I'm creating a filter in Java that filters out words based on skills.

The skill I'm working on now is call Phonetic Skill #1.  The rule is that a vowel is short if it is followed by a single consonant sound.  I'm using the following regex to determin this.

^(y??(\w*?[\w&&[^aeiouy]]+?)*?)([aeiouy])(ch|sh|wh|th|ph|gn|kn|ck|wr|[\w&&[^aeiouyrw]])([\w&&[^aeiou]]+?[aeiouy]+?\w*?)*?$

 This works except for one exception (that I've found so far anyway).  Blends (br, cr, sc, sk, bl, pl, etc.) in multi-syllable words (zebra).  Typically consonants in multi-syllable words split (zeb-ra) but blends (br) stay together (ze-bra).  So, my regex catches this word and says it works but it really shouldn't.  How do I modify my regex to factor in the blend exceptions?

 Sorry for the lesson in Phonics, but thanks for any help.

Well I'm not 100% clear on what you are trying to match but this won't match zebra

Original Expression ^(y??(\w*?[\w&&[^aeiouy]]+?)*?)([aeiouy](?!br|cr|sc|sk|bl|pl))(ch|sh|wh|th|ph|gn|kn|ck|wr|[\w&&[^aeiouyrw]])([\w&&[^aeiou]]+?[aeiouy]+?\w*?)*?$

as a Java string "^(y??(\\w*?[\\w&&[^aeiouy]]+?)*?)([aeiouy](?!br|cr|sc|sk|bl|pl))(ch|sh|wh|th|ph|gn|kn|ck|wr|[\\w&&[^aeiouyrw]])([\\w&&[^aeiou]]+?[aeiouy]+?\\w*?)*?$"

---

Michael

"In theory, theory and practice are the same. In practice, they are not."
Albert Einstein

Wow, it worked.

If it's not evident by now I'm not very experienced in regular expressions.  Can you explain what the added statement does?

Let me give some examples of what I'm doing.  If you have any suggestions on improving my regex I'd appreciate it.

ad : true, because it has a vowel followed by a single consonant - [aeiouy][\w&&[^aeiouyrw]]

add: false, because it has two consonants - $

ban: true, single consonant after vowel, any consonant can come before the vowel - (\w*?[\w&&[^aeiouy]]+?)*?)

band: false, two consonants

bath: true, th is a digraph and makes a single sound - [aeiouy](ch|sh|wh|th|ph|gn|kn|ck|wr|[\w&&[^aeiouyrw]]) (the r & w are for words like war (ar) and bow (ow) these are another whole different case)

mad: true

maid: false, can't have adjacent vowels - (\w*?[\w&&[^aeiouy]]+?)*?[aeiouy]

yap: true, y is a consonant at the begining of a word - (y??(\w*?[\w&&[^aeiouy]]+?)*?)

When you get into multi-syllable words you have to take each syllable individually

diner: false, di-ner single consonant goes with next syllable

dinner: true, din-ner consonants split then use the above rules on each syllable -  ([\\w&&[^aeiou]]+?[aeiouy]+?\\w*?)*? (tacked on to disregard any following syllables)

zebra: false, ze-bra typically consonants split, but blends (br) stay together and go with next syllable

 

Let me know if I can explain the rules or parts of my regex better.

Buddha24:

Wow, it worked.

If it's not evident by now I'm not very experienced in regular expressions.  Can you explain what the added statement does?

 The construction I added to your pattern is called a negative lookahead. (?! pattern goes here) , which basically says continue trying to complete the match is the negative lookahead's patten is not found further ahead in the string. A more detailed explaination can be found here http://www.regular-expressions.info/lookaround.html

I simply added it after the character class of vowels with the values you listed as blends.

---

Michael

"In theory, theory and practice are the same. In practice, they are not."
Albert Einstein

Thanks again!

Buddha24:

Let me give some examples of what I'm doing.  If you have any suggestions on improving my regex I'd appreciate it.

ad : true, because it has a vowel followed by a single consonant - [aeiouy][\w&&[^aeiouyrw]]

add: false, because it has two consonants - $

ban: true, single consonant after vowel, any consonant can come before the vowel - (\w*?[\w&&[^aeiouy]]+?)*?)

band: false, two consonants

bath: true, th is a digraph and makes a single sound - [aeiouy](ch|sh|wh|th|ph|gn|kn|ck|wr|[\w&&[^aeiouyrw]]) (the r & w are for words like war (ar) and bow (ow) these are another whole different case)

mad: true

maid: false, can't have adjacent vowels - (\w*?[\w&&[^aeiouy]]+?)*?[aeiouy]

yap: true, y is a consonant at the begining of a word - (y??(\w*?[\w&&[^aeiouy]]+?)*?)

When you get into multi-syllable words you have to take each syllable individually

diner: false, di-ner single consonant goes with next syllable

dinner: true, din-ner consonants split then use the above rules on each syllable -  ([\\w&&[^aeiou]]+?[aeiouy]+?\\w*?)*? (tacked on to disregard any following syllables)

zebra: false, ze-bra typically consonants split, but blends (br) stay together and go with next syllable

 

Let me know if I can explain the rules or parts of my regex better.

While I'm sure the rule of the English language are too complex to handle purely with a regex this pattern handles all the test cases you provided. And since you are just dealing English words I replace \w with a-z in the character classes. \w includes digits, the underscore and other Unicode alphas which don't seem to be relevant to your task. 

^(?:(?:[a-z&&[aeiou]]+|\b)([aeiou](?!br|cr|sc|sk|bl|pl))(?:(?:ch|sh|wh|th|ph|gn|kn|ck|wr)|(?:[a-z&&[aeiouy]](?![aeiouy]))))*\s*$

Regular Expression
Original Expression	^(?:(?:[a-z&&[^aeiou]]+|\b)([aeiou](?!br|cr|sc|sk|bl|pl))(?:(?:ch|sh|wh|th|ph|gn|kn|ck|wr)|(?:[a-z&&[^aeiouy]](?![aeiouy]))))*\s*$
as a Java string	"^(?:(?:[a-z&&[^aeiou]]+|\\b)([aeiou](?!br|cr|sc|sk|bl|pl))(?:(?:ch|sh|wh|th|ph|gn|kn|ck|wr)|(?:[a-z&&[^aeiouy]](?![aeiouy]))))*\\s*$"
Replacement
groupCount()	1

Test	Target String	matches()	replaceFirst()	replaceAll()	lookingAt()	find()	group(0)	group(1)
1	ad	Yes			Yes	Yes	ad	a
2	add	No	add	add	No	No
3	ban	Yes			Yes	Yes	ban	a
4	band	No	band	band	No	No
5	mad	Yes			Yes	Yes	mad	a
6	maid	No	maid	maid	No	No
7	yap	Yes			Yes	Yes	yap	a
8	diner	No	diner	diner	No	No
9	dinner	Yes			Yes	Yes	dinner	e
10	zebra	No	zebra	zebra	No	No

 It probably doesn't follow all the rules will want to ultimately use but it is at least a starting point.

---

Michael

"In theory, theory and practice are the same. In practice, they are not."
Albert Einstein