diff --git a/assets/bg-utf8.txt b/assets/bg-utf8.txt
index 2f3120ec..02d1ded0 100644
--- a/assets/bg-utf8.txt
+++ b/assets/bg-utf8.txt
@@ -230042,7 +230042,6 @@
 домопритежателка
 домостроителство
 домоуправителите
-донякогадоникъде
 дореволюционната
 дореволюционните
 дореволюционното
diff --git a/assets/en-utf8.txt b/assets/en-utf8.txt
index 5dad8233..a25a5698 100644
--- a/assets/en-utf8.txt
+++ b/assets/en-utf8.txt
@@ -15658,6 +15658,7 @@ lester
 lester's
 lestrade
 lestrade's
+let's
 leta
 leta's
 letha
@@ -93569,6 +93570,7 @@ isthmus's
 isthmuses
 istle
 it'd
+it's
 it'll
 itacolumite
 italicization
diff --git a/src/io/github/sspanak/tt9/ime/InputModeValidator.java b/src/io/github/sspanak/tt9/ime/InputModeValidator.java
index c129587a..c12d502e 100644
--- a/src/io/github/sspanak/tt9/ime/InputModeValidator.java
+++ b/src/io/github/sspanak/tt9/ime/InputModeValidator.java
@@ -58,18 +58,12 @@ public class InputModeValidator {
 		return newMode;
 	}
 
-	public static int validateTextCase(T9Preferences prefs, int textCase, ArrayList<Integer> allowedTextCases) {
-		if (allowedTextCases.size() > 0 && allowedTextCases.contains(textCase)) {
-			return textCase;
+	public static void validateTextCase(T9Preferences prefs, InputMode inputMode, int newTextCase) {
+		if (!inputMode.setTextCase(newTextCase)) {
+			inputMode.defaultTextCase();
+			Logger.w("tt9/validateTextCase", "Invalid text case: " + newTextCase + " Enforcing: " + inputMode.getTextCase());
 		}
 
-		int newCase = allowedTextCases.size() > 0 ? allowedTextCases.get(0) : InputMode.CASE_LOWER;
-		prefs.saveTextCase(newCase);
-
-		if (textCase != newCase) {
-			Logger.w("tt9/validateTextCase", "Invalid text case: " + textCase + " Enforcing: " + newCase);
-		}
-
-		return newCase;
+		prefs.saveTextCase(inputMode.getTextCase());
 	}
 }
diff --git a/src/io/github/sspanak/tt9/ime/TraditionalT9.java b/src/io/github/sspanak/tt9/ime/TraditionalT9.java
index 79478d84..17ffde38 100644
--- a/src/io/github/sspanak/tt9/ime/TraditionalT9.java
+++ b/src/io/github/sspanak/tt9/ime/TraditionalT9.java
@@ -25,10 +25,6 @@ public class TraditionalT9 extends KeyPadHandler {
 	private ArrayList<Integer> allowedInputModes = new ArrayList<>();
 	private InputMode mInputMode;
 
-	// text case
-	private ArrayList<Integer> allowedTextCases = new ArrayList<>();
-	private int mTextCase = InputMode.CASE_LOWER;
-
 	// language
 	protected ArrayList<Integer> mEnabledLanguages;
 	protected Language mLanguage;
@@ -45,21 +41,20 @@ public class TraditionalT9 extends KeyPadHandler {
 	private void loadPreferences() {
 		mLanguage = LanguageCollection.getLanguage(prefs.getInputLanguage());
 		mEnabledLanguages = prefs.getEnabledLanguages();
+		validateLanguages();
+
 		mInputMode = InputMode.getInstance(prefs.getInputMode());
-		mTextCase = prefs.getTextCase();
+		mInputMode = InputModeValidator.validateMode(prefs, mInputMode, allowedInputModes);
+
+		InputModeValidator.validateTextCase(prefs, mInputMode, prefs.getTextCase());
 	}
 
+
 	private void validateLanguages() {
 		mEnabledLanguages = InputModeValidator.validateEnabledLanguages(prefs, mEnabledLanguages);
 		mLanguage = InputModeValidator.validateLanguage(prefs, mLanguage, mEnabledLanguages);
 	}
 
-	private void validatePreferences() {
-		validateLanguages();
-		mInputMode = InputModeValidator.validateMode(prefs, mInputMode, allowedInputModes);
-		mTextCase = InputModeValidator.validateTextCase(prefs, mTextCase, allowedTextCases);
-	}
-
 
 	protected void onInit() {
 		self = this;
@@ -74,19 +69,14 @@ public class TraditionalT9 extends KeyPadHandler {
 
 
 	protected void onRestart(EditorInfo inputField) {
-		determineNextTextCase();
-
 		// determine the valid state for the current input field and preferences
-		determineAllowedInputModes(inputField);
-		determineAllowedTextCases();
 		mEnabledLanguages = prefs.getEnabledLanguages(); // in case we are back from Preferences screen, update the language list
-
-		// enforce a valid initial state
-		validatePreferences();
-		clearSuggestions();
+		determineAllowedInputModes(inputField);
+		determineNextTextCase(); // Only in some modes. If they support it, let's overwrite the default.
 
