges.push({ start: phraseStartWordOffset, end: sentenceEndWordOffset })
|
|
183
|
-
}
|
|
184
|
-
|
|
185
|
-
for (const wordRange of phraseWordRanges) {
|
|
186
|
-
const phraseWordSequence = wordSequence.slice(wordRange.start, wordRange.end)
|
|
187
|
-
|
|
188
|
-
sentence.phrases.push(new Phrase(wordRange, phraseWordSequence))
|
|
189
|
-
}
|
|
190
|
-
}
|
|
191
|
-
|
|
192
|
-
const result: SegmentationResult = {
|
|
193
|
-
words: wordSequence,
|
|
194
|
-
sentences,
|
|
195
|
-
segmentSentenceRanges,
|
|
196
|
-
}
|
|
197
|
-
|
|
198
|
-
return result
|
|
199
|
-
}
|
|
200
|
-
|
|
201
|
-
const cachedWordSplitterRegExps = new Map<string, RegExp>()
|
|
202
|
-
|
|
203
|
-
export async function splitToWords(text: string, options?: SegmentationOptions) {
|
|
204
|
-
if (!options) {
|
|
205
|
-
options = {}
|
|
206
|
-
}
|
|
207
|
-
|
|
208
|
-
options = { ...defaultSegmentationOptions, ...options }
|
|
209
|
-
|
|
210
|
-
if (options.language) {
|
|
211
|
-
options.language = getShortLanguageCode(options.language)
|
|
212
|
-
}
|
|
213
|
-
|
|
214
|
-
const optionsAsJson = JSON.stringify(options)
|
|
215
|
-
|
|
216
|
-
let wordSplitterRegExp = cachedWordSplitterRegExps.get(optionsAsJson)
|
|
217
|
-
|
|
218
|
-
if (!wordSplitterRegExp) {
|
|
219
|
-
wordSplitterRegExp = buildWordSplitterRegExpForOptions(options)
|
|
220
|
-
|
|
221
|
-
cachedWordSplitterRegExps.set(optionsAsJson, wordSplitterRegExp)
|
|
222
|
-
}
|
|
223
|
-
|
|
224
|
-
let wordSequence = new WordSequence()
|
|
225
|
-
|
|
226
|
-
function addPunctuationWordsBetween(startOffset: number, endOffset: number) {
|
|
227
|
-
const punctuationWordSubstring = text.substring(startOffset, endOffset)
|
|
228
|
-
|
|
229
|
-
let charOffset = startOffset
|
|
230
|
-
let punctuationWordStartOffset = startOffset
|
|
231
|
-
|
|
232
|
-
function addPunctuationWordIfNeeded() {
|
|
233
|
-
if (charOffset > punctuationWordStartOffset) {
|
|
234
|
-
const wordText = text.substring(punctuationWordStartOffset, charOffset)
|
|
235
|
-
wordSequence.addWord(wordText, punctuationWordStartOffset, true)
|
|
236
|
-
|
|
237
|
-
punctuationWordStartOffset = charOffset
|
|
238
|
-
}
|
|
239
|
-
}
|
|
240
|
-
|
|
241
|
-
for (const char of punctuationWordSubstring) {
|
|
242
|
-
if (char === ' ') {
|
|
243
|
-
charOffset += 1
|
|
244
|
-
|
|
245
|
-
continue
|
|
246
|
-
}
|
|
247
|
-
|
|
248
|
-
addPunctuationWordIfNeeded()
|
|
249
|
-
|
|
250
|
-
charOffset += char.length
|
|
251
|
-
|
|
252
|
-
addPunctuationWordIfNeeded()
|
|
253
|
-
}
|
|
254
|
-
|
|
255
|
-
addPunctuationWordIfNeeded()
|
|
256
|
-
}
|
|
257
|
-
|
|
258
|
-
const wordMatches = text.matchAll(wordSplitterRegExp)
|
|
259
|
-
|
|
260
|
-
let lastMatchEndOffset = 0
|
|
261
|
-
|
|
262
|
-
if (wordMatches) {
|
|
263
|
-
for (const match of wordMatches) {
|
|
264
|
-
const offsets = match.indices![0]!
