How to Correct Text to Speech Pronunciation Errors

Computerized voices will correctly pronounce the majority of words. However there are always exceptions that will be mispronounced. Common examples include brand names and industry-specific terms, especially those that are made up of compound words (e.g. myspace).

Checking and Correcting Pronunciation Before Converting Text to Speech

It would be nice if you could check an article for pronunciation errors before converting it to speech. Text2Go makes this possible. Firstly, you choose the text you want to check. The following 4 options are available.

Text2Go will now do the following to make it easy to identify mispronunciations.

1. Firstly a list of unique words is extracted from the text.
2. Then it's filtered to remove words that have already been corrected and words that are known to be correctly pronounced.
3. The remaining words are then spell-checked.
4. Finally they are sorted, with misspelt words being placed at the top of the list, as these are also the most likely to be mispronounced.

Watch a video of this in action

You will then be presented with the following window that allows you to check and correct words.

Right-click on any of the original words to display a popup menu of available commands.

The first task is to listen to the words you wish to check.

Select the words to listen to. You can either click on individual words (hold down the Ctrl and or Shift keys to select multiple words) or right-click to use one of the selection commands from the menu.

Right-click and select the 'Check' command.

Text2Go will now read each word in turn. Every time you hear a mispronounced word, left-click anyway in the 'Check Pronunciation' window and it will be marked as incorrect. Once all words have been read aloud, you can enter corrections. Words that are correctly pronounced will be marked in green. These will be added to the list of correct words, so they need never be checked again.

Tips

• Wait until the mispronounced word has finished and the next word is being spoken before clicking.
• If you make a mistake and mark the wrong word as mispronounced or miss a mispronounced word, right-click to pause. You can then use the Undo command to go back to the word you missed and resume from that point.
• If words are being spoken too quickly, adjust the speech rate to add a longer delay between words.

Once the words you wish to check have been marked as correct or incorrect, click the apply button. This will add all correct words to the correct word list.

You can now enter corrections for any mispronounced words.

Enter a word in the correction column. Often you will need to try spelling it phonetically. Sometimes it takes a bit of experimentation to get it right. To listen to the correction, right-click on the original word and select the 'Speak Correction' command.

If you have more than one voice, you can use the Voice combo box to choose a voice to listen to the word in.

Tip. Always check your correction sounds correct with all your voices. If a correction only works for a particular voice, make sure you only apply the correction to that voice (read on to see how this is done).

Each correction has a number of attributes that control which voice(s) the correction applies to, the dictionary it's stored in, the type of correction and whether it's applied to the plural form of the original word. These attributes are displayed in the 'Correction Details' panel.

Correction Details

The correction details apply to the currently selected correction. If you select more than one correction, then the details apply to all selected corrections. When the details differ between the selected corrections, those fields will be displayed as blank. If you change a setting, it will be applied to all the selected corrections.

Case Sensitive

• If checked, A case-sensitive comparison is used to match the original word. i.e. 'cat' will only match 'cat', NOT 'Cat', 'CAT', 'CaT', etc. This is useful when matching acronymns.
• If not checked, a case insensitive comparison is used to match the original word. i.e 'cat' will match 'cat', 'Cat', 'CAT', 'CaT', etc.

Match Plurals

This field allows you to automatically generate an entry for the plural form of a word. The following options are available.

Type

Voice(s)

Used to restrict a correction to a particular voice or voice(s). This allows you to correct words that are mispronounced by a particular voice. By default, a correction will be used for all voices. Note that you don't need to specify the entire voice name and case is ignored. You can also specify more than one voice.

e.g. 'karen' will restrict the entry to the voice 'RealSpeak Karen'.

Language Variant

Used to restrict an entry to a particular language variant. Current English language variants include

By default, an entry will match all language variants. To restrict an entry to Australian English voices use

en-AU

To restrict an entry to Indian English voices, due to the fact it's classified as en-GB (UK English), use the following code

ENI

Vendor

Used to restrict an entry to a particular vendor/manufacturer of voices (e.g. RealSpeak). By default, an entry will match voices supplied by all vendors. e.g. To restrict an entry to just RealSpeak voices use

RealSpeak

Dictionary

A 'User Dictionary' is created the first time you go to check some text. By default all corrections will be placed in this dictionary. However it is possible to organise your corrections into multiple dictionaries. This field lets you choose the dictionary to place a new correction in or move an existing correction from one dictionary to another.

Note: The dictionaries that come with Text2Go are read-only. The corrections within them cannot be modified or removed and new entries cannot be added. The reason for this is that these dictionaries are updated regularly and if you modified them you would lose all your changes the next time you received a new version from Text2Go.

Manually Adding Corrections

If you click on the 'Add' button you can enter both the original word and the correction. This allows you to add corrections for mispronunciations that you have you've already listened to.

Sorting the Word List

By default the original word list is sorted by category (Unrecognised, Correctly Spelt, Corrected Pronunciation and Correct Pronunciation) and then alphabetically. It's also possible to sort by word length, frequency and compound words. Sometimes it's quite useful to identify and place compound words at the top of the list, as they are more likely to be mispronounced.

Showing and Hiding Word Categories

By default Correctly Pronounced and Correct Pronunciation word categories are not listed when checking a block of text. However you can use the checkboxes in the Show panel to hide or show each category.

Editing Existing Corrections

You can edit existing corrections stored in the dictionaries using the Edit Corrections... command. This will allow you to change or remove existing corrections or move corrections between dictionaries. It can also be used to remove words that have been incorrectly added to the correct word list (see below).

Deleting a Correction

If you need to delete a correction all you need to do is erase the word in the Correction column. When you hit apply, the correction will be deleted from the dictionary.

To delete a word from the correct word list, first show the 'Correct Pronunciation' word category. Then select the word to remove, right-click and select the 'Reset to Unchecked' command. It will be removed from the correct word list when you apply the changes.

Duplicate Entries

It's quite possible to have more than one entry for the same original word. Text2Go will choose the entry most appropriate for the current voice. Entries that specify the voice will be given priority over entries that specify just a vendor followed by entries that specify a language variant. Entries that don't specify voice, vendor or language variant are given the lowest priority. If there are duplicate entries with the same priority then the entry in the highest dictionary in the list will be used.

Sharing Pronunciation Corrections

Text2Go uses an automatic-update like service to share pronunciation corrections amongst Text2Go users. By default an update is done every 2 days. However this can be changed on the Pronunciation Options tab.

You can change how often you wish to check for updates or turn the service off completely. You can also decide whether you want to share your corrections with others. By default this is enabled, but once again you can turn it off. You can also decide whether you would like to contribute anonymously (the default) or identify your contributions with your email address.

Checking Pronunciation Before Creating an Audio Track

If you would like to be prompted to check pronunciation every time you go to create an audio track, you can enable this from the Pronunciation Options tab. By default this is not enabled.

Creating a New Dictionary

Sometimes you may want to place a certain set of corrections into a separate dictionary. You can create your own dictionary to do this.

Select the Dictionaries... command from the Pronunciation menu.

Click on Add... to create a new dictionary.

Enter some details to identify your dictionary and click OK.

You have now created a new dictionary. You can add new entries to this dictionary or move entries from an existing dictionary into this dictionary.

Managing Dictionaries

 Dictionaries can be enabled, disabled and the order changed from the Text2Go Dictionaries window. This window is available from the Pronunciation -> Dictionaries... menu.

It's also possible to view and edit just the entries that belong to a single dictionary, by selecting the 'Edit Entries' command.