 		// build the UI
-		UI.updateStatusIcon(this, mLanguage, mInputMode, mTextCase);
+		clearSuggestions();
+		UI.updateStatusIcon(this, mLanguage, mInputMode);
 		softKeyHandler.show();
 		if (!isInputViewShown()) {
 			showWindow(true);
@@ -163,7 +153,7 @@ public class TraditionalT9 extends KeyPadHandler {
 
 	protected boolean onLeft() {
 		if (mInputMode.clearWordStem()) {
-			mInputMode.getSuggestionsAsync(handleSuggestionsAsync, mLanguage, getComposingText());
+			mInputMode.loadSuggestions(handleSuggestionsAsync, mLanguage, getComposingText());
 		} else {
 			jumpBeforeComposingText();
 		}
@@ -175,7 +165,7 @@ public class TraditionalT9 extends KeyPadHandler {
 		String filter = repeat ? mSuggestionView.getSuggestion(1) : getComposingText();
 
 		if (mInputMode.setWordStem(mLanguage, filter, repeat)) {
-			mInputMode.getSuggestionsAsync(handleSuggestionsAsync, mLanguage, filter);
+			mInputMode.loadSuggestions(handleSuggestionsAsync, mLanguage, filter);
 		} else if (filter.length() == 0) {
 			mInputMode.reset();
 		}
@@ -333,14 +323,14 @@ public class TraditionalT9 extends KeyPadHandler {
 
 
 	private void getSuggestions() {
-		if (!mInputMode.getSuggestionsAsync(handleSuggestionsAsync, mLanguage, mSuggestionView.getCurrentSuggestion())) {
-			handleSuggestions(mInputMode.getSuggestions());
+		if (!mInputMode.loadSuggestions(handleSuggestionsAsync, mLanguage, mSuggestionView.getCurrentSuggestion())) {
+			handleSuggestions();
 		}
 	}
 
 
-	private void handleSuggestions(ArrayList<String> suggestions) {
-		setSuggestions(suggestions);
+	private void handleSuggestions() {
+		setSuggestions(mInputMode.getSuggestions(mLanguage));
 
 		// Put the first suggestion in the text field,
 		// but cut it off to the length of the sequence (how many keys were pressed),
@@ -353,21 +343,24 @@ public class TraditionalT9 extends KeyPadHandler {
 
 	private final Handler handleSuggestionsAsync = new Handler(Looper.getMainLooper()) {
 		@Override
-		public void handleMessage(Message msg) {
-			handleSuggestions(msg.getData().getStringArrayList("suggestions"));
+		public void handleMessage(Message m) {
+			handleSuggestions();
 		}
 	};
 
 
 	private void setSuggestions(List<String> suggestions) {
+		setSuggestions(suggestions, 0);
+	}
+
+	private void setSuggestions(List<String> suggestions, int selectedIndex) {
 		if (mSuggestionView == null) {
 			return;
 		}
 
 		boolean show = suggestions != null && suggestions.size() > 0;
 
-		mSuggestionView.setSuggestions(suggestions, 0);
-		mSuggestionView.changeCase(mTextCase, mLanguage.getLocale());
+		mSuggestionView.setSuggestions(suggestions, selectedIndex);
 		setCandidatesViewShown(show);
 	}
 
@@ -410,29 +403,27 @@ public class TraditionalT9 extends KeyPadHandler {
 		}
 		// when typing a word or viewing scrolling the suggestions, only change the case
 		else if (!isSuggestionViewHidden()) {
-			determineAllowedTextCases();
+			mInputMode.nextTextCase();
+			ArrayList<String> switchedSuggestions = mInputMode.getSuggestions(mLanguage);
 
-			int modeIndex = (allowedTextCases.indexOf(mTextCase) + 1) % allowedTextCases.size();
-			mTextCase = allowedTextCases.get(modeIndex);
-
-			mSuggestionView.changeCase(mTextCase, mLanguage.getLocale());
+			setSuggestions(switchedSuggestions, mSuggestionView.getCurrentIndex());
 			setComposingText(getComposingText()); // no mistake, this forces the new text case
 		}
 		// make "abc" and "ABC" separate modes from user perspective
-		else if (mInputMode.isABC() && mTextCase == InputMode.CASE_LOWER) {
-			mTextCase = InputMode.CASE_UPPER;
+		else if (mInputMode.isABC() && mInputMode.getTextCase() == InputMode.CASE_LOWER) {
+			mInputMode.setTextCase(InputMode.CASE_UPPER);
 		} else {
 			int modeIndex = (allowedInputModes.indexOf(mInputMode.getId()) + 1) % allowedInputModes.size();
 			mInputMode = InputMode.getInstance(allowedInputModes.get(modeIndex));
 