|
|
265
|
-
const matchStartOffset = offsets[0]
|
|
266
|
-
const matchEndOffset = offsets[1]
|
|
267
|
-
|
|
268
|
-
if (matchStartOffset > lastMatchEndOffset) {
|
|
269
|
-
addPunctuationWordsBetween(lastMatchEndOffset, matchStartOffset)
|
|
270
|
-
}
|
|
271
|
-
|
|
272
|
-
const wordText = text.substring(matchStartOffset, matchEndOffset)
|
|
273
|
-
wordSequence.addWord(wordText, matchStartOffset, false)
|
|
274
|
-
|
|
275
|
-
lastMatchEndOffset = matchEndOffset
|
|
276
|
-
}
|
|
277
|
-
|
|
278
|
-
addPunctuationWordsBetween(lastMatchEndOffset, text.length)
|
|
279
|
-
}
|
|
280
|
-
|
|
281
|
-
if (options.enableEastAsianPostprocessing) {
|
|
282
|
-
wordSequence = await postprocessEastAsianWords(text, wordSequence)
|
|
283
|
-
}
|
|
284
|
-
|
|
285
|
-
return wordSequence
|
|
286
|
-
}
|
|
287
|
-
|
|
288
|
-
async function postprocessEastAsianWords(containingText: string, wordSequence: WordSequence) {
|
|
289
|
-
const icuSegmentation = await getIcuSegmentation()
|
|
290
|
-
|
|
291
|
-
if (icuSegmentation === undefined) {
|
|
292
|
-
return wordSequence
|
|
293
|
-
}
|
|
294
|
-
|
|
295
|
-
let icuInitialized = false
|
|
296
|
-
|
|
297
|
-
const newWordSequence = new WordSequence()
|
|
298
|
-
|
|
299
|
-
for (let wordIndex = 0; wordIndex < wordSequence.length; wordIndex++) {
|
|
300
|
-
const wordEntry = wordSequence.entries[wordIndex]
|
|
301
|
-
const wordStartOffset = wordEntry.startOffset
|
|
302
|
-
|
|
303
|
-
const word = wordSequence.getWordAt(wordIndex)
|
|
304
|
-
|
|
305
|
-
if (eastAsianCharRangesRegExp.test(word)) {
|
|
306
|
-
if (!icuInitialized) {
|
|
307
|
-
await icuSegmentation.initialize()
|
|
308
|
-
|
|
309
|
-
icuInitialized = true
|
|
310
|
-
}
|
|
311
|
-
|
|
312
|
-
const wordBreaks = [...icuSegmentation.createWordBreakIterator(word)]
|
|
313
|
-
|
|
314
|
-
for (let i = 0; i < wordBreaks.length - 1; i++) {
|
|
315
|
-
const subwordStartOffset = wordStartOffset + wordBreaks[i]
|
|
316
|
-
const subwordEndOffset = wordStartOffset + wordBreaks[i + 1]
|
|
317
|
-
|
|
318
|
-
const subwordText = containingText.substring(subwordStartOffset, subwordEndOffset)
|
|
319
|
-
|
|
320
|
-
newWordSequence.addWord(
|
|
321
|
-
subwordText,
|
|
322
|
-
subwordStartOffset,
|
|
323
|
-
false,
|
|
324
|
-
)
|
|
325
|
-
}
|
|
326
|
-
} else {
|
|
327
|
-
const wordText = containingText.substring(wordEntry.startOffset, wordEntry.endOffset)
|
|
328
|
-
|
|
329
|
-
newWordSequence.addWord(wordText, wordEntry.startOffset, wordEntry.isPunctuation)
|
|
330
|
-
}
|
|
331
|
-
}
|
|
332
|
-
|
|
333
|
-
return newWordSequence
|
|
334
|
-
}
|
|
335
|
-
|
|
336
|
-
// Add any missing punctuation words to a word sequence
|
|
337
|
-
export function addMissingPunctuationWordsToWordSequence(wordSequence: WordSequence, sourceText: string) {
|
|
338
|
-
const originalWordsReverseMapping = new Map<number, number>()
|
|
339
|
-
|
|
340
|
-
const wordSequenceWithPunctuation = new WordSequence()
|
|
341
|
-
|
|
342
|
-
function addWordEntriesForTextSlice(textSlice: string, initialCharOffset: number) {
|
|
343
|
-
let charOffset = initialCharOffset
|
|
344
|
-
|
|
345
|
-
// Add entry for every codepoint (this will correctly treat characters beyond BMP)
|
|
346
|
-
for (const char of textSlice) {
|
|
347
|
-
const charEndOffset = charOffset + char.length
|
|
348
|
-
|
|
349
|
-
const lastEntry = wordSequenceWithPunctuation.lastEntry
|
|
350
|
-
|
|
351
|
-
if (char === ' ' && lastEntry && lastEntry.isPunctuation && lastEntry.text[0] === ' ') {
