Count the number of sentences in the text

At the risk of incurring the wrath of the whole tape I’ll still write:

And you, friends, no matter how you sit down,
All the musicians are not fit

Well, seriously, you do not need amateur work, but you have to solve the problem correctly.

I see 2 possible ways:

1. Or stick your text in the NLP network and train it. For starters, you can take OpenNLP and try to train it.
2. Take Apache Lucene and try the SentenceTokenizer - which will break the text into sentences. Lucene, as I understand it, supports the Russian language , by the way, there are also external Lucene extenders for Russian morphology

Something like this.

with such various kinds of abbreviations (interim, so-called, IS Turgenev ...). It is necessary to take into account the number of characters between the points, or the number of characters to the point. There are no such short sentences, therefore they can be ignored. You can also lay on proposals like: "Yes.", "No." I do not think that such a short set up a lot.

In any case, the task is useless, IMHO.

Offhand a few moments that would seem to make a simple task unsolvable:

1. Parcelry - it is not clear, this is one sentence or several:

   He went too. To the store. Buy cigarettes. (Shukshin)

2. Direct speech sentences:

   Ignat whispered: "Well, to hell with this task," he laughed nervously.

3. Scanning errors, typos, lack of punctuation marks as in typical Internet communication. When the border is on, the sentences of the sentence are not labeled in any way.

   EVERYTHING WAS EVERYTHING SO JUST EIGHT FIVE MINUTES BACK I HATE

4. Interspersing code:

   To display the username, type echo $ name. ''. $ surname.

This is an open science task.

Here , for example, people are trying to solve it somehow.

There is a library for Python , which copes well with this task (at least for English). If you still need Java, see Jython.

You can use regex:

 Matcher m = Pattern.compile("\\.\s*[A-ZА-Я]").matcher(yourText); //Вместо yourText должна быть переменная с текстом, в котором нужно посчитать предложения int count = 1; while(m.find()) count++; //Будут сосчитаны все предложения начинающиеся с большой буквы 

Here I wrote, in my opinion it was interesting. This code understands this kind of test sentences.

 Привет тест. T..... a Тестирование 34. WHAT??? 

Conclusion

Предложений по моему мнению:4

There is also a debugging fitch for the test, after each sentence it displays the sentence itself. They can be removed, and you can use.

 package javaapplication3; import java.io.ByteArrayOutputStream; public class JavaApplication3 { public static void main(String[] args) { String str = "Привет тест. T..... a Тестирование 34. WHAT???"; byte[] bytes = str.getBytes(); byte tocka = '.'; //заставляем интерпретатор из символа точки получить 1 байт(можно цифру, но так понятнее вам будет) byte vopr = '?'; byte voscl = '!'; int count = 0; //храним количество int max = bytes.length; if(max>0){ int i = 0; //сдвиг byte a; //активный байт //ДЕБАГ, В СЛУЧАЕ ИСПОЛЬЗОВАНИЯ СКРИПТА БЕЗ ДЕБАГА ЗАКОМЕНТИТЬ ОТ И ДО ByteArrayOutputStream out = new ByteArrayOutputStream(); //буфер для предложений //дебаг режим выводит уже предложения //РАЗБИРАЕМ while(i<max){ //ЦИКЛ ОСНОВА while(i<max){ //ЦИКЛ ПРЕРЫВАЕТСЯ ОЗНАЧАЮЩИЙ КОНЕЦ ПРЕДЛОЖЕНИЯ a = bytes[i]; i = i + 1; if(i<max){ //ЕСТЬ СИМВОЛЫ ДЛЯ ДЕБАГИНГА if(a == tocka || a == vopr || a == voscl){ //ПРЕДЛОЖЕНИЕ. //проверяем следующий символ a = bytes[i]; i = i + 1; if(i<max){ if(a == tocka || a == vopr || a == voscl){ //возможно это троеточие, почемубы и нет? или ???:) while(i<max){ a = bytes[i]; if(a!=tocka && a!=vopr && a!=voscl){ break; } i = i + 1; } } } break; } } //ДЕБАГ out.write(a); // } count = count + 1; //ПРЕРЫВАНИЕ БЫЛО ИЛИ ПРЕДЛОЖЕНИЕ ОКОНЧЕНО ЗНАЧИТ ПРЕДЛОЖЕНИЕ //ДЕБАГ, МОЖНО ЗАКОМЕНТИТЬ System.err.println(out.toString()); out.reset(); // } } System.out.println("Предложений по моему мнению:"+String.valueOf(count)); } } 

I managed to solve my question. For those who are interested here is my answer. It is worth noting that the task was for the texts of the German language and cases with abbreviations of the initials of the name (for example, AS Pushkin) I did not need to take into account.

 package ir_ub2; import java.io.BufferedReader; import java.io.FileReader; import java.io.IOException; public class TextCounts { public static void main(String[] args) throws IOException { FileReader file = new FileReader("inputDE.txt"); // datei lesen BufferedReader reader = new BufferedReader(file); int sentenceCount = 0; String line; String delimiter = ".?!"; String[] singlePointExceptions = { "19. Jahrhundert", "allg.", "bzw.", "bspw.", "etc.", "evtl.", "geb.", "ggf.", "n.Chr", "od.", "s.", "u.", "usw.", "vgl." }; String[] doublePointExceptions = { "bw", "dh", "di", "n.Chr.", "sa", "so", "su", "ua", "u.Ä.", "uU", "uz", "va", "v.Chr.", "zB", "zT", "z.Zt." }; while ((line = reader.readLine()) != null) { // Continue reading until end of file is reached int countQuestionsAndExclamations = countMatchesOfSpecialCharacters(line, "?", "!"); int countSingles = countMatchesOfSpecialCharacters(line, singlePointExceptions); int countDoubles = countMatchesOfSpecialCharacters(line, doublePointExceptions); int countPoints = countMatchesOfSpecialCharacters(line, "."); sentenceCount += (countQuestionsAndExclamations + (countPoints - (countSingles + countDoubles))); } reader.close(); file.close(); System.out.println("# of founded Sentences: " + sentenceCount); } private static int countMatchesOfSpecialCharacters(final String str, final String...specialCharacters) { if (null == str || str.isEmpty()) { return 0; } if (null == specialCharacters || specialCharacters.length == 0) { return 0; } int count = 0; int index = 0; for (int i = 0; i < specialCharacters.length; i++) { String special = specialCharacters[i]; index = 0; while ((index = str.indexOf(special, index)) != -1) { count++; index += special.length(); if (index >= str.length()) { break; } } } return count; } } 

You can use the function String.split ("."); The function returns an array of sentences, and with predlidzenja.length() you can get the length of the array.