-			mTextCase = mInputMode.isPredictive() ? InputMode.CASE_CAPITALIZE : InputMode.CASE_LOWER;
+			mInputMode.defaultTextCase();
 		}
 
 		// save the settings for the next time
 		prefs.saveInputMode(mInputMode);
-		prefs.saveTextCase(mTextCase);
+		prefs.saveTextCase(mInputMode.getTextCase());
 
-		UI.updateStatusIcon(this, mLanguage, mInputMode, mTextCase);
+		UI.updateStatusIcon(this, mLanguage, mInputMode);
 	}
 
 
@@ -453,7 +444,7 @@ public class TraditionalT9 extends KeyPadHandler {
 		// save it for the next time
 		prefs.saveInputLanguage(mLanguage.getId());
 
-		UI.updateStatusIcon(this, mLanguage, mInputMode, mTextCase);
+		UI.updateStatusIcon(this, mLanguage, mInputMode);
 	}
 
 
@@ -490,20 +481,11 @@ public class TraditionalT9 extends KeyPadHandler {
 	}
 
 
-	private void determineAllowedTextCases() {
-		allowedTextCases = mInputMode.getAllowedTextCases();
-		// @todo: determine the text case of the input and validate using the allowed ones [ https://github.com/sspanak/tt9/issues/48 ]
-	}
-
-
 	private void determineNextTextCase() {
-		int nextTextCase = mInputMode.getNextWordTextCase(
-			mTextCase,
+		mInputMode.determineNextWordTextCase(
 			InputFieldHelper.isThereText(currentInputConnection),
 			(String) currentInputConnection.getTextBeforeCursor(50, 0)
 		);
-
-		mTextCase = nextTextCase != -1 ? nextTextCase : mTextCase;
 	}
 
 
diff --git a/src/io/github/sspanak/tt9/ime/modes/InputMode.java b/src/io/github/sspanak/tt9/ime/modes/InputMode.java
index 326a5c7d..686ce02d 100644
--- a/src/io/github/sspanak/tt9/ime/modes/InputMode.java
+++ b/src/io/github/sspanak/tt9/ime/modes/InputMode.java
@@ -1,8 +1,6 @@
 package io.github.sspanak.tt9.ime.modes;
 
-import android.os.Bundle;
 import android.os.Handler;
-import android.os.Message;
 
 import java.util.ArrayList;
 
@@ -19,7 +17,9 @@ abstract public class InputMode {
 	public static final int CASE_UPPER = 0;
 	public static final int CASE_CAPITALIZE = 1;
 	public static final int CASE_LOWER = 2;
+	public static final int CASE_DICTIONARY = 3; // do not force it, but use the dictionary word as-is
 	protected ArrayList<Integer> allowedTextCases = new ArrayList<>();
+	protected int textCase = CASE_LOWER;
 
 	// data
 	protected ArrayList<String> suggestions = new ArrayList<>();
@@ -45,14 +45,16 @@ abstract public class InputMode {
 
 	// Suggestions
 	public void onAcceptSuggestion(Language language, String suggestion) {}
-	public ArrayList<String> getSuggestions() { return suggestions; }
-	public boolean getSuggestionsAsync(Handler handler, Language language, String lastWord) { return false; }
-	protected void sendSuggestions(Handler handler, ArrayList<String> suggestions) {
-		Bundle bundle = new Bundle();
-		bundle.putStringArrayList("suggestions", suggestions);
-		Message msg = new Message();
-		msg.setData(bundle);
-		handler.sendMessage(msg);
+	protected void onSuggestionsUpdated(Handler handler) { handler.sendEmptyMessage(0); }
+	public boolean loadSuggestions(Handler handler, Language language, String lastWord) { return false; }
+
+	public ArrayList<String> getSuggestions(Language language) {
+		ArrayList<String> newSuggestions = new ArrayList<>();
+		for (String s : suggestions) {
+			newSuggestions.add(adjustSuggestionTextCase(s, textCase, language));
+		}
+
+		return newSuggestions;
 	}
 
 	// Word
@@ -65,15 +67,38 @@ abstract public class InputMode {
 
 	// Utility
 	abstract public int getId();
-	public ArrayList<Integer> getAllowedTextCases() { return allowedTextCases; }
-	// Perform any special logic to determine the text case of the next word, or return "-1" if there is no need to change it.
-	public int getNextWordTextCase(int currentTextCase, boolean isThereText, String textBeforeCursor) { return -1; }
 	abstract public int getSequenceLength(); // The number of key presses for the current word.
 	public void reset() {
 		suggestions = new ArrayList<>();
 		word = null;
 	}
 