|
|
352
|
-
wordSequenceWithPunctuation.lastEntry.text += ' '
|
|
353
|
-
wordSequenceWithPunctuation.lastEntry.endOffset = charEndOffset
|
|
354
|
-
} else {
|
|
355
|
-
const wordText = sourceText.substring(charOffset, charEndOffset)
|
|
356
|
-
|
|
357
|
-
wordSequenceWithPunctuation.addWord(wordText, charOffset, true)
|
|
358
|
-
}
|
|
359
|
-
|
|
360
|
-
charOffset = charEndOffset
|
|
361
|
-
}
|
|
362
|
-
}
|
|
363
|
-
|
|
364
|
-
for (let wordIndex = 0; wordIndex < wordSequence.length; wordIndex++) {
|
|
365
|
-
const wordEntry = wordSequence.getEntryAt(wordIndex)
|
|
366
|
-
const wordStartOffset = wordEntry.startOffset
|
|
367
|
-
|
|
368
|
-
const previousWordEndOffset = wordIndex > 0 ? wordSequence.entries[wordIndex - 1].endOffset : 0
|
|
369
|
-
|
|
370
|
-
// Add entries for any punctuation characters between the current and previous word (or start)
|
|
371
|
-
if (previousWordEndOffset !== wordStartOffset) {
|
|
372
|
-
const textSlice = sourceText.substring(previousWordEndOffset, wordStartOffset)
|
|
373
|
-
|
|
374
|
-
addWordEntriesForTextSlice(textSlice, previousWordEndOffset)
|
|
375
|
-
}
|
|
376
|
-
|
|
377
|
-
wordSequenceWithPunctuation.entries.push(wordEntry)
|
|
378
|
-
originalWordsReverseMapping.set(wordSequenceWithPunctuation.length - 1, wordIndex)
|
|
379
|
-
|
|
380
|
-
// If last word, add entries for any trailing punctuation characters
|
|
381
|
-
if (wordIndex === wordSequence.length - 1) {
|
|
382
|
-
if (sourceText.length !== wordEntry.endOffset) {
|
|
383
|
-
const textSlice = sourceText.substring(wordEntry.endOffset, sourceText.length)
|
|
384
|
-
|
|
385
|
-
addWordEntriesForTextSlice(textSlice, wordEntry.endOffset)
|
|
386
|
-
}
|
|
387
|
-
}
|
|
388
|
-
}
|
|
389
|
-
|
|
390
|
-
return { wordSequenceWithPunctuation, originalWordsReverseMapping }
|
|
391
|
-
}
|
|
392
|
-
|
|
393
|
-
function getPunctuationRanges(wordSequence: WordSequence, text: string) {
|
|
394
|
-
const punctuationRanges: Range[] = []
|
|
395
|
-
const wordEntries = wordSequence.entries
|
|
396
|
-
|
|
397
|
-
if (wordEntries[0].startOffset > 0) {
|
|
398
|
-
punctuationRanges.push({ start: 0, end: wordEntries[0].startOffset })
|
|
399
|
-
}
|
|
400
|
-
|
|
401
|
-
for (let i = 0; i < wordEntries.length; i++) {
|
|
402
|
-
const entry = wordEntries[i]
|
|
403
|
-
|
|
404
|
-
const previousEndOffset = wordEntries[i - 1]?.endOffset ?? 0
|
|
405
|
-
|
|
406
|
-
if (entry.startOffset > previousEndOffset) {
|
|
407
|
-
punctuationRanges.push({ start: previousEndOffset, end: entry.startOffset })
|
|
408
|
-
}
|
|
409
|
-
|
|
410
|
-
if (entry.isPunctuation) {
|
|
411
|
-
punctuationRanges.push({ start: entry.startOffset, end: entry.endOffset })
|
|
412
|
-
}
|
|
413
|
-
}
|
|
414
|
-
|
|
415
|
-
{
|
|
416
|
-
const lastEndOffset = wordEntries[wordEntries.length - 1]?.endOffset
|
|
417
|
-
|
|
418
|
-
if (lastEndOffset && lastEndOffset < text.length) {
|
|
419
|
-
punctuationRanges.push({ start: lastEndOffset, end: text.length })
|
|
420
|
-
}
|
|
421
|
-
}
|
|
422
|
-
|
|
423
|
-
return punctuationRanges
|
|
424
|
-
}
|
|
425
|
-
|
|
426
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
427
|
-
// Helper methods
|
|
428
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
429
|
-
function buildWordSplitterRegExpForOptions(options: SegmentationOptions) {
|
|
430
|
-
const cldrSuppressionsForLang = cldrSuppressions[options.language ?? ''] ?? []
|
|
431
|
-
const extendedSuppressionsForLang = additionalSuppressions[options.language ?? ''] ?? []
