Predictive Text Engines in Windows Mobile 6.5 Keyboards

Windows Mobile 6.5 relied on a combination of Microsoft's built-in text prediction system, licensed OEM solutions, and advanced third-party engines to deliver auto-correct and predictive text across its soft input panels (SIPs). While Microsoft provided a basic native dictionary-based engine, the vast majority of consumer devices utilized sophisticated commercial engines—most notably Nuance's XT9—or dedicated third-party keyboards to provide modern touch correction, word completion, and next-word prediction.

Microsoft Native Word Completion Engine

Windows Mobile 6.5 included an updated, finger-friendly on-screen keyboard alongside legacy stylus-focused input panels. Powering the default Microsoft keyboard was the built-in Windows Mobile Word Completion engine.

This native system was relatively basic:

  • Dictionary Matching: It matched prefix characters against a static system dictionary and displayed candidate words in a horizontal suggestion bar above the keyboard.
  • User Dictionaries: The engine logged frequently typed words into a local user dictionary to prioritize custom vocabulary over time.
  • Auto-Correction: The stock engine provided rudimentary spelling correction and automatic capitalization, but it lacked contextual language modeling and spatial error tolerance for missed keystrokes.

Nuance XT9 (The OEM Standard)

Because the native Windows Mobile engine was limited, major hardware manufacturers (OEMs) bypassed it in favor of XT9, developed by Nuance Communications (originally created by Tegic Communications). XT9 was the undisputed standard engine powering soft keyboards from brands like HTC, Samsung, and Sony Ericsson.

XT9 served as the underlying predictive backbone for custom keyboard skins, such as HTC’s EzInput (Touch Input) and Samsung’s TouchWiz keyboard:

  • Spatial Error Concealment: XT9 analyzed touch coordinates rather than just discrete key hits. If a user tapped near the border of a letter, the engine calculated the probability of the intended letter based on language modeling.
  • Multi-Layout Support: It handled prediction across full QWERTY, compact QWERTY (two letters per key), and traditional 12-key phone keypads.
  • Next-Word Prediction and Auto-Correction: XT9 dynamically offered the next logical word in a sentence and automatically substituted mistyped words upon pressing the spacebar.

Third-Party Predictive Engines

Power users and enterprise deployments frequently replaced the stock and OEM input methods with third-party soft keyboards, each driven by proprietary prediction algorithms:

  • TouchPal (CooTek): Driven by CooTek's proprietary engine, TouchPal introduced context-aware prediction, mistype compensation, and the patented "T+" layout, which paired high-accuracy word disambiguation with swipe-based input.
  • Swype: Before its acquisition by Nuance, Swype gained massive popularity on Windows Mobile 6.5 via public beta releases. Its engine translated continuous finger-tracing paths across the keyboard into words, using spatial analysis and language frequency dictionaries.
  • Spb Keyboard: Developed by Spb Software, this keyboard used a proprietary smart-prediction engine that adapted to user typing habits, offered multi-language auto-switching, and provided aggressive typo correction.
  • SlideIT (Dasur): SlideIT utilized Dasur’s text input engine, which focused on pattern recognition of continuous gestures combined with predictive vocabulary databases.