+	// Text case
+	public int getTextCase() { return textCase; }
+
+	public boolean setTextCase(int newTextCase) {
+		if (!allowedTextCases.contains(newTextCase)) {
+			return false;
+		}
+
+		textCase = newTextCase;
+		return true;
+	}
+
+	public void defaultTextCase() {
+		textCase = allowedTextCases.get(0);
+	}
+
+	public void nextTextCase() {
+		int nextIndex = (allowedTextCases.indexOf(textCase) + 1) % allowedTextCases.size();
+		textCase = allowedTextCases.get(nextIndex);
+	}
+
+	public void determineNextWordTextCase(boolean isThereText, String textBeforeCursor) {}
+
+	// Based on the internal logic of the mode (punctuation or grammar rules), re-adjust the text case for when getSuggestions() is called.
+	protected String adjustSuggestionTextCase(String word, int newTextCase, Language language) { return word; }
+
 	// Stem filtering.
 	// Where applicable, return "true" if the mode supports it and the operation was possible.
 	public boolean clearWordStem() { return false; }
diff --git a/src/io/github/sspanak/tt9/ime/modes/ModeABC.java b/src/io/github/sspanak/tt9/ime/modes/ModeABC.java
index 7d75059e..c024446d 100644
--- a/src/io/github/sspanak/tt9/ime/modes/ModeABC.java
+++ b/src/io/github/sspanak/tt9/ime/modes/ModeABC.java
@@ -1,5 +1,7 @@
 package io.github.sspanak.tt9.ime.modes;
 
+import java.util.ArrayList;
+
 import io.github.sspanak.tt9.languages.Language;
 
 public class ModeABC extends InputMode {
@@ -19,10 +21,10 @@ public class ModeABC extends InputMode {
 		word = null;
 
 		if (hold) {
-			suggestions = null;
+			suggestions = new ArrayList<>();
 			word = String.valueOf(key);
 		} else if (repeat) {
-			suggestions = null;
+			suggestions = new ArrayList<>();
 			shouldSelectNextLetter = true;
 		}
 
@@ -30,6 +32,11 @@ public class ModeABC extends InputMode {
 	}
 
 
+	protected String adjustSuggestionTextCase(String word, int newTextCase, Language language) {
+		return newTextCase == CASE_UPPER ? word.toUpperCase(language.getLocale()) : word.toLowerCase(language.getLocale());
+	}
+
+
 	final public boolean isABC() { return true; }
 	public int getSequenceLength() { return 1; }
 
diff --git a/src/io/github/sspanak/tt9/ime/modes/ModePredictive.java b/src/io/github/sspanak/tt9/ime/modes/ModePredictive.java
index 106d44f6..7dff60a4 100644
--- a/src/io/github/sspanak/tt9/ime/modes/ModePredictive.java
+++ b/src/io/github/sspanak/tt9/ime/modes/ModePredictive.java
@@ -5,7 +5,6 @@ import android.os.Looper;
 import android.os.Message;
 
 import java.util.ArrayList;
-import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
 import io.github.sspanak.tt9.Logger;
@@ -29,11 +28,14 @@ public class ModePredictive extends InputMode {
 	private String lastInputFieldWord = "";
 	private static Handler handleSuggestionsExternal;
 
+	// auto text case selection
+	private final Pattern endOfSentence = Pattern.compile("(?<!\\.)[.?!]\\s*$");
+
 
 	ModePredictive() {
-		allowedTextCases.add(CASE_UPPER);
-		allowedTextCases.add(CASE_CAPITALIZE);
 		allowedTextCases.add(CASE_LOWER);
+		allowedTextCases.add(CASE_CAPITALIZE);
+		allowedTextCases.add(CASE_UPPER);
 	}
 
 
@@ -88,16 +90,6 @@ public class ModePredictive extends InputMode {
 	}
 
 
-	final public boolean isPredictive() {
-		return true;
-	}
-
-	public int getSequenceLength() { return isEmoji ? 2 : digitSequence.length(); }
-
-	public boolean shouldTrackUpDown() { return true; }
-	public boolean shouldTrackLeftRight() { return true; }
-
-
 	/**
 	 * shouldAcceptCurrentSuggestion
 	 * In this mode, In addition to confirming the suggestion in the input field,
@@ -166,20 +158,21 @@ public class ModePredictive extends InputMode {
 
 
 	/**
-	 * getSuggestionsAsync
+	 * loadSuggestions
 	 * Queries the dictionary database for a list of suggestions matching the current language and
 	 * sequence. Returns "false" when there is nothing to do.
 	 *
 	 * "lastWord" is used for generating suggestions when there are no results.
 	 * See: generatePossibleCompletions()
 	 */
-	public boolean getSuggestionsAsync(Handler handler, Language language, String lastWord) {
+	public boolean loadSuggestions(Handler handler, Language language, String lastWord) {
 		if (isEmoji) {
-			super.sendSuggestions(handler, suggestions);
+			super.onSuggestionsUpdated(handler);
 			return true;
 		}
 
 		if (digitSequence.length() == 0) {
+			suggestions.clear();
 			return false;
 		}
 