|
|
432
|
-
const contractionSuppressionsForLang = leadingApostropheContractionSuppressions[options.language ?? ''] ?? []
|
|
433
|
-
const contractionSuppressionsForLangWithSingleQuote = contractionSuppressionsForLang.map(str => str.replaceAll(`'`, `’`))
|
|
434
|
-
const customSuppressions = options.customSuppressions ?? []
|
|
435
|
-
|
|
436
|
-
let suppressions = [
|
|
437
|
-
...customSuppressions,
|
|
438
|
-
...cldrSuppressionsForLang,
|
|
439
|
-
...extendedSuppressionsForLang,
|
|
440
|
-
...contractionSuppressionsForLang,
|
|
441
|
-
...contractionSuppressionsForLangWithSingleQuote,
|
|
442
|
-
...nounSuppressions,
|
|
443
|
-
...tldSuppressions,
|
|
444
|
-
]
|
|
445
|
-
|
|
446
|
-
const wordPattern = buildWordSplitterPattern([
|
|
447
|
-
...suppressions,
|
|
448
|
-
...suppressions.map(word => word.toLocaleLowerCase()),
|
|
449
|
-
...suppressions.map(word => word.toLocaleUpperCase()),
|
|
450
|
-
])
|
|
451
|
-
|
|
452
|
-
const wordSplitterRegExp = buildRegExp(wordPattern, { global: true })
|
|
453
|
-
|
|
454
|
-
return wordSplitterRegExp
|
|
455
|
-
}
|
|
456
|
-
|
|
457
|
-
async function getIcuSegmentation() {
|
|
458
|
-
try {
|
|
459
|
-
const icuSegmentation = await import('@echogarden/icu-segmentation-wasm')
|
|
460
|
-
|
|
461
|
-
return icuSegmentation
|
|
462
|
-
} catch {
|
|
463
|
-
return undefined
|
|
464
|
-
}
|
|
465
|
-
}
|
|
466
|
-
|
|
467
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
468
|
-
// Types
|
|
469
|
-
////////////////////////////////////////////////////////////////////////////////////////////////
|
|
470
|
-
export interface SegmentationResult {
|
|
471
|
-
words: WordSequence
|
|
472
|
-
sentences: Sentence[]
|
|
473
|
-
segmentSentenceRanges: Range[]
|
|
474
|
-
}
|
|
475
|
-
|
|
476
|
-
export class TextFragment {
|
|
477
|
-
wordRange: Range
|
|
478
|
-
words: WordSequence
|
|
479
|
-
|
|
480
|
-
constructor(wordRange: Range, words: WordSequence) {
|
|
481
|
-
this.wordRange = wordRange
|
|
482
|
-
this.words = words
|
|
483
|
-
}
|
|
484
|
-
|
|
485
|
-
get text() {
|
|
486
|
-
return this.words.text
|
|
487
|
-
}
|
|
488
|
-
|
|
489
|
-
get charRange(): Range {
|
|
490
|
-
return {
|
|
491
|
-
start: this.words.firstEntry.startOffset,
|
|
492
|
-
end: this.words.lastEntry.endOffset
|
|
493
|
-
}
|
|
494
|
-
}
|
|
495
|
-
}
|
|
496
|
-
|
|
497
|
-
export class Sentence extends TextFragment {
|
|
498
|
-
phrases: Phrase[] = []
|
|
499
|
-
}
|
|
500
|
-
|
|
501
|
-
export class Phrase extends TextFragment {
|
|
502
|
-
}
|
|
503
|
-
|
|
504
|
-
export interface Range {
|
|
505
|
-
start: number
|
|
506
|
-
end: number
|
|
507
|
-
}
|
|
508
|
-
|
|
509
|
-
export interface SegmentationOptions {
|
|
510
|
-
language?: string
|
|
511
|
-
customSuppressions?: string[]
|
|
512
|
-
enableEastAsianPostprocessing?: boolean
|
|
513
|
-
}
|
|
514
|
-
|
|
515
|
-
export const defaultSegmentationOptions: SegmentationOptions = {
|
|
516
|
-
language: '',
|
|
517
|
-
customSuppressions: [],
|
|
518
|
-
enableEastAsianPostprocessing: true,
|
|
519
|
-
}
|
|
520
|
-
|
|
521
|
-
export interface WordSequenceSegmentationOptions {
|
|
522
|
-
requireSpaceAfterColonsOrSemicolons: boolean
|
|
523
|
-
}
|
|
524
|
-
|
|
525
|
-
export const defaultWordSequenceSegmentationOptions: WordSequenceSegmentationOptions = {
|
|
526
|
-
requireSpaceAfterColonsOrSemicolons: true
|
|
527
|
-
}
|
|
File without changes
|
|
File without changes
|
|
File without changes
|