@@ -209,17 +202,19 @@ public class ModePredictive extends InputMode {
 	private final Handler handleSuggestions = new Handler(Looper.getMainLooper()) {
 		@Override
 		public void handleMessage(Message msg) {
-			ArrayList<String> suggestions = msg.getData().getStringArrayList("suggestions");
-			suggestions = suggestions == null ? new ArrayList<>() : suggestions;
+			ArrayList<String> dbSuggestions = msg.getData().getStringArrayList("suggestions");
+			dbSuggestions = dbSuggestions == null ? new ArrayList<>() : dbSuggestions;
 
-			if (suggestions.size() == 0 && digitSequence.length() > 0) {
-				suggestions = generatePossibleCompletions(currentLanguage, lastInputFieldWord);
+			if (dbSuggestions.size() == 0 && digitSequence.length() > 0) {
+				dbSuggestions = generatePossibleCompletions(currentLanguage, lastInputFieldWord);
 			}
 
-			ArrayList<String> stemVariations = generatePossibleStemVariations(currentLanguage, suggestions);
-			stemVariations.addAll(suggestions);
+			suggestions.clear();
+			suggestStem();
+			suggestions.addAll(generatePossibleStemVariations(currentLanguage, dbSuggestions));
+			suggestMoreWords(dbSuggestions);
 
-			ModePredictive.super.sendSuggestions(handleSuggestionsExternal, stemVariations);
+			ModePredictive.super.onSuggestionsUpdated(handleSuggestionsExternal);
 		}
 	};
 
@@ -269,22 +264,28 @@ public class ModePredictive extends InputMode {
 	 *
 	 * For example, if the filter is "extr", the current word is "extr_" and the user has pressed "1",
 	 * the database would have returned only "extra", but this function would also
-	 * generate: "extrb" and "extrc". This is useful for typing an unknown word, similar to the ones
-	 * in the dictionary.
+	 * generate: "extrb" and "extrc". This is useful for typing an unknown word, that is similar to
+	 * the ones in the dictionary.
 	 */
-	private ArrayList<String> generatePossibleStemVariations(Language language, ArrayList<String> currentSuggestions) {
+	private ArrayList<String> generatePossibleStemVariations(Language language, ArrayList<String> dbSuggestions) {
 		ArrayList<String> variations = new ArrayList<>();
 		if (stem.length() == 0) {
 			return variations;
 		}
 
-		if (stem.length() == digitSequence.length() && !currentSuggestions.contains(stem)) {
-			variations.add(stem);
-		}
-
 		if (isStemFuzzy && stem.length() == digitSequence.length() - 1) {
-			for (String word : generatePossibleCompletions(language, stem)) {
-				if (!currentSuggestions.contains(word)) {
+			ArrayList<String> allPossibleVariations = generatePossibleCompletions(language, stem);
+
+			// first add the known words, because it makes more sense to see them first
+			for (String word : allPossibleVariations) {
+				if (dbSuggestions.contains(word)) {
+					variations.add(word);
+				}
+			}
+
+			// then add the unknown ones, so they can be used as possible beginnings of new words.
+			for (String word : allPossibleVariations) {
+				if (!dbSuggestions.contains(word)) {
 					variations.add(word);
 				}
 			}
@@ -293,6 +294,30 @@ public class ModePredictive extends InputMode {
 		return variations;
 	}
 
+	/**
+	 * suggestStem
+	 * Add the current stem filter to the suggestion list, when it has length of X and
+	 * the user has pressed X keys.
+	 */
+	public void suggestStem() {
+		if (stem.length() > 0 && stem.length() == digitSequence.length()) {
+			suggestions.add(stem);
+		}
+	}
+
+
+	/**
+	 * suggestMoreWords
+	 * Takes a list of words and appends them to the suggestion list, if they are missing.
+	 */
+	public void suggestMoreWords(ArrayList<String> newSuggestions) {
+		for (String word : newSuggestions) {
+			if (!suggestions.contains(word)) {
+				suggestions.add(word);
+			}
+		}
+	}
+
 
 	/**
 	 * onAcceptSuggestion
@@ -316,28 +341,61 @@ public class ModePredictive extends InputMode {
 
 
 	/**
-	 * getNextWordTextCase
-	 * Dynamically determine text case of words as the user types to reduce key presses.
+	 * adjustSuggestionTextCase
+	 * In addition to uppercase/lowercase, here we use the result from determineNextWordTextCase(),
+	 * to conveniently start sentences with capitals or whatnot.
+	 *
+	 * Also, by default we preserve any  mixed case words in the dictionary,
+	 * for example: "dB", "Mb", proper names, German nouns, that always start with a capital,
+	 * or Dutch words such as: "'s-Hertogenbosch".
+	 */
+	protected String adjustSuggestionTextCase(String word, int newTextCase, Language language) {
+		switch (newTextCase) {
+			case CASE_UPPER:
+				return word.toUpperCase(language.getLocale());
+			case CASE_LOWER:
+				return word.toLowerCase(language.getLocale());
+			case CASE_CAPITALIZE:
+				return language.isMixedCaseWord(word) ? word : language.capitalize(word);
+			case CASE_DICTIONARY:
+				return language.isMixedCaseWord(word) ? word : word.toLowerCase(language.getLocale());
+			default:
+				return word;
+		}
+	}
+
+
+	/**
+	 * determineNextWordTextCase
+	 * Dynamically determine text case of words as the user types, to reduce key presses.
 	 * For example, this function will return CASE_LOWER by default, but CASE_UPPER at the beginning
 	 * of a sentence.
 	 */
-	public int getNextWordTextCase(int currentTextCase, boolean isThereText, String textBeforeCursor) {
+	public void determineNextWordTextCase(boolean isThereText, String textBeforeCursor) {
 		// If the user wants to type in uppercase, this must be for a reason, so we better not override it.
-		if (currentTextCase == CASE_UPPER) {
-			return -1;
+		if (textCase == CASE_UPPER) {
+			return;
 		}
 
 		// start of text
 		if (!isThereText) {
-			return CASE_CAPITALIZE;
+			textCase = CASE_CAPITALIZE;
+			return;
 		}
 
 		// start of sentence, excluding after "..."
-		Matcher endOfSentenceMatch = Pattern.compile("(?<!\\.)[.?!]\\s*$").matcher(textBeforeCursor);
-		if (endOfSentenceMatch.find()) {
-			return CASE_CAPITALIZE;
+		if (endOfSentence.matcher(textBeforeCursor).find()) {
+			textCase = CASE_CAPITALIZE;
+			return;
 		}
 
-		return CASE_LOWER;
+		textCase = CASE_DICTIONARY;
 	}
+
+
+	final public boolean isPredictive() { return true; }
+	public int getSequenceLength() { return isEmoji ? 2 : digitSequence.length(); }
+	public boolean shouldTrackUpDown() { return true; }
+	public boolean shouldTrackLeftRight() { return true; }
+
 }
diff --git a/src/io/github/sspanak/tt9/languages/Language.java b/src/io/github/sspanak/tt9/languages/Language.java
index 8a2d4498..a9182fa2 100644
--- a/src/io/github/sspanak/tt9/languages/Language.java
+++ b/src/io/github/sspanak/tt9/languages/Language.java
@@ -8,13 +8,15 @@ public class Language {
 	protected int id;
 	protected String name;
 	protected Locale locale;
-	protected boolean isPunctuationPartOfWords; // see the getter for more info
 	protected int icon;
 	protected String dictionaryFile;
 	protected int abcLowerCaseIcon;
 	protected int abcUpperCaseIcon;
 	protected ArrayList<ArrayList<String>> characterMap = new ArrayList<>();
 
+	// settings
+	protected boolean isPunctuationPartOfWords; // see the getter for more info
+
 	final public int getId() {
 		return id;
 	}
@@ -31,6 +33,15 @@ public class Language {
 		return icon;
 	}
 
+	final public String getDictionaryFile() {
+		return dictionaryFile;
+	}
+
+	final public int getAbcIcon(boolean lowerCase) {
+		return lowerCase ? abcLowerCaseIcon : abcUpperCaseIcon;
+	}
+
+
 	/**
 	 * isPunctuationPartOfWords
 	 * This plays a role in Predictive mode only.
@@ -49,12 +60,15 @@ public class Language {
 	 */
 	final public boolean isPunctuationPartOfWords() { return isPunctuationPartOfWords; }
 
-	final public String getDictionaryFile() {
-		return dictionaryFile;
+
+	/************* utility *************/
+
+	public String capitalize(String word) {
+		return word != null ? word.substring(0, 1).toUpperCase(locale) + word.substring(1).toLowerCase(locale) : null;
 	}
 
-	final public int getAbcIcon(boolean lowerCase) {
-		return lowerCase ? abcLowerCaseIcon : abcUpperCaseIcon;
+	public boolean isMixedCaseWord(String word) {
+		return word != null && !word.toLowerCase(locale).equals(word) && !word.toUpperCase(locale).equals(word);
 	}
 
 	public ArrayList<String> getKeyCharacters(int key) {
diff --git a/src/io/github/sspanak/tt9/languages/definitions/Bulgarian.java b/src/io/github/sspanak/tt9/languages/definitions/Bulgarian.java
index 1a043e9a..8224bb66 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/Bulgarian.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/Bulgarian.java
@@ -14,11 +14,12 @@ public class Bulgarian extends Language {
 		name = "български";
 		locale = new Locale("bg","BG");
 		dictionaryFile = "bg-utf8.txt";
-		isPunctuationPartOfWords = false;
 		icon = R.drawable.ime_lang_bg;
 		abcLowerCaseIcon = R.drawable.ime_lang_cyrillic_lower;
 		abcUpperCaseIcon = R.drawable.ime_lang_cyrillic_upper;
 
+		isPunctuationPartOfWords = false;
+
 		characterMap = new ArrayList<>(Arrays.asList(
 			Punctuation.Secondary, // 0
 			Punctuation.Main, // 1
diff --git a/src/io/github/sspanak/tt9/languages/definitions/Dutch.java b/src/io/github/sspanak/tt9/languages/definitions/Dutch.java
index b8fe5f83..1fd81dc6 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/Dutch.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/Dutch.java
@@ -12,10 +12,11 @@ public class Dutch extends English {
 		id = 8;
 		name = "Nederlands";
 		locale = new Locale("nl","NL");
-		isPunctuationPartOfWords = true;
 		dictionaryFile = "nl-utf8.txt";
 		icon = R.drawable.ime_lang_nl;
 
+		isPunctuationPartOfWords = true;
+
 		characterMap.get(2).addAll(Arrays.asList("à", "ä", "ç"));
 		characterMap.get(3).addAll(Arrays.asList("é", "è", "ê", "ë"));
 		characterMap.get(4).addAll(Arrays.asList("î", "ï"));
diff --git a/src/io/github/sspanak/tt9/languages/definitions/English.java b/src/io/github/sspanak/tt9/languages/definitions/English.java
index 17d96964..1a46e955 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/English.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/English.java
@@ -14,11 +14,12 @@ public class English extends Language {
 		name = "English";
 		locale = Locale.ENGLISH;
 		dictionaryFile = "en-utf8.txt";
-		isPunctuationPartOfWords = true;
 		icon = R.drawable.ime_lang_en;
 		abcLowerCaseIcon = R.drawable.ime_lang_latin_lower;
 		abcUpperCaseIcon = R.drawable.ime_lang_latin_upper;
 
+		isPunctuationPartOfWords = true;
+
 		characterMap = new ArrayList<>(Arrays.asList(
 			Punctuation.Secondary, // 0
 			Punctuation.Main, // 1
diff --git a/src/io/github/sspanak/tt9/languages/definitions/French.java b/src/io/github/sspanak/tt9/languages/definitions/French.java
index 93e11aac..844d672f 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/French.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/French.java
@@ -14,6 +14,7 @@ public class French extends English {
 		locale = Locale.FRENCH;
 		dictionaryFile = "fr-utf8.txt";
 		icon = R.drawable.ime_lang_fr;
+
 		isPunctuationPartOfWords = false;
 
 		characterMap.get(2).addAll(Arrays.asList("à", "â", "æ", "ç"));
diff --git a/src/io/github/sspanak/tt9/languages/definitions/German.java b/src/io/github/sspanak/tt9/languages/definitions/German.java
index 3a05f1fe..ec52221b 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/German.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/German.java
@@ -13,6 +13,7 @@ public class German extends English {
 		locale = Locale.GERMAN;
 		dictionaryFile = "de-utf8.txt";
 		icon = R.drawable.ime_lang_de;
+
 		isPunctuationPartOfWords = false;
 
 		characterMap.get(2).add("ä");
diff --git a/src/io/github/sspanak/tt9/languages/definitions/Italian.java b/src/io/github/sspanak/tt9/languages/definitions/Italian.java
index cf882d46..c5181c49 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/Italian.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/Italian.java
@@ -14,6 +14,7 @@ public class Italian extends English {
 		locale = Locale.ITALIAN;
 		dictionaryFile = "it-utf8.txt";
 		icon = R.drawable.ime_lang_it;
+
 		isPunctuationPartOfWords = false;
 
 		characterMap.get(2).add("à");
diff --git a/src/io/github/sspanak/tt9/languages/definitions/Russian.java b/src/io/github/sspanak/tt9/languages/definitions/Russian.java
index 0bded11b..c039720c 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/Russian.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/Russian.java
@@ -14,11 +14,12 @@ public class Russian extends Language {
 		name = "русский";
 		locale = new Locale("ru","RU");
 		dictionaryFile = "ru-utf8.txt";
-		isPunctuationPartOfWords = false;
 		icon = R.drawable.ime_lang_ru;
 		abcLowerCaseIcon = R.drawable.ime_lang_cyrillic_lower;
 		abcUpperCaseIcon = R.drawable.ime_lang_cyrillic_upper;
 
+		isPunctuationPartOfWords = false;
+
 		characterMap = new ArrayList<>(Arrays.asList(
 			Punctuation.Secondary, // 0
 			Punctuation.Main, // 1
diff --git a/src/io/github/sspanak/tt9/languages/definitions/Ukrainian.java b/src/io/github/sspanak/tt9/languages/definitions/Ukrainian.java
index a6a4ff2b..0ce93175 100644
--- a/src/io/github/sspanak/tt9/languages/definitions/Ukrainian.java
+++ b/src/io/github/sspanak/tt9/languages/definitions/Ukrainian.java
@@ -14,11 +14,12 @@ public class Ukrainian extends Language {
 		name = "українська";
 		locale = new Locale("uk","UA");
 		dictionaryFile = "uk-utf8.txt";
-		isPunctuationPartOfWords = true;
 		icon = R.drawable.ime_lang_uk;
 		abcLowerCaseIcon = R.drawable.ime_lang_cyrillic_lower;
 		abcUpperCaseIcon = R.drawable.ime_lang_cyrillic_upper;
 
+		isPunctuationPartOfWords = true;
+
 		characterMap = new ArrayList<>(Arrays.asList(
 			Punctuation.Secondary, // 0
 			Punctuation.Main, // 1
diff --git a/src/io/github/sspanak/tt9/ui/CandidateView.java b/src/io/github/sspanak/tt9/ui/CandidateView.java
index 258fc069..d3a17996 100644
--- a/src/io/github/sspanak/tt9/ui/CandidateView.java
+++ b/src/io/github/sspanak/tt9/ui/CandidateView.java
@@ -10,14 +10,12 @@ import android.view.View;
 
 import java.util.ArrayList;
 import java.util.List;
-import java.util.Locale;
 
 import io.github.sspanak.tt9.R;
-import io.github.sspanak.tt9.ime.modes.InputMode;
 
 public class CandidateView extends View {
 
-	private List<String> mSuggestions;
+	private List<String> mSuggestions = new ArrayList<>();
 	protected int mSelectedIndex;
 
 	private Drawable mSelectionHighlight;
@@ -32,7 +30,7 @@ public class CandidateView extends View {
 
 	private static final int X_GAP = 10;
 
-	private static final List<String> EMPTY_LIST = new ArrayList<String>();
+	private static final List<String> EMPTY_LIST = new ArrayList<>();
 
 	private int mColorNormal;
 	private int mColorRecommended;
@@ -47,8 +45,6 @@ public class CandidateView extends View {
 
 	/**
 	 * Construct a CandidateView for showing suggested words for completion.
-	 *
-	 * @param context
 	 */
 	public CandidateView(Context context) {
 		super(context);
@@ -204,23 +200,6 @@ public class CandidateView extends View {
 		requestLayout();
 	}
 
-	public void changeCase(int textCase, Locale locale) {
-		ArrayList<String> newSuggestions = new ArrayList<>();
-
-		for (String s : mSuggestions) {
-			if (textCase == InputMode.CASE_LOWER) {
-				newSuggestions.add(s.toLowerCase(locale));
-			} else if (textCase == InputMode.CASE_CAPITALIZE) {
-				String cs = s.substring(0, 1).toUpperCase(locale) + s.substring(1).toLowerCase(locale);
-				newSuggestions.add(cs);
-			} else {
-				newSuggestions.add(s.toUpperCase(locale));
-			}
-		}
-
-		setSuggestions(newSuggestions, mSelectedIndex);
-	}
-
 	protected void clear() {
 		mSuggestions = EMPTY_LIST;
 		mSelectedIndex = -1;
diff --git a/src/io/github/sspanak/tt9/ui/UI.java b/src/io/github/sspanak/tt9/ui/UI.java
index cc7ff2bc..e3476ba2 100644
--- a/src/io/github/sspanak/tt9/ui/UI.java
+++ b/src/io/github/sspanak/tt9/ui/UI.java
@@ -35,9 +35,9 @@ public class UI {
 	 * Set the status icon that is appropriate in current mode (based on
 	 * openwmm-legacy)
 	 */
-	public static void updateStatusIcon(TraditionalT9 tt9, Language inputLanguage, InputMode inputMode, int textCase) {
+	public static void updateStatusIcon(TraditionalT9 tt9, Language inputLanguage, InputMode inputMode) {
 		if (inputMode.isABC()) {
-			tt9.showStatusIcon(inputLanguage.getAbcIcon(textCase == InputMode.CASE_LOWER));
+			tt9.showStatusIcon(inputLanguage.getAbcIcon(inputMode.getTextCase() == InputMode.CASE_LOWER));
 		} else if (inputMode.isPredictive()) {
 			tt9.showStatusIcon(inputLanguage.getIcon());
 		} else if (inputMode.is